text
stringlengths 102
105k
| source
stringlengths 44
152
|
|---|---|
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Image-to-Image Task Guide
[[open-in-colab]]
Image-to-Image task is the task where an application receives an image and outputs another image. This has various subtasks, including image enhancement (super resolution, low light enhancement, deraining and so on), image inpainting, and more.
This guide will show you how to:
- Use an image-to-image pipeline for super resolution task,
- Run image-to-image models for same task without a pipeline.
Note that as of the time this guide is released, `image-to-image` pipeline only supports super resolution task.
Let's begin by installing the necessary libraries.
```bash
pip install transformers
```
We can now initialize the pipeline with a [Swin2SR model](https://huggingface.co/caidas/swin2SR-lightweight-x2-64). We can then infer with the pipeline by calling it with an image. As of now, only [Swin2SR models](https://huggingface.co/models?sort=trending&search=swin2sr) are supported in this pipeline.
```python
from transformers import pipeline
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
pipe = pipeline(task="image-to-image", model="caidas/swin2SR-lightweight-x2-64", device=device)
```
Now, let's load an image.
```python
from PIL import Image
import requests
url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/cat.jpg"
image = Image.open(requests.get(url, stream=True).raw)
print(image.size)
```
```bash
# (532, 432)
```
<div class="flex justify-center">
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/cat.jpg" alt="Photo of a cat"/>
</div>
We can now do inference with the pipeline. We will get an upscaled version of the cat image.
```python
upscaled = pipe(image)
print(upscaled.size)
```
```bash
# (1072, 880)
```
If you wish to do inference yourself with no pipeline, you can use the `Swin2SRForImageSuperResolution` and `Swin2SRImageProcessor` classes of transformers. We will use the same model checkpoint for this. Let's initialize the model and the processor.
```python
from transformers import Swin2SRForImageSuperResolution, Swin2SRImageProcessor
model = Swin2SRForImageSuperResolution.from_pretrained("caidas/swin2SR-lightweight-x2-64").to(device)
processor = Swin2SRImageProcessor("caidas/swin2SR-lightweight-x2-64")
```
`pipeline` abstracts away the preprocessing and postprocessing steps that we have to do ourselves, so let's preprocess the image. We will pass the image to the processor and then move the pixel values to GPU.
```python
pixel_values = processor(image, return_tensors="pt").pixel_values
print(pixel_values.shape)
pixel_values = pixel_values.to(device)
```
We can now infer the image by passing pixel values to the model.
```python
import torch
with torch.no_grad():
outputs = model(pixel_values)
```
Output is an object of type `ImageSuperResolutionOutput` that looks like below 👇
```
(loss=None, reconstruction=tensor([[[[0.8270, 0.8269, 0.8275, ..., 0.7463, 0.7446, 0.7453],
[0.8287, 0.8278, 0.8283, ..., 0.7451, 0.7448, 0.7457],
[0.8280, 0.8273, 0.8269, ..., 0.7447, 0.7446, 0.7452],
...,
[0.5923, 0.5933, 0.5924, ..., 0.0697, 0.0695, 0.0706],
[0.5926, 0.5932, 0.5926, ..., 0.0673, 0.0687, 0.0705],
[0.5927, 0.5914, 0.5922, ..., 0.0664, 0.0694, 0.0718]]]],
device='cuda:0'), hidden_states=None, attentions=None)
```
We need to get the `reconstruction` and post-process it for visualization. Let's see how it looks like.
```python
outputs.reconstruction.data.shape
# torch.Size([1, 3, 880, 1072])
```
We need to squeeze the output and get rid of axis 0, clip the values, then convert it to be numpy float. Then we will arrange axes to have the shape [1072, 880], and finally, bring the output back to range [0, 255].
```python
import numpy as np
# squeeze, take to CPU and clip the values
output = outputs.reconstruction.data.squeeze().cpu().clamp_(0, 1).numpy()
# rearrange the axes
output = np.moveaxis(output, source=0, destination=-1)
# bring values back to pixel values range
output = (output * 255.0).round().astype(np.uint8)
Image.fromarray(output)
```
<div class="flex justify-center">
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/cat_upscaled.png" alt="Upscaled photo of a cat"/>
</div>
| huggingface/transformers/blob/main/docs/source/en/tasks/image_to_image.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Utilities for Tokenizers
This page lists all the utility functions used by the tokenizers, mainly the class
[`~tokenization_utils_base.PreTrainedTokenizerBase`] that implements the common methods between
[`PreTrainedTokenizer`] and [`PreTrainedTokenizerFast`] and the mixin
[`~tokenization_utils_base.SpecialTokensMixin`].
Most of those are only useful if you are studying the code of the tokenizers in the library.
## PreTrainedTokenizerBase
[[autodoc]] tokenization_utils_base.PreTrainedTokenizerBase
- __call__
- all
## SpecialTokensMixin
[[autodoc]] tokenization_utils_base.SpecialTokensMixin
## Enums and namedtuples
[[autodoc]] tokenization_utils_base.TruncationStrategy
[[autodoc]] tokenization_utils_base.CharSpan
[[autodoc]] tokenization_utils_base.TokenSpan
| huggingface/transformers/blob/main/docs/source/en/internal/tokenization_utils.md |
!--Copyright 2021 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# LUKE
## Overview
The LUKE model was proposed in [LUKE: Deep Contextualized Entity Representations with Entity-aware Self-attention](https://arxiv.org/abs/2010.01057) by Ikuya Yamada, Akari Asai, Hiroyuki Shindo, Hideaki Takeda and Yuji Matsumoto.
It is based on RoBERTa and adds entity embeddings as well as an entity-aware self-attention mechanism, which helps
improve performance on various downstream tasks involving reasoning about entities such as named entity recognition,
extractive and cloze-style question answering, entity typing, and relation classification.
The abstract from the paper is the following:
*Entity representations are useful in natural language tasks involving entities. In this paper, we propose new
pretrained contextualized representations of words and entities based on the bidirectional transformer. The proposed
model treats words and entities in a given text as independent tokens, and outputs contextualized representations of
them. Our model is trained using a new pretraining task based on the masked language model of BERT. The task involves
predicting randomly masked words and entities in a large entity-annotated corpus retrieved from Wikipedia. We also
propose an entity-aware self-attention mechanism that is an extension of the self-attention mechanism of the
transformer, and considers the types of tokens (words or entities) when computing attention scores. The proposed model
achieves impressive empirical performance on a wide range of entity-related tasks. In particular, it obtains
state-of-the-art results on five well-known datasets: Open Entity (entity typing), TACRED (relation classification),
CoNLL-2003 (named entity recognition), ReCoRD (cloze-style question answering), and SQuAD 1.1 (extractive question
answering).*
This model was contributed by [ikuyamada](https://huggingface.co/ikuyamada) and [nielsr](https://huggingface.co/nielsr). The original code can be found [here](https://github.com/studio-ousia/luke).
## Usage tips
- This implementation is the same as [`RobertaModel`] with the addition of entity embeddings as well
as an entity-aware self-attention mechanism, which improves performance on tasks involving reasoning about entities.
- LUKE treats entities as input tokens; therefore, it takes `entity_ids`, `entity_attention_mask`,
`entity_token_type_ids` and `entity_position_ids` as extra input. You can obtain those using
[`LukeTokenizer`].
- [`LukeTokenizer`] takes `entities` and `entity_spans` (character-based start and end
positions of the entities in the input text) as extra input. `entities` typically consist of [MASK] entities or
Wikipedia entities. The brief description when inputting these entities are as follows:
- *Inputting [MASK] entities to compute entity representations*: The [MASK] entity is used to mask entities to be
predicted during pretraining. When LUKE receives the [MASK] entity, it tries to predict the original entity by
gathering the information about the entity from the input text. Therefore, the [MASK] entity can be used to address
downstream tasks requiring the information of entities in text such as entity typing, relation classification, and
named entity recognition.
- *Inputting Wikipedia entities to compute knowledge-enhanced token representations*: LUKE learns rich information
(or knowledge) about Wikipedia entities during pretraining and stores the information in its entity embedding. By
using Wikipedia entities as input tokens, LUKE outputs token representations enriched by the information stored in
the embeddings of these entities. This is particularly effective for tasks requiring real-world knowledge, such as
question answering.
- There are three head models for the former use case:
- [`LukeForEntityClassification`], for tasks to classify a single entity in an input text such as
entity typing, e.g. the [Open Entity dataset](https://www.cs.utexas.edu/~eunsol/html_pages/open_entity.html).
This model places a linear head on top of the output entity representation.
- [`LukeForEntityPairClassification`], for tasks to classify the relationship between two entities
such as relation classification, e.g. the [TACRED dataset](https://nlp.stanford.edu/projects/tacred/). This
model places a linear head on top of the concatenated output representation of the pair of given entities.
- [`LukeForEntitySpanClassification`], for tasks to classify the sequence of entity spans, such as
named entity recognition (NER). This model places a linear head on top of the output entity representations. You
can address NER using this model by inputting all possible entity spans in the text to the model.
[`LukeTokenizer`] has a `task` argument, which enables you to easily create an input to these
head models by specifying `task="entity_classification"`, `task="entity_pair_classification"`, or
`task="entity_span_classification"`. Please refer to the example code of each head models.
Usage example:
```python
>>> from transformers import LukeTokenizer, LukeModel, LukeForEntityPairClassification
>>> model = LukeModel.from_pretrained("studio-ousia/luke-base")
>>> tokenizer = LukeTokenizer.from_pretrained("studio-ousia/luke-base")
# Example 1: Computing the contextualized entity representation corresponding to the entity mention "Beyoncé"
>>> text = "Beyoncé lives in Los Angeles."
>>> entity_spans = [(0, 7)] # character-based entity span corresponding to "Beyoncé"
>>> inputs = tokenizer(text, entity_spans=entity_spans, add_prefix_space=True, return_tensors="pt")
>>> outputs = model(**inputs)
>>> word_last_hidden_state = outputs.last_hidden_state
>>> entity_last_hidden_state = outputs.entity_last_hidden_state
# Example 2: Inputting Wikipedia entities to obtain enriched contextualized representations
>>> entities = [
... "Beyoncé",
... "Los Angeles",
... ] # Wikipedia entity titles corresponding to the entity mentions "Beyoncé" and "Los Angeles"
>>> entity_spans = [(0, 7), (17, 28)] # character-based entity spans corresponding to "Beyoncé" and "Los Angeles"
>>> inputs = tokenizer(text, entities=entities, entity_spans=entity_spans, add_prefix_space=True, return_tensors="pt")
>>> outputs = model(**inputs)
>>> word_last_hidden_state = outputs.last_hidden_state
>>> entity_last_hidden_state = outputs.entity_last_hidden_state
# Example 3: Classifying the relationship between two entities using LukeForEntityPairClassification head model
>>> model = LukeForEntityPairClassification.from_pretrained("studio-ousia/luke-large-finetuned-tacred")
>>> tokenizer = LukeTokenizer.from_pretrained("studio-ousia/luke-large-finetuned-tacred")
>>> entity_spans = [(0, 7), (17, 28)] # character-based entity spans corresponding to "Beyoncé" and "Los Angeles"
>>> inputs = tokenizer(text, entity_spans=entity_spans, return_tensors="pt")
>>> outputs = model(**inputs)
>>> logits = outputs.logits
>>> predicted_class_idx = int(logits[0].argmax())
>>> print("Predicted class:", model.config.id2label[predicted_class_idx])
```
## Resources
- [A demo notebook on how to fine-tune [`LukeForEntityPairClassification`] for relation classification](https://github.com/NielsRogge/Transformers-Tutorials/tree/master/LUKE)
- [Notebooks showcasing how you to reproduce the results as reported in the paper with the HuggingFace implementation of LUKE](https://github.com/studio-ousia/luke/tree/master/notebooks)
- [Text classification task guide](../tasks/sequence_classification)
- [Token classification task guide](../tasks/token_classification)
- [Question answering task guide](../tasks/question_answering)
- [Masked language modeling task guide](../tasks/masked_language_modeling)
- [Multiple choice task guide](../tasks/multiple_choice)
## LukeConfig
[[autodoc]] LukeConfig
## LukeTokenizer
[[autodoc]] LukeTokenizer
- __call__
- save_vocabulary
## LukeModel
[[autodoc]] LukeModel
- forward
## LukeForMaskedLM
[[autodoc]] LukeForMaskedLM
- forward
## LukeForEntityClassification
[[autodoc]] LukeForEntityClassification
- forward
## LukeForEntityPairClassification
[[autodoc]] LukeForEntityPairClassification
- forward
## LukeForEntitySpanClassification
[[autodoc]] LukeForEntitySpanClassification
- forward
## LukeForSequenceClassification
[[autodoc]] LukeForSequenceClassification
- forward
## LukeForMultipleChoice
[[autodoc]] LukeForMultipleChoice
- forward
## LukeForTokenClassification
[[autodoc]] LukeForTokenClassification
- forward
## LukeForQuestionAnswering
[[autodoc]] LukeForQuestionAnswering
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/luke.md |
!--Copyright 2022 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Image classification
[[open-in-colab]]
<Youtube id="tjAIM7BOYhw"/>
Image classification assigns a label or class to an image. Unlike text or audio classification, the inputs are the
pixel values that comprise an image. There are many applications for image classification, such as detecting damage
after a natural disaster, monitoring crop health, or helping screen medical images for signs of disease.
This guide illustrates how to:
1. Fine-tune [ViT](model_doc/vit) on the [Food-101](https://huggingface.co/datasets/food101) dataset to classify a food item in an image.
2. Use your fine-tuned model for inference.
<Tip>
The task illustrated in this tutorial is supported by the following model architectures:
<!--This tip is automatically generated by `make fix-copies`, do not fill manually!-->
[BEiT](../model_doc/beit), [BiT](../model_doc/bit), [ConvNeXT](../model_doc/convnext), [ConvNeXTV2](../model_doc/convnextv2), [CvT](../model_doc/cvt), [Data2VecVision](../model_doc/data2vec-vision), [DeiT](../model_doc/deit), [DiNAT](../model_doc/dinat), [DINOv2](../model_doc/dinov2), [EfficientFormer](../model_doc/efficientformer), [EfficientNet](../model_doc/efficientnet), [FocalNet](../model_doc/focalnet), [ImageGPT](../model_doc/imagegpt), [LeViT](../model_doc/levit), [MobileNetV1](../model_doc/mobilenet_v1), [MobileNetV2](../model_doc/mobilenet_v2), [MobileViT](../model_doc/mobilevit), [MobileViTV2](../model_doc/mobilevitv2), [NAT](../model_doc/nat), [Perceiver](../model_doc/perceiver), [PoolFormer](../model_doc/poolformer), [PVT](../model_doc/pvt), [RegNet](../model_doc/regnet), [ResNet](../model_doc/resnet), [SegFormer](../model_doc/segformer), [SwiftFormer](../model_doc/swiftformer), [Swin Transformer](../model_doc/swin), [Swin Transformer V2](../model_doc/swinv2), [VAN](../model_doc/van), [ViT](../model_doc/vit), [ViT Hybrid](../model_doc/vit_hybrid), [ViTMSN](../model_doc/vit_msn)
<!--End of the generated tip-->
</Tip>
Before you begin, make sure you have all the necessary libraries installed:
```bash
pip install transformers datasets evaluate
```
We encourage you to log in to your Hugging Face account to upload and share your model with the community. When prompted, enter your token to log in:
```py
>>> from huggingface_hub import notebook_login
>>> notebook_login()
```
## Load Food-101 dataset
Start by loading a smaller subset of the Food-101 dataset from the 🤗 Datasets library. This will give you a chance to
experiment and make sure everything works before spending more time training on the full dataset.
```py
>>> from datasets import load_dataset
>>> food = load_dataset("food101", split="train[:5000]")
```
Split the dataset's `train` split into a train and test set with the [`~datasets.Dataset.train_test_split`] method:
```py
>>> food = food.train_test_split(test_size=0.2)
```
Then take a look at an example:
```py
>>> food["train"][0]
{'image': <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=512x512 at 0x7F52AFC8AC50>,
'label': 79}
```
Each example in the dataset has two fields:
- `image`: a PIL image of the food item
- `label`: the label class of the food item
To make it easier for the model to get the label name from the label id, create a dictionary that maps the label name
to an integer and vice versa:
```py
>>> labels = food["train"].features["label"].names
>>> label2id, id2label = dict(), dict()
>>> for i, label in enumerate(labels):
... label2id[label] = str(i)
... id2label[str(i)] = label
```
Now you can convert the label id to a label name:
```py
>>> id2label[str(79)]
'prime_rib'
```
## Preprocess
The next step is to load a ViT image processor to process the image into a tensor:
```py
>>> from transformers import AutoImageProcessor
>>> checkpoint = "google/vit-base-patch16-224-in21k"
>>> image_processor = AutoImageProcessor.from_pretrained(checkpoint)
```
<frameworkcontent>
<pt>
Apply some image transformations to the images to make the model more robust against overfitting. Here you'll use torchvision's [`transforms`](https://pytorch.org/vision/stable/transforms.html) module, but you can also use any image library you like.
Crop a random part of the image, resize it, and normalize it with the image mean and standard deviation:
```py
>>> from torchvision.transforms import RandomResizedCrop, Compose, Normalize, ToTensor
>>> normalize = Normalize(mean=image_processor.image_mean, std=image_processor.image_std)
>>> size = (
... image_processor.size["shortest_edge"]
... if "shortest_edge" in image_processor.size
... else (image_processor.size["height"], image_processor.size["width"])
... )
>>> _transforms = Compose([RandomResizedCrop(size), ToTensor(), normalize])
```
Then create a preprocessing function to apply the transforms and return the `pixel_values` - the inputs to the model - of the image:
```py
>>> def transforms(examples):
... examples["pixel_values"] = [_transforms(img.convert("RGB")) for img in examples["image"]]
... del examples["image"]
... return examples
```
To apply the preprocessing function over the entire dataset, use 🤗 Datasets [`~datasets.Dataset.with_transform`] method. The transforms are applied on the fly when you load an element of the dataset:
```py
>>> food = food.with_transform(transforms)
```
Now create a batch of examples using [`DefaultDataCollator`]. Unlike other data collators in 🤗 Transformers, the `DefaultDataCollator` does not apply additional preprocessing such as padding.
```py
>>> from transformers import DefaultDataCollator
>>> data_collator = DefaultDataCollator()
```
</pt>
</frameworkcontent>
<frameworkcontent>
<tf>
To avoid overfitting and to make the model more robust, add some data augmentation to the training part of the dataset.
Here we use Keras preprocessing layers to define the transformations for the training data (includes data augmentation),
and transformations for the validation data (only center cropping, resizing and normalizing). You can use `tf.image`or
any other library you prefer.
```py
>>> from tensorflow import keras
>>> from tensorflow.keras import layers
>>> size = (image_processor.size["height"], image_processor.size["width"])
>>> train_data_augmentation = keras.Sequential(
... [
... layers.RandomCrop(size[0], size[1]),
... layers.Rescaling(scale=1.0 / 127.5, offset=-1),
... layers.RandomFlip("horizontal"),
... layers.RandomRotation(factor=0.02),
... layers.RandomZoom(height_factor=0.2, width_factor=0.2),
... ],
... name="train_data_augmentation",
... )
>>> val_data_augmentation = keras.Sequential(
... [
... layers.CenterCrop(size[0], size[1]),
... layers.Rescaling(scale=1.0 / 127.5, offset=-1),
... ],
... name="val_data_augmentation",
... )
```
Next, create functions to apply appropriate transformations to a batch of images, instead of one image at a time.
```py
>>> import numpy as np
>>> import tensorflow as tf
>>> from PIL import Image
>>> def convert_to_tf_tensor(image: Image):
... np_image = np.array(image)
... tf_image = tf.convert_to_tensor(np_image)
... # `expand_dims()` is used to add a batch dimension since
... # the TF augmentation layers operates on batched inputs.
... return tf.expand_dims(tf_image, 0)
>>> def preprocess_train(example_batch):
... """Apply train_transforms across a batch."""
... images = [
... train_data_augmentation(convert_to_tf_tensor(image.convert("RGB"))) for image in example_batch["image"]
... ]
... example_batch["pixel_values"] = [tf.transpose(tf.squeeze(image)) for image in images]
... return example_batch
... def preprocess_val(example_batch):
... """Apply val_transforms across a batch."""
... images = [
... val_data_augmentation(convert_to_tf_tensor(image.convert("RGB"))) for image in example_batch["image"]
... ]
... example_batch["pixel_values"] = [tf.transpose(tf.squeeze(image)) for image in images]
... return example_batch
```
Use 🤗 Datasets [`~datasets.Dataset.set_transform`] to apply the transformations on the fly:
```py
food["train"].set_transform(preprocess_train)
food["test"].set_transform(preprocess_val)
```
As a final preprocessing step, create a batch of examples using `DefaultDataCollator`. Unlike other data collators in 🤗 Transformers, the
`DefaultDataCollator` does not apply additional preprocessing, such as padding.
```py
>>> from transformers import DefaultDataCollator
>>> data_collator = DefaultDataCollator(return_tensors="tf")
```
</tf>
</frameworkcontent>
## Evaluate
Including a metric during training is often helpful for evaluating your model's performance. You can quickly load an
evaluation method with the 🤗 [Evaluate](https://huggingface.co/docs/evaluate/index) library. For this task, load
the [accuracy](https://huggingface.co/spaces/evaluate-metric/accuracy) metric (see the 🤗 Evaluate [quick tour](https://huggingface.co/docs/evaluate/a_quick_tour) to learn more about how to load and compute a metric):
```py
>>> import evaluate
>>> accuracy = evaluate.load("accuracy")
```
Then create a function that passes your predictions and labels to [`~evaluate.EvaluationModule.compute`] to calculate the accuracy:
```py
>>> import numpy as np
>>> def compute_metrics(eval_pred):
... predictions, labels = eval_pred
... predictions = np.argmax(predictions, axis=1)
... return accuracy.compute(predictions=predictions, references=labels)
```
Your `compute_metrics` function is ready to go now, and you'll return to it when you set up your training.
## Train
<frameworkcontent>
<pt>
<Tip>
If you aren't familiar with finetuning a model with the [`Trainer`], take a look at the basic tutorial [here](../training#train-with-pytorch-trainer)!
</Tip>
You're ready to start training your model now! Load ViT with [`AutoModelForImageClassification`]. Specify the number of labels along with the number of expected labels, and the label mappings:
```py
>>> from transformers import AutoModelForImageClassification, TrainingArguments, Trainer
>>> model = AutoModelForImageClassification.from_pretrained(
... checkpoint,
... num_labels=len(labels),
... id2label=id2label,
... label2id=label2id,
... )
```
At this point, only three steps remain:
1. Define your training hyperparameters in [`TrainingArguments`]. It is important you don't remove unused columns because that'll drop the `image` column. Without the `image` column, you can't create `pixel_values`. Set `remove_unused_columns=False` to prevent this behavior! The only other required parameter is `output_dir` which specifies where to save your model. You'll push this model to the Hub by setting `push_to_hub=True` (you need to be signed in to Hugging Face to upload your model). At the end of each epoch, the [`Trainer`] will evaluate the accuracy and save the training checkpoint.
2. Pass the training arguments to [`Trainer`] along with the model, dataset, tokenizer, data collator, and `compute_metrics` function.
3. Call [`~Trainer.train`] to finetune your model.
```py
>>> training_args = TrainingArguments(
... output_dir="my_awesome_food_model",
... remove_unused_columns=False,
... evaluation_strategy="epoch",
... save_strategy="epoch",
... learning_rate=5e-5,
... per_device_train_batch_size=16,
... gradient_accumulation_steps=4,
... per_device_eval_batch_size=16,
... num_train_epochs=3,
... warmup_ratio=0.1,
... logging_steps=10,
... load_best_model_at_end=True,
... metric_for_best_model="accuracy",
... push_to_hub=True,
... )
>>> trainer = Trainer(
... model=model,
... args=training_args,
... data_collator=data_collator,
... train_dataset=food["train"],
... eval_dataset=food["test"],
... tokenizer=image_processor,
... compute_metrics=compute_metrics,
... )
>>> trainer.train()
```
Once training is completed, share your model to the Hub with the [`~transformers.Trainer.push_to_hub`] method so everyone can use your model:
```py
>>> trainer.push_to_hub()
```
</pt>
</frameworkcontent>
<frameworkcontent>
<tf>
<Tip>
If you are unfamiliar with fine-tuning a model with Keras, check out the [basic tutorial](./training#train-a-tensorflow-model-with-keras) first!
</Tip>
To fine-tune a model in TensorFlow, follow these steps:
1. Define the training hyperparameters, and set up an optimizer and a learning rate schedule.
2. Instantiate a pre-trained model.
3. Convert a 🤗 Dataset to a `tf.data.Dataset`.
4. Compile your model.
5. Add callbacks and use the `fit()` method to run the training.
6. Upload your model to 🤗 Hub to share with the community.
Start by defining the hyperparameters, optimizer and learning rate schedule:
```py
>>> from transformers import create_optimizer
>>> batch_size = 16
>>> num_epochs = 5
>>> num_train_steps = len(food["train"]) * num_epochs
>>> learning_rate = 3e-5
>>> weight_decay_rate = 0.01
>>> optimizer, lr_schedule = create_optimizer(
... init_lr=learning_rate,
... num_train_steps=num_train_steps,
... weight_decay_rate=weight_decay_rate,
... num_warmup_steps=0,
... )
```
Then, load ViT with [`TFAutoModelForImageClassification`] along with the label mappings:
```py
>>> from transformers import TFAutoModelForImageClassification
>>> model = TFAutoModelForImageClassification.from_pretrained(
... checkpoint,
... id2label=id2label,
... label2id=label2id,
... )
```
Convert your datasets to the `tf.data.Dataset` format using the [`~datasets.Dataset.to_tf_dataset`] and your `data_collator`:
```py
>>> # converting our train dataset to tf.data.Dataset
>>> tf_train_dataset = food["train"].to_tf_dataset(
... columns="pixel_values", label_cols="label", shuffle=True, batch_size=batch_size, collate_fn=data_collator
... )
>>> # converting our test dataset to tf.data.Dataset
>>> tf_eval_dataset = food["test"].to_tf_dataset(
... columns="pixel_values", label_cols="label", shuffle=True, batch_size=batch_size, collate_fn=data_collator
... )
```
Configure the model for training with `compile()`:
```py
>>> from tensorflow.keras.losses import SparseCategoricalCrossentropy
>>> loss = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)
>>> model.compile(optimizer=optimizer, loss=loss)
```
To compute the accuracy from the predictions and push your model to the 🤗 Hub, use [Keras callbacks](../main_classes/keras_callbacks).
Pass your `compute_metrics` function to [KerasMetricCallback](../main_classes/keras_callbacks#transformers.KerasMetricCallback),
and use the [PushToHubCallback](../main_classes/keras_callbacks#transformers.PushToHubCallback) to upload the model:
```py
>>> from transformers.keras_callbacks import KerasMetricCallback, PushToHubCallback
>>> metric_callback = KerasMetricCallback(metric_fn=compute_metrics, eval_dataset=tf_eval_dataset)
>>> push_to_hub_callback = PushToHubCallback(
... output_dir="food_classifier",
... tokenizer=image_processor,
... save_strategy="no",
... )
>>> callbacks = [metric_callback, push_to_hub_callback]
```
Finally, you are ready to train your model! Call `fit()` with your training and validation datasets, the number of epochs,
and your callbacks to fine-tune the model:
```py
>>> model.fit(tf_train_dataset, validation_data=tf_eval_dataset, epochs=num_epochs, callbacks=callbacks)
Epoch 1/5
250/250 [==============================] - 313s 1s/step - loss: 2.5623 - val_loss: 1.4161 - accuracy: 0.9290
Epoch 2/5
250/250 [==============================] - 265s 1s/step - loss: 0.9181 - val_loss: 0.6808 - accuracy: 0.9690
Epoch 3/5
250/250 [==============================] - 252s 1s/step - loss: 0.3910 - val_loss: 0.4303 - accuracy: 0.9820
Epoch 4/5
250/250 [==============================] - 251s 1s/step - loss: 0.2028 - val_loss: 0.3191 - accuracy: 0.9900
Epoch 5/5
250/250 [==============================] - 238s 949ms/step - loss: 0.1232 - val_loss: 0.3259 - accuracy: 0.9890
```
Congratulations! You have fine-tuned your model and shared it on the 🤗 Hub. You can now use it for inference!
</tf>
</frameworkcontent>
<Tip>
For a more in-depth example of how to finetune a model for image classification, take a look at the corresponding [PyTorch notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/image_classification.ipynb).
</Tip>
## Inference
Great, now that you've fine-tuned a model, you can use it for inference!
Load an image you'd like to run inference on:
```py
>>> ds = load_dataset("food101", split="validation[:10]")
>>> image = ds["image"][0]
```
<div class="flex justify-center">
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/beignets-task-guide.png" alt="image of beignets"/>
</div>
The simplest way to try out your finetuned model for inference is to use it in a [`pipeline`]. Instantiate a `pipeline` for image classification with your model, and pass your image to it:
```py
>>> from transformers import pipeline
>>> classifier = pipeline("image-classification", model="my_awesome_food_model")
>>> classifier(image)
[{'score': 0.31856709718704224, 'label': 'beignets'},
{'score': 0.015232225880026817, 'label': 'bruschetta'},
{'score': 0.01519392803311348, 'label': 'chicken_wings'},
{'score': 0.013022331520915031, 'label': 'pork_chop'},
{'score': 0.012728818692266941, 'label': 'prime_rib'}]
```
You can also manually replicate the results of the `pipeline` if you'd like:
<frameworkcontent>
<pt>
Load an image processor to preprocess the image and return the `input` as PyTorch tensors:
```py
>>> from transformers import AutoImageProcessor
>>> import torch
>>> image_processor = AutoImageProcessor.from_pretrained("my_awesome_food_model")
>>> inputs = image_processor(image, return_tensors="pt")
```
Pass your inputs to the model and return the logits:
```py
>>> from transformers import AutoModelForImageClassification
>>> model = AutoModelForImageClassification.from_pretrained("my_awesome_food_model")
>>> with torch.no_grad():
... logits = model(**inputs).logits
```
Get the predicted label with the highest probability, and use the model's `id2label` mapping to convert it to a label:
```py
>>> predicted_label = logits.argmax(-1).item()
>>> model.config.id2label[predicted_label]
'beignets'
```
</pt>
</frameworkcontent>
<frameworkcontent>
<tf>
Load an image processor to preprocess the image and return the `input` as TensorFlow tensors:
```py
>>> from transformers import AutoImageProcessor
>>> image_processor = AutoImageProcessor.from_pretrained("MariaK/food_classifier")
>>> inputs = image_processor(image, return_tensors="tf")
```
Pass your inputs to the model and return the logits:
```py
>>> from transformers import TFAutoModelForImageClassification
>>> model = TFAutoModelForImageClassification.from_pretrained("MariaK/food_classifier")
>>> logits = model(**inputs).logits
```
Get the predicted label with the highest probability, and use the model's `id2label` mapping to convert it to a label:
```py
>>> predicted_class_id = int(tf.math.argmax(logits, axis=-1)[0])
>>> model.config.id2label[predicted_class_id]
'beignets'
```
</tf>
</frameworkcontent>
| huggingface/transformers/blob/main/docs/source/en/tasks/image_classification.md |
!--Copyright 2022 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Whisper
## Overview
The Whisper model was proposed in [Robust Speech Recognition via Large-Scale Weak Supervision](https://cdn.openai.com/papers/whisper.pdf) by Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, Ilya Sutskever.
The abstract from the paper is the following:
*We study the capabilities of speech processing systems trained simply to predict large amounts of transcripts of audio on the internet. When scaled to 680,000 hours of multilingual and multitask supervision, the resulting models generalize well to standard benchmarks and are often competitive with prior fully supervised results but in a zeroshot transfer setting without the need for any finetuning. When compared to humans, the models approach their accuracy and robustness. We are releasing models and inference code to serve as a foundation for further work on robust speech processing.*
This model was contributed by [Arthur Zucker](https://huggingface.co/ArthurZ). The Tensorflow version of this model was contributed by [amyeroberts](https://huggingface.co/amyeroberts).
The original code can be found [here](https://github.com/openai/whisper).
## Usage tips
- The model usually performs well without requiring any finetuning.
- The architecture follows a classic encoder-decoder architecture, which means that it relies on the [`~generation.GenerationMixin.generate`] function for inference.
- Inference is currently only implemented for short-form i.e. audio is pre-segmented into <=30s segments. Long-form (including timestamps) will be implemented in a future release.
- One can use [`WhisperProcessor`] to prepare audio for the model, and decode the predicted ID's back into text.
- To convert the model and the processor, we recommend using the following:
```bash
python src/transformers/models/whisper/convert_openai_to_hf.py --checkpoint_path "" --pytorch_dump_folder_path "Arthur/whisper-3" --convert_preprocessor True
```
The script will automatically determine all necessary parameters from the OpenAI checkpoint. A `tiktoken` library needs to be installed
to perform the conversion of the OpenAI tokenizer to the `tokenizers` version.
## Inference
Here is a step-by-step guide to transcribing an audio sample using a pre-trained Whisper model:
```python
>>> from datasets import load_dataset
>>> from transformers import WhisperProcessor, WhisperForConditionalGeneration
>>> # Select an audio file and read it:
>>> ds = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation")
>>> audio_sample = ds[0]["audio"]
>>> waveform = audio_sample["array"]
>>> sampling_rate = audio_sample["sampling_rate"]
>>> # Load the Whisper model in Hugging Face format:
>>> processor = WhisperProcessor.from_pretrained("openai/whisper-tiny.en")
>>> model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-tiny.en")
>>> # Use the model and processor to transcribe the audio:
>>> input_features = processor(
... waveform, sampling_rate=sampling_rate, return_tensors="pt"
... ).input_features
>>> # Generate token ids
>>> predicted_ids = model.generate(input_features)
>>> # Decode token ids to text
>>> transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)
>>> transcription[0]
' Mr. Quilter is the apostle of the middle classes, and we are glad to welcome his gospel.'
```
## Resources
A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with Whisper. If you're interested in submitting a resource to be included here, please feel free to open a Pull Request and we'll review it! The resource should ideally demonstrate something new instead of duplicating an existing resource.
- A fork with a script to [convert a Whisper model in Hugging Face format to OpenAI format](https://github.com/zuazo-forks/transformers/blob/convert_hf_to_openai/src/transformers/models/whisper/convert_hf_to_openai.py). 🌎
Usage example:
```bash
pip install -U openai-whisper
python convert_hf_to_openai.py \
--checkpoint openai/whisper-tiny \
--whisper_dump_path whisper-tiny-openai.pt
```
## WhisperConfig
[[autodoc]] WhisperConfig
## WhisperTokenizer
[[autodoc]] WhisperTokenizer
- set_prefix_tokens
- build_inputs_with_special_tokens
- get_special_tokens_mask
- create_token_type_ids_from_sequences
- save_vocabulary
- batch_decode
- decode
## WhisperTokenizerFast
[[autodoc]] WhisperTokenizerFast
- set_prefix_tokens
- build_inputs_with_special_tokens
- get_special_tokens_mask
- create_token_type_ids_from_sequences
- save_vocabulary
- batch_decode
- decode
## WhisperFeatureExtractor
[[autodoc]] WhisperFeatureExtractor
- __call__
## WhisperProcessor
[[autodoc]] WhisperProcessor
- __call__
- from_pretrained
- save_pretrained
- batch_decode
- decode
<frameworkcontent>
<pt>
## WhisperModel
[[autodoc]] WhisperModel
- forward
- _mask_input_features
## WhisperForConditionalGeneration
[[autodoc]] WhisperForConditionalGeneration
- forward
- generate
## WhisperForCausalLM
[[autodoc]] WhisperForCausalLM
- forward
## WhisperForAudioClassification
[[autodoc]] WhisperForAudioClassification
- forward
</pt>
<tf>
## TFWhisperModel
[[autodoc]] TFWhisperModel
- call
## TFWhisperForConditionalGeneration
[[autodoc]] TFWhisperForConditionalGeneration
- call
</tf>
<jax>
## FlaxWhisperModel
[[autodoc]] FlaxWhisperModel
- __call__
## FlaxWhisperForConditionalGeneration
[[autodoc]] FlaxWhisperForConditionalGeneration
- __call__
## FlaxWhisperForAudioClassification
[[autodoc]] FlaxWhisperForAudioClassification
- __call__
</jax>
</frameworkcontent>
| huggingface/transformers/blob/main/docs/source/en/model_doc/whisper.md |
!--Copyright 2021 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# GPT-J
## Overview
The GPT-J model was released in the [kingoflolz/mesh-transformer-jax](https://github.com/kingoflolz/mesh-transformer-jax) repository by Ben Wang and Aran Komatsuzaki. It is a GPT-2-like
causal language model trained on [the Pile](https://pile.eleuther.ai/) dataset.
This model was contributed by [Stella Biderman](https://huggingface.co/stellaathena).
## Usage tips
- To load [GPT-J](https://huggingface.co/EleutherAI/gpt-j-6B) in float32 one would need at least 2x model size
RAM: 1x for initial weights and another 1x to load the checkpoint. So for GPT-J it would take at least 48GB
RAM to just load the model. To reduce the RAM usage there are a few options. The `torch_dtype` argument can be
used to initialize the model in half-precision on a CUDA device only. There is also a fp16 branch which stores the fp16 weights,
which could be used to further minimize the RAM usage:
```python
>>> from transformers import GPTJForCausalLM
>>> import torch
>>> device = "cuda"
>>> model = GPTJForCausalLM.from_pretrained(
... "EleutherAI/gpt-j-6B",
... revision="float16",
... torch_dtype=torch.float16,
... ).to(device)
```
- The model should fit on 16GB GPU for inference. For training/fine-tuning it would take much more GPU RAM. Adam
optimizer for example makes four copies of the model: model, gradients, average and squared average of the gradients.
So it would need at least 4x model size GPU memory, even with mixed precision as gradient updates are in fp32. This
is not including the activations and data batches, which would again require some more GPU RAM. So one should explore
solutions such as DeepSpeed, to train/fine-tune the model. Another option is to use the original codebase to
train/fine-tune the model on TPU and then convert the model to Transformers format for inference. Instructions for
that could be found [here](https://github.com/kingoflolz/mesh-transformer-jax/blob/master/howto_finetune.md)
- Although the embedding matrix has a size of 50400, only 50257 entries are used by the GPT-2 tokenizer. These extra
tokens are added for the sake of efficiency on TPUs. To avoid the mismatch between embedding matrix size and vocab
size, the tokenizer for [GPT-J](https://huggingface.co/EleutherAI/gpt-j-6B) contains 143 extra tokens
`<|extratoken_1|>... <|extratoken_143|>`, so the `vocab_size` of tokenizer also becomes 50400.
## Usage examples
The [`~generation.GenerationMixin.generate`] method can be used to generate text using GPT-J
model.
```python
>>> from transformers import AutoModelForCausalLM, AutoTokenizer
>>> model = AutoModelForCausalLM.from_pretrained("EleutherAI/gpt-j-6B")
>>> tokenizer = AutoTokenizer.from_pretrained("EleutherAI/gpt-j-6B")
>>> prompt = (
... "In a shocking finding, scientists discovered a herd of unicorns living in a remote, "
... "previously unexplored valley, in the Andes Mountains. Even more surprising to the "
... "researchers was the fact that the unicorns spoke perfect English."
... )
>>> input_ids = tokenizer(prompt, return_tensors="pt").input_ids
>>> gen_tokens = model.generate(
... input_ids,
... do_sample=True,
... temperature=0.9,
... max_length=100,
... )
>>> gen_text = tokenizer.batch_decode(gen_tokens)[0]
```
...or in float16 precision:
```python
>>> from transformers import GPTJForCausalLM, AutoTokenizer
>>> import torch
>>> device = "cuda"
>>> model = GPTJForCausalLM.from_pretrained("EleutherAI/gpt-j-6B", torch_dtype=torch.float16).to(device)
>>> tokenizer = AutoTokenizer.from_pretrained("EleutherAI/gpt-j-6B")
>>> prompt = (
... "In a shocking finding, scientists discovered a herd of unicorns living in a remote, "
... "previously unexplored valley, in the Andes Mountains. Even more surprising to the "
... "researchers was the fact that the unicorns spoke perfect English."
... )
>>> input_ids = tokenizer(prompt, return_tensors="pt").input_ids.to(device)
>>> gen_tokens = model.generate(
... input_ids,
... do_sample=True,
... temperature=0.9,
... max_length=100,
... )
>>> gen_text = tokenizer.batch_decode(gen_tokens)[0]
```
## Resources
A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with GPT-J. If you're interested in submitting a resource to be included here, please feel free to open a Pull Request and we'll review it! The resource should ideally demonstrate something new instead of duplicating an existing resource.
<PipelineTag pipeline="text-generation"/>
- Description of [GPT-J](https://huggingface.co/EleutherAI/gpt-j-6B).
- A blog on how to [Deploy GPT-J 6B for inference using Hugging Face Transformers and Amazon SageMaker](https://huggingface.co/blog/gptj-sagemaker).
- A blog on how to [Accelerate GPT-J inference with DeepSpeed-Inference on GPUs](https://www.philschmid.de/gptj-deepspeed-inference).
- A blog post introducing [GPT-J-6B: 6B JAX-Based Transformer](https://arankomatsuzaki.wordpress.com/2021/06/04/gpt-j/). 🌎
- A notebook for [GPT-J-6B Inference Demo](https://colab.research.google.com/github/kingoflolz/mesh-transformer-jax/blob/master/colab_demo.ipynb). 🌎
- Another notebook demonstrating [Inference with GPT-J-6B](https://colab.research.google.com/github/NielsRogge/Transformers-Tutorials/blob/master/GPT-J-6B/Inference_with_GPT_J_6B.ipynb).
- [Causal language modeling](https://huggingface.co/course/en/chapter7/6?fw=pt#training-a-causal-language-model-from-scratch) chapter of the 🤗 Hugging Face Course.
- [`GPTJForCausalLM`] is supported by this [causal language modeling example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/language-modeling#gpt-2gpt-and-causal-language-modeling), [text generation example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/text-generation), and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/language_modeling.ipynb).
- [`TFGPTJForCausalLM`] is supported by this [causal language modeling example script](https://github.com/huggingface/transformers/tree/main/examples/tensorflow/language-modeling#run_clmpy) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/language_modeling-tf.ipynb).
- [`FlaxGPTJForCausalLM`] is supported by this [causal language modeling example script](https://github.com/huggingface/transformers/tree/main/examples/flax/language-modeling#causal-language-modeling) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/causal_language_modeling_flax.ipynb).
**Documentation resources**
- [Text classification task guide](../tasks/sequence_classification)
- [Question answering task guide](../tasks/question_answering)
- [Causal language modeling task guide](../tasks/language_modeling)
## GPTJConfig
[[autodoc]] GPTJConfig
- all
<frameworkcontent>
<pt>
## GPTJModel
[[autodoc]] GPTJModel
- forward
## GPTJForCausalLM
[[autodoc]] GPTJForCausalLM
- forward
## GPTJForSequenceClassification
[[autodoc]] GPTJForSequenceClassification
- forward
## GPTJForQuestionAnswering
[[autodoc]] GPTJForQuestionAnswering
- forward
</pt>
<tf>
## TFGPTJModel
[[autodoc]] TFGPTJModel
- call
## TFGPTJForCausalLM
[[autodoc]] TFGPTJForCausalLM
- call
## TFGPTJForSequenceClassification
[[autodoc]] TFGPTJForSequenceClassification
- call
## TFGPTJForQuestionAnswering
[[autodoc]] TFGPTJForQuestionAnswering
- call
</tf>
<jax>
## FlaxGPTJModel
[[autodoc]] FlaxGPTJModel
- __call__
## FlaxGPTJForCausalLM
[[autodoc]] FlaxGPTJForCausalLM
- __call__
</jax>
</frameworkcontent>
| huggingface/transformers/blob/main/docs/source/en/model_doc/gptj.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Model outputs
All models have outputs that are instances of subclasses of [`~utils.ModelOutput`]. Those are
data structures containing all the information returned by the model, but that can also be used as tuples or
dictionaries.
Let's see how this looks in an example:
```python
from transformers import BertTokenizer, BertForSequenceClassification
import torch
tokenizer = BertTokenizer.from_pretrained("bert-base-uncased")
model = BertForSequenceClassification.from_pretrained("bert-base-uncased")
inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
labels = torch.tensor([1]).unsqueeze(0) # Batch size 1
outputs = model(**inputs, labels=labels)
```
The `outputs` object is a [`~modeling_outputs.SequenceClassifierOutput`], as we can see in the
documentation of that class below, it means it has an optional `loss`, a `logits`, an optional `hidden_states` and
an optional `attentions` attribute. Here we have the `loss` since we passed along `labels`, but we don't have
`hidden_states` and `attentions` because we didn't pass `output_hidden_states=True` or
`output_attentions=True`.
<Tip>
When passing `output_hidden_states=True` you may expect the `outputs.hidden_states[-1]` to match `outputs.last_hidden_states` exactly.
However, this is not always the case. Some models apply normalization or subsequent process to the last hidden state when it's returned.
</Tip>
You can access each attribute as you would usually do, and if that attribute has not been returned by the model, you
will get `None`. Here for instance `outputs.loss` is the loss computed by the model, and `outputs.attentions` is
`None`.
When considering our `outputs` object as tuple, it only considers the attributes that don't have `None` values.
Here for instance, it has two elements, `loss` then `logits`, so
```python
outputs[:2]
```
will return the tuple `(outputs.loss, outputs.logits)` for instance.
When considering our `outputs` object as dictionary, it only considers the attributes that don't have `None`
values. Here for instance, it has two keys that are `loss` and `logits`.
We document here the generic model outputs that are used by more than one model type. Specific output types are
documented on their corresponding model page.
## ModelOutput
[[autodoc]] utils.ModelOutput
- to_tuple
## BaseModelOutput
[[autodoc]] modeling_outputs.BaseModelOutput
## BaseModelOutputWithPooling
[[autodoc]] modeling_outputs.BaseModelOutputWithPooling
## BaseModelOutputWithCrossAttentions
[[autodoc]] modeling_outputs.BaseModelOutputWithCrossAttentions
## BaseModelOutputWithPoolingAndCrossAttentions
[[autodoc]] modeling_outputs.BaseModelOutputWithPoolingAndCrossAttentions
## BaseModelOutputWithPast
[[autodoc]] modeling_outputs.BaseModelOutputWithPast
## BaseModelOutputWithPastAndCrossAttentions
[[autodoc]] modeling_outputs.BaseModelOutputWithPastAndCrossAttentions
## Seq2SeqModelOutput
[[autodoc]] modeling_outputs.Seq2SeqModelOutput
## CausalLMOutput
[[autodoc]] modeling_outputs.CausalLMOutput
## CausalLMOutputWithCrossAttentions
[[autodoc]] modeling_outputs.CausalLMOutputWithCrossAttentions
## CausalLMOutputWithPast
[[autodoc]] modeling_outputs.CausalLMOutputWithPast
## MaskedLMOutput
[[autodoc]] modeling_outputs.MaskedLMOutput
## Seq2SeqLMOutput
[[autodoc]] modeling_outputs.Seq2SeqLMOutput
## NextSentencePredictorOutput
[[autodoc]] modeling_outputs.NextSentencePredictorOutput
## SequenceClassifierOutput
[[autodoc]] modeling_outputs.SequenceClassifierOutput
## Seq2SeqSequenceClassifierOutput
[[autodoc]] modeling_outputs.Seq2SeqSequenceClassifierOutput
## MultipleChoiceModelOutput
[[autodoc]] modeling_outputs.MultipleChoiceModelOutput
## TokenClassifierOutput
[[autodoc]] modeling_outputs.TokenClassifierOutput
## QuestionAnsweringModelOutput
[[autodoc]] modeling_outputs.QuestionAnsweringModelOutput
## Seq2SeqQuestionAnsweringModelOutput
[[autodoc]] modeling_outputs.Seq2SeqQuestionAnsweringModelOutput
## Seq2SeqSpectrogramOutput
[[autodoc]] modeling_outputs.Seq2SeqSpectrogramOutput
## SemanticSegmenterOutput
[[autodoc]] modeling_outputs.SemanticSegmenterOutput
## ImageClassifierOutput
[[autodoc]] modeling_outputs.ImageClassifierOutput
## ImageClassifierOutputWithNoAttention
[[autodoc]] modeling_outputs.ImageClassifierOutputWithNoAttention
## DepthEstimatorOutput
[[autodoc]] modeling_outputs.DepthEstimatorOutput
## Wav2Vec2BaseModelOutput
[[autodoc]] modeling_outputs.Wav2Vec2BaseModelOutput
## XVectorOutput
[[autodoc]] modeling_outputs.XVectorOutput
## Seq2SeqTSModelOutput
[[autodoc]] modeling_outputs.Seq2SeqTSModelOutput
## Seq2SeqTSPredictionOutput
[[autodoc]] modeling_outputs.Seq2SeqTSPredictionOutput
## SampleTSPredictionOutput
[[autodoc]] modeling_outputs.SampleTSPredictionOutput
## TFBaseModelOutput
[[autodoc]] modeling_tf_outputs.TFBaseModelOutput
## TFBaseModelOutputWithPooling
[[autodoc]] modeling_tf_outputs.TFBaseModelOutputWithPooling
## TFBaseModelOutputWithPoolingAndCrossAttentions
[[autodoc]] modeling_tf_outputs.TFBaseModelOutputWithPoolingAndCrossAttentions
## TFBaseModelOutputWithPast
[[autodoc]] modeling_tf_outputs.TFBaseModelOutputWithPast
## TFBaseModelOutputWithPastAndCrossAttentions
[[autodoc]] modeling_tf_outputs.TFBaseModelOutputWithPastAndCrossAttentions
## TFSeq2SeqModelOutput
[[autodoc]] modeling_tf_outputs.TFSeq2SeqModelOutput
## TFCausalLMOutput
[[autodoc]] modeling_tf_outputs.TFCausalLMOutput
## TFCausalLMOutputWithCrossAttentions
[[autodoc]] modeling_tf_outputs.TFCausalLMOutputWithCrossAttentions
## TFCausalLMOutputWithPast
[[autodoc]] modeling_tf_outputs.TFCausalLMOutputWithPast
## TFMaskedLMOutput
[[autodoc]] modeling_tf_outputs.TFMaskedLMOutput
## TFSeq2SeqLMOutput
[[autodoc]] modeling_tf_outputs.TFSeq2SeqLMOutput
## TFNextSentencePredictorOutput
[[autodoc]] modeling_tf_outputs.TFNextSentencePredictorOutput
## TFSequenceClassifierOutput
[[autodoc]] modeling_tf_outputs.TFSequenceClassifierOutput
## TFSeq2SeqSequenceClassifierOutput
[[autodoc]] modeling_tf_outputs.TFSeq2SeqSequenceClassifierOutput
## TFMultipleChoiceModelOutput
[[autodoc]] modeling_tf_outputs.TFMultipleChoiceModelOutput
## TFTokenClassifierOutput
[[autodoc]] modeling_tf_outputs.TFTokenClassifierOutput
## TFQuestionAnsweringModelOutput
[[autodoc]] modeling_tf_outputs.TFQuestionAnsweringModelOutput
## TFSeq2SeqQuestionAnsweringModelOutput
[[autodoc]] modeling_tf_outputs.TFSeq2SeqQuestionAnsweringModelOutput
## FlaxBaseModelOutput
[[autodoc]] modeling_flax_outputs.FlaxBaseModelOutput
## FlaxBaseModelOutputWithPast
[[autodoc]] modeling_flax_outputs.FlaxBaseModelOutputWithPast
## FlaxBaseModelOutputWithPooling
[[autodoc]] modeling_flax_outputs.FlaxBaseModelOutputWithPooling
## FlaxBaseModelOutputWithPastAndCrossAttentions
[[autodoc]] modeling_flax_outputs.FlaxBaseModelOutputWithPastAndCrossAttentions
## FlaxSeq2SeqModelOutput
[[autodoc]] modeling_flax_outputs.FlaxSeq2SeqModelOutput
## FlaxCausalLMOutputWithCrossAttentions
[[autodoc]] modeling_flax_outputs.FlaxCausalLMOutputWithCrossAttentions
## FlaxMaskedLMOutput
[[autodoc]] modeling_flax_outputs.FlaxMaskedLMOutput
## FlaxSeq2SeqLMOutput
[[autodoc]] modeling_flax_outputs.FlaxSeq2SeqLMOutput
## FlaxNextSentencePredictorOutput
[[autodoc]] modeling_flax_outputs.FlaxNextSentencePredictorOutput
## FlaxSequenceClassifierOutput
[[autodoc]] modeling_flax_outputs.FlaxSequenceClassifierOutput
## FlaxSeq2SeqSequenceClassifierOutput
[[autodoc]] modeling_flax_outputs.FlaxSeq2SeqSequenceClassifierOutput
## FlaxMultipleChoiceModelOutput
[[autodoc]] modeling_flax_outputs.FlaxMultipleChoiceModelOutput
## FlaxTokenClassifierOutput
[[autodoc]] modeling_flax_outputs.FlaxTokenClassifierOutput
## FlaxQuestionAnsweringModelOutput
[[autodoc]] modeling_flax_outputs.FlaxQuestionAnsweringModelOutput
## FlaxSeq2SeqQuestionAnsweringModelOutput
[[autodoc]] modeling_flax_outputs.FlaxSeq2SeqQuestionAnsweringModelOutput
| huggingface/transformers/blob/main/docs/source/en/main_classes/output.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# DeBERTa
## Overview
The DeBERTa model was proposed in [DeBERTa: Decoding-enhanced BERT with Disentangled Attention](https://arxiv.org/abs/2006.03654) by Pengcheng He, Xiaodong Liu, Jianfeng Gao, Weizhu Chen It is based on Google's
BERT model released in 2018 and Facebook's RoBERTa model released in 2019.
It builds on RoBERTa with disentangled attention and enhanced mask decoder training with half of the data used in
RoBERTa.
The abstract from the paper is the following:
*Recent progress in pre-trained neural language models has significantly improved the performance of many natural
language processing (NLP) tasks. In this paper we propose a new model architecture DeBERTa (Decoding-enhanced BERT with
disentangled attention) that improves the BERT and RoBERTa models using two novel techniques. The first is the
disentangled attention mechanism, where each word is represented using two vectors that encode its content and
position, respectively, and the attention weights among words are computed using disentangled matrices on their
contents and relative positions. Second, an enhanced mask decoder is used to replace the output softmax layer to
predict the masked tokens for model pretraining. We show that these two techniques significantly improve the efficiency
of model pretraining and performance of downstream tasks. Compared to RoBERTa-Large, a DeBERTa model trained on half of
the training data performs consistently better on a wide range of NLP tasks, achieving improvements on MNLI by +0.9%
(90.2% vs. 91.1%), on SQuAD v2.0 by +2.3% (88.4% vs. 90.7%) and RACE by +3.6% (83.2% vs. 86.8%). The DeBERTa code and
pre-trained models will be made publicly available at https://github.com/microsoft/DeBERTa.*
This model was contributed by [DeBERTa](https://huggingface.co/DeBERTa). This model TF 2.0 implementation was
contributed by [kamalkraj](https://huggingface.co/kamalkraj) . The original code can be found [here](https://github.com/microsoft/DeBERTa).
## Resources
A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with DeBERTa. If you're interested in submitting a resource to be included here, please feel free to open a Pull Request and we'll review it! The resource should ideally demonstrate something new instead of duplicating an existing resource.
<PipelineTag pipeline="text-classification"/>
- A blog post on how to [Accelerate Large Model Training using DeepSpeed](https://huggingface.co/blog/accelerate-deepspeed) with DeBERTa.
- A blog post on [Supercharged Customer Service with Machine Learning](https://huggingface.co/blog/supercharge-customer-service-with-machine-learning) with DeBERTa.
- [`DebertaForSequenceClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/text-classification) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/text_classification.ipynb).
- [`TFDebertaForSequenceClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/tensorflow/text-classification) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/text_classification-tf.ipynb).
- [Text classification task guide](../tasks/sequence_classification)
<PipelineTag pipeline="token-classification" />
- [`DebertaForTokenClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/token-classification) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/token_classification.ipynb).
- [`TFDebertaForTokenClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/tensorflow/token-classification) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/token_classification-tf.ipynb).
- [Token classification](https://huggingface.co/course/chapter7/2?fw=pt) chapter of the 🤗 Hugging Face Course.
- [Byte-Pair Encoding tokenization](https://huggingface.co/course/chapter6/5?fw=pt) chapter of the 🤗 Hugging Face Course.
- [Token classification task guide](../tasks/token_classification)
<PipelineTag pipeline="fill-mask"/>
- [`DebertaForMaskedLM`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/language-modeling#robertabertdistilbert-and-masked-language-modeling) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/language_modeling.ipynb).
- [`TFDebertaForMaskedLM`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/tensorflow/language-modeling#run_mlmpy) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/language_modeling-tf.ipynb).
- [Masked language modeling](https://huggingface.co/course/chapter7/3?fw=pt) chapter of the 🤗 Hugging Face Course.
- [Masked language modeling task guide](../tasks/masked_language_modeling)
<PipelineTag pipeline="question-answering"/>
- [`DebertaForQuestionAnswering`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/question-answering) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/question_answering.ipynb).
- [`TFDebertaForQuestionAnswering`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/tensorflow/question-answering) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/question_answering-tf.ipynb).
- [Question answering](https://huggingface.co/course/chapter7/7?fw=pt) chapter of the 🤗 Hugging Face Course.
- [Question answering task guide](../tasks/question_answering)
## DebertaConfig
[[autodoc]] DebertaConfig
## DebertaTokenizer
[[autodoc]] DebertaTokenizer
- build_inputs_with_special_tokens
- get_special_tokens_mask
- create_token_type_ids_from_sequences
- save_vocabulary
## DebertaTokenizerFast
[[autodoc]] DebertaTokenizerFast
- build_inputs_with_special_tokens
- create_token_type_ids_from_sequences
<frameworkcontent>
<pt>
## DebertaModel
[[autodoc]] DebertaModel
- forward
## DebertaPreTrainedModel
[[autodoc]] DebertaPreTrainedModel
## DebertaForMaskedLM
[[autodoc]] DebertaForMaskedLM
- forward
## DebertaForSequenceClassification
[[autodoc]] DebertaForSequenceClassification
- forward
## DebertaForTokenClassification
[[autodoc]] DebertaForTokenClassification
- forward
## DebertaForQuestionAnswering
[[autodoc]] DebertaForQuestionAnswering
- forward
</pt>
<tf>
## TFDebertaModel
[[autodoc]] TFDebertaModel
- call
## TFDebertaPreTrainedModel
[[autodoc]] TFDebertaPreTrainedModel
- call
## TFDebertaForMaskedLM
[[autodoc]] TFDebertaForMaskedLM
- call
## TFDebertaForSequenceClassification
[[autodoc]] TFDebertaForSequenceClassification
- call
## TFDebertaForTokenClassification
[[autodoc]] TFDebertaForTokenClassification
- call
## TFDebertaForQuestionAnswering
[[autodoc]] TFDebertaForQuestionAnswering
- call
</tf>
</frameworkcontent>
| huggingface/transformers/blob/main/docs/source/en/model_doc/deberta.md |
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# EnCodec
## Overview
The EnCodec neural codec model was proposed in [High Fidelity Neural Audio Compression](https://arxiv.org/abs/2210.13438) by Alexandre Défossez, Jade Copet, Gabriel Synnaeve, Yossi Adi.
The abstract from the paper is the following:
*We introduce a state-of-the-art real-time, high-fidelity, audio codec leveraging neural networks. It consists in a streaming encoder-decoder architecture with quantized latent space trained in an end-to-end fashion. We simplify and speed-up the training by using a single multiscale spectrogram adversary that efficiently reduces artifacts and produce high-quality samples. We introduce a novel loss balancer mechanism to stabilize training: the weight of a loss now defines the fraction of the overall gradient it should represent, thus decoupling the choice of this hyper-parameter from the typical scale of the loss. Finally, we study how lightweight Transformer models can be used to further compress the obtained representation by up to 40%, while staying faster than real time. We provide a detailed description of the key design choices of the proposed model including: training objective, architectural changes and a study of various perceptual loss functions. We present an extensive subjective evaluation (MUSHRA tests) together with an ablation study for a range of bandwidths and audio domains, including speech, noisy-reverberant speech, and music. Our approach is superior to the baselines methods across all evaluated settings, considering both 24 kHz monophonic and 48 kHz stereophonic audio.*
This model was contributed by [Matthijs](https://huggingface.co/Matthijs), [Patrick Von Platen](https://huggingface.co/patrickvonplaten) and [Arthur Zucker](https://huggingface.co/ArthurZ).
The original code can be found [here](https://github.com/facebookresearch/encodec).
## Usage example
Here is a quick example of how to encode and decode an audio using this model:
```python
>>> from datasets import load_dataset, Audio
>>> from transformers import EncodecModel, AutoProcessor
>>> librispeech_dummy = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation")
>>> model = EncodecModel.from_pretrained("facebook/encodec_24khz")
>>> processor = AutoProcessor.from_pretrained("facebook/encodec_24khz")
>>> librispeech_dummy = librispeech_dummy.cast_column("audio", Audio(sampling_rate=processor.sampling_rate))
>>> audio_sample = librispeech_dummy[-1]["audio"]["array"]
>>> inputs = processor(raw_audio=audio_sample, sampling_rate=processor.sampling_rate, return_tensors="pt")
>>> encoder_outputs = model.encode(inputs["input_values"], inputs["padding_mask"])
>>> audio_values = model.decode(encoder_outputs.audio_codes, encoder_outputs.audio_scales, inputs["padding_mask"])[0]
>>> # or the equivalent with a forward pass
>>> audio_values = model(inputs["input_values"], inputs["padding_mask"]).audio_values
```
## EncodecConfig
[[autodoc]] EncodecConfig
## EncodecFeatureExtractor
[[autodoc]] EncodecFeatureExtractor
- __call__
## EncodecModel
[[autodoc]] EncodecModel
- decode
- encode
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/encodec.md |
!---
Copyright 2021 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-->
# Model parallel language model training example
The following example showcases how to train/fine-tune GPTNeo model with model parallelism using
the JAX/Flax backend and the [`pjit`](https://jax.readthedocs.io/en/latest/jax.experimental.pjit.html) transformation.
> Note: The example is experimental and might have bugs. Also currently it only supports single V3-8.
The `partition.py` file defines the `PyTree` of `ParitionSpec` for the GPTNeo model which describes how the model will be sharded.
The actual sharding is auto-matically handled by `pjit`. The weights are sharded across all local devices.
To adapt the script for other models, we need to also change the `ParitionSpec` accordingly.
TODO: Add more explantion.
Before training, let's prepare our model first. To be able to shard the model, the sharded dimention needs to be a multiple of devices it'll be sharded on. But GPTNeo's vocab size is 50257, so we need to resize the embeddings accordingly.
```python
from transformers import FlaxGPTNeoForCausalLM, GPTNeoConfig
model = FlaxGPTNeoForCausalLM.from_pretrained("EleutherAI/gpt-neo-1.3B")
emb = jnp.zeros((50264, model.config.hidden_size))
# update the first 50257 weights using pre-trained weights
emb = emb.at[:50257, :].set(model.params["transformer"]["wte"]["embedding"])
params = model.params
params["transformer"]["wte"]["embedding"] = emb
# initialize a random model with the right vocab_size
config = GPTNeoConfig.from_pretrained("EleutherAI/gpt-neo-1.3B", vocab_size=50264)
model = FlaxGPTNeoForCausalLM(config)
# assign the pre-trained weights and save the model.
model.params = params
model.save_pretrained("gpt-neo-1.3B")
```
### Train Model
```bash
python run_clm_mp.py \
--model_name_or_path gpt-neo-1.3B \
--tokenizer_name gpt2 \
--dataset_name wikitext --dataset_config_name wikitext-2-raw-v1 \
--do_train --do_eval \
--block_size 1024 \
--num_train_epochs 5 \
--learning_rate 4e-6 \
--per_device_train_batch_size 3 --per_device_eval_batch_size 3 \
--overwrite_output_dir --output_dir ~/tmp/flax-clm \
--cache_dir ~/datasets_cache/wikitext --dtype bfloat16 \
--logging_steps 96 --eval_steps 96
``` | huggingface/transformers/blob/main/examples/research_projects/jax-projects/model_parallel/README.md |
VisualBERT Demo
This demo shows usage of VisualBERT VQA model and is adapted from LXMERT demo present [here](https://github.com/huggingface/transformers/blob/main/examples/research_projects/lxmert/demo.ipynb).
1. make a virtualenv: ``virtualenv venv`` and activate ``source venv/bin/activate``
2. install reqs: ``pip install -r ./requirements.txt``
3. usage is as shown in demo.ipynb
| huggingface/transformers/blob/main/examples/research_projects/visual_bert/README.md |
!--Copyright 2021 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# VisualBERT
## Overview
The VisualBERT model was proposed in [VisualBERT: A Simple and Performant Baseline for Vision and Language](https://arxiv.org/pdf/1908.03557) by Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang.
VisualBERT is a neural network trained on a variety of (image, text) pairs.
The abstract from the paper is the following:
*We propose VisualBERT, a simple and flexible framework for modeling a broad range of vision-and-language tasks.
VisualBERT consists of a stack of Transformer layers that implicitly align elements of an input text and regions in an
associated input image with self-attention. We further propose two visually-grounded language model objectives for
pre-training VisualBERT on image caption data. Experiments on four vision-and-language tasks including VQA, VCR, NLVR2,
and Flickr30K show that VisualBERT outperforms or rivals with state-of-the-art models while being significantly
simpler. Further analysis demonstrates that VisualBERT can ground elements of language to image regions without any
explicit supervision and is even sensitive to syntactic relationships, tracking, for example, associations between
verbs and image regions corresponding to their arguments.*
This model was contributed by [gchhablani](https://huggingface.co/gchhablani). The original code can be found [here](https://github.com/uclanlp/visualbert).
## Usage tips
1. Most of the checkpoints provided work with the [`VisualBertForPreTraining`] configuration. Other
checkpoints provided are the fine-tuned checkpoints for down-stream tasks - VQA ('visualbert-vqa'), VCR
('visualbert-vcr'), NLVR2 ('visualbert-nlvr2'). Hence, if you are not working on these downstream tasks, it is
recommended that you use the pretrained checkpoints.
2. For the VCR task, the authors use a fine-tuned detector for generating visual embeddings, for all the checkpoints.
We do not provide the detector and its weights as a part of the package, but it will be available in the research
projects, and the states can be loaded directly into the detector provided.
VisualBERT is a multi-modal vision and language model. It can be used for visual question answering, multiple choice,
visual reasoning and region-to-phrase correspondence tasks. VisualBERT uses a BERT-like transformer to prepare
embeddings for image-text pairs. Both the text and visual features are then projected to a latent space with identical
dimension.
To feed images to the model, each image is passed through a pre-trained object detector and the regions and the
bounding boxes are extracted. The authors use the features generated after passing these regions through a pre-trained
CNN like ResNet as visual embeddings. They also add absolute position embeddings, and feed the resulting sequence of
vectors to a standard BERT model. The text input is concatenated in the front of the visual embeddings in the embedding
layer, and is expected to be bound by [CLS] and a [SEP] tokens, as in BERT. The segment IDs must also be set
appropriately for the textual and visual parts.
The [`BertTokenizer`] is used to encode the text. A custom detector/image processor must be used
to get the visual embeddings. The following example notebooks show how to use VisualBERT with Detectron-like models:
- [VisualBERT VQA demo notebook](https://github.com/huggingface/transformers/tree/main/examples/research_projects/visual_bert) : This notebook
contains an example on VisualBERT VQA.
- [Generate Embeddings for VisualBERT (Colab Notebook)](https://colab.research.google.com/drive/1bLGxKdldwqnMVA5x4neY7-l_8fKGWQYI?usp=sharing) : This notebook contains
an example on how to generate visual embeddings.
The following example shows how to get the last hidden state using [`VisualBertModel`]:
```python
>>> import torch
>>> from transformers import BertTokenizer, VisualBertModel
>>> model = VisualBertModel.from_pretrained("uclanlp/visualbert-vqa-coco-pre")
>>> tokenizer = BertTokenizer.from_pretrained("bert-base-uncased")
>>> inputs = tokenizer("What is the man eating?", return_tensors="pt")
>>> # this is a custom function that returns the visual embeddings given the image path
>>> visual_embeds = get_visual_embeddings(image_path)
>>> visual_token_type_ids = torch.ones(visual_embeds.shape[:-1], dtype=torch.long)
>>> visual_attention_mask = torch.ones(visual_embeds.shape[:-1], dtype=torch.float)
>>> inputs.update(
... {
... "visual_embeds": visual_embeds,
... "visual_token_type_ids": visual_token_type_ids,
... "visual_attention_mask": visual_attention_mask,
... }
... )
>>> outputs = model(**inputs)
>>> last_hidden_state = outputs.last_hidden_state
```
## VisualBertConfig
[[autodoc]] VisualBertConfig
## VisualBertModel
[[autodoc]] VisualBertModel
- forward
## VisualBertForPreTraining
[[autodoc]] VisualBertForPreTraining
- forward
## VisualBertForQuestionAnswering
[[autodoc]] VisualBertForQuestionAnswering
- forward
## VisualBertForMultipleChoice
[[autodoc]] VisualBertForMultipleChoice
- forward
## VisualBertForVisualReasoning
[[autodoc]] VisualBertForVisualReasoning
- forward
## VisualBertForRegionToPhraseAlignment
[[autodoc]] VisualBertForRegionToPhraseAlignment
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/visual_bert.md |
!--Copyright 2022 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Nyströmformer
## Overview
The Nyströmformer model was proposed in [*Nyströmformer: A Nyström-Based Algorithm for Approximating Self-Attention*](https://arxiv.org/abs/2102.03902) by Yunyang Xiong, Zhanpeng Zeng, Rudrasis Chakraborty, Mingxing Tan, Glenn
Fung, Yin Li, and Vikas Singh.
The abstract from the paper is the following:
*Transformers have emerged as a powerful tool for a broad range of natural language processing tasks. A key component
that drives the impressive performance of Transformers is the self-attention mechanism that encodes the influence or
dependence of other tokens on each specific token. While beneficial, the quadratic complexity of self-attention on the
input sequence length has limited its application to longer sequences -- a topic being actively studied in the
community. To address this limitation, we propose Nyströmformer -- a model that exhibits favorable scalability as a
function of sequence length. Our idea is based on adapting the Nyström method to approximate standard self-attention
with O(n) complexity. The scalability of Nyströmformer enables application to longer sequences with thousands of
tokens. We perform evaluations on multiple downstream tasks on the GLUE benchmark and IMDB reviews with standard
sequence length, and find that our Nyströmformer performs comparably, or in a few cases, even slightly better, than
standard self-attention. On longer sequence tasks in the Long Range Arena (LRA) benchmark, Nyströmformer performs
favorably relative to other efficient self-attention methods. Our code is available at this https URL.*
This model was contributed by [novice03](https://huggingface.co/novice03). The original code can be found [here](https://github.com/mlpen/Nystromformer).
## Resources
- [Text classification task guide](../tasks/sequence_classification)
- [Token classification task guide](../tasks/token_classification)
- [Question answering task guide](../tasks/question_answering)
- [Masked language modeling task guide](../tasks/masked_language_modeling)
- [Multiple choice task guide](../tasks/multiple_choice)
## NystromformerConfig
[[autodoc]] NystromformerConfig
## NystromformerModel
[[autodoc]] NystromformerModel
- forward
## NystromformerForMaskedLM
[[autodoc]] NystromformerForMaskedLM
- forward
## NystromformerForSequenceClassification
[[autodoc]] NystromformerForSequenceClassification
- forward
## NystromformerForMultipleChoice
[[autodoc]] NystromformerForMultipleChoice
- forward
## NystromformerForTokenClassification
[[autodoc]] NystromformerForTokenClassification
- forward
## NystromformerForQuestionAnswering
[[autodoc]] NystromformerForQuestionAnswering
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/nystromformer.md |
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# SAM
## Overview
SAM (Segment Anything Model) was proposed in [Segment Anything](https://arxiv.org/pdf/2304.02643v1.pdf) by Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer Whitehead, Alex Berg, Wan-Yen Lo, Piotr Dollar, Ross Girshick.
The model can be used to predict segmentation masks of any object of interest given an input image.
The abstract from the paper is the following:
*We introduce the Segment Anything (SA) project: a new task, model, and dataset for image segmentation. Using our efficient model in a data collection loop, we built the largest segmentation dataset to date (by far), with over 1 billion masks on 11M licensed and privacy respecting images. The model is designed and trained to be promptable, so it can transfer zero-shot to new image distributions and tasks. We evaluate its capabilities on numerous tasks and find that its zero-shot performance is impressive -- often competitive with or even superior to prior fully supervised results. We are releasing the Segment Anything Model (SAM) and corresponding dataset (SA-1B) of 1B masks and 11M images at [https://segment-anything.com](https://segment-anything.com) to foster research into foundation models for computer vision.*
Tips:
- The model predicts binary masks that states the presence or not of the object of interest given an image.
- The model predicts much better results if input 2D points and/or input bounding boxes are provided
- You can prompt multiple points for the same image, and predict a single mask.
- Fine-tuning the model is not supported yet
- According to the paper, textual input should be also supported. However, at this time of writing this seems to be not supported according to [the official repository](https://github.com/facebookresearch/segment-anything/issues/4#issuecomment-1497626844).
This model was contributed by [ybelkada](https://huggingface.co/ybelkada) and [ArthurZ](https://huggingface.co/ArthurZ).
The original code can be found [here](https://github.com/facebookresearch/segment-anything).
Below is an example on how to run mask generation given an image and a 2D point:
```python
import torch
from PIL import Image
import requests
from transformers import SamModel, SamProcessor
device = "cuda" if torch.cuda.is_available() else "cpu"
model = SamModel.from_pretrained("facebook/sam-vit-huge").to(device)
processor = SamProcessor.from_pretrained("facebook/sam-vit-huge")
img_url = "https://huggingface.co/ybelkada/segment-anything/resolve/main/assets/car.png"
raw_image = Image.open(requests.get(img_url, stream=True).raw).convert("RGB")
input_points = [[[450, 600]]] # 2D location of a window in the image
inputs = processor(raw_image, input_points=input_points, return_tensors="pt").to(device)
with torch.no_grad():
outputs = model(**inputs)
masks = processor.image_processor.post_process_masks(
outputs.pred_masks.cpu(), inputs["original_sizes"].cpu(), inputs["reshaped_input_sizes"].cpu()
)
scores = outputs.iou_scores
```
Resources:
- [Demo notebook](https://github.com/huggingface/notebooks/blob/main/examples/segment_anything.ipynb) for using the model.
- [Demo notebook](https://github.com/huggingface/notebooks/blob/main/examples/automatic_mask_generation.ipynb) for using the automatic mask generation pipeline.
- [Demo notebook](https://github.com/NielsRogge/Transformers-Tutorials/blob/master/SAM/Run_inference_with_MedSAM_using_HuggingFace_Transformers.ipynb) for inference with MedSAM, a fine-tuned version of SAM on the medical domain.
- [Demo notebook](https://github.com/NielsRogge/Transformers-Tutorials/blob/master/SAM/Fine_tune_SAM_(segment_anything)_on_a_custom_dataset.ipynb) for fine-tuning the model on custom data.
## SamConfig
[[autodoc]] SamConfig
## SamVisionConfig
[[autodoc]] SamVisionConfig
## SamMaskDecoderConfig
[[autodoc]] SamMaskDecoderConfig
## SamPromptEncoderConfig
[[autodoc]] SamPromptEncoderConfig
## SamProcessor
[[autodoc]] SamProcessor
## SamImageProcessor
[[autodoc]] SamImageProcessor
## SamModel
[[autodoc]] SamModel
- forward
## TFSamModel
[[autodoc]] TFSamModel
- call
| huggingface/transformers/blob/main/docs/source/en/model_doc/sam.md |
!--Copyright 2022 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# UPerNet
## Overview
The UPerNet model was proposed in [Unified Perceptual Parsing for Scene Understanding](https://arxiv.org/abs/1807.10221)
by Tete Xiao, Yingcheng Liu, Bolei Zhou, Yuning Jiang, Jian Sun. UPerNet is a general framework to effectively segment
a wide range of concepts from images, leveraging any vision backbone like [ConvNeXt](convnext) or [Swin](swin).
The abstract from the paper is the following:
*Humans recognize the visual world at multiple levels: we effortlessly categorize scenes and detect objects inside, while also identifying the textures and surfaces of the objects along with their different compositional parts. In this paper, we study a new task called Unified Perceptual Parsing, which requires the machine vision systems to recognize as many visual concepts as possible from a given image. A multi-task framework called UPerNet and a training strategy are developed to learn from heterogeneous image annotations. We benchmark our framework on Unified Perceptual Parsing and show that it is able to effectively segment a wide range of concepts from images. The trained networks are further applied to discover visual knowledge in natural scenes.*
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/upernet_architecture.jpg"
alt="drawing" width="600"/>
<small> UPerNet framework. Taken from the <a href="https://arxiv.org/abs/1807.10221">original paper</a>. </small>
This model was contributed by [nielsr](https://huggingface.co/nielsr). The original code is based on OpenMMLab's mmsegmentation [here](https://github.com/open-mmlab/mmsegmentation/blob/master/mmseg/models/decode_heads/uper_head.py).
## Usage examples
UPerNet is a general framework for semantic segmentation. It can be used with any vision backbone, like so:
```py
from transformers import SwinConfig, UperNetConfig, UperNetForSemanticSegmentation
backbone_config = SwinConfig(out_features=["stage1", "stage2", "stage3", "stage4"])
config = UperNetConfig(backbone_config=backbone_config)
model = UperNetForSemanticSegmentation(config)
```
To use another vision backbone, like [ConvNeXt](convnext), simply instantiate the model with the appropriate backbone:
```py
from transformers import ConvNextConfig, UperNetConfig, UperNetForSemanticSegmentation
backbone_config = ConvNextConfig(out_features=["stage1", "stage2", "stage3", "stage4"])
config = UperNetConfig(backbone_config=backbone_config)
model = UperNetForSemanticSegmentation(config)
```
Note that this will randomly initialize all the weights of the model.
## Resources
A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with UPerNet.
- Demo notebooks for UPerNet can be found [here](https://github.com/NielsRogge/Transformers-Tutorials/tree/master/UPerNet).
- [`UperNetForSemanticSegmentation`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/semantic-segmentation) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/semantic_segmentation.ipynb).
- See also: [Semantic segmentation task guide](../tasks/semantic_segmentation)
If you're interested in submitting a resource to be included here, please feel free to open a Pull Request and we'll review it! The resource should ideally demonstrate something new instead of duplicating an existing resource.
## UperNetConfig
[[autodoc]] UperNetConfig
## UperNetForSemanticSegmentation
[[autodoc]] UperNetForSemanticSegmentation
- forward | huggingface/transformers/blob/main/docs/source/en/model_doc/upernet.md |
!--Copyright 2022 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# PoolFormer
## Overview
The PoolFormer model was proposed in [MetaFormer is Actually What You Need for Vision](https://arxiv.org/abs/2111.11418) by Sea AI Labs. Instead of designing complicated token mixer to achieve SOTA performance, the target of this work is to demonstrate the competence of transformer models largely stem from the general architecture MetaFormer.
The abstract from the paper is the following:
*Transformers have shown great potential in computer vision tasks. A common belief is their attention-based token mixer module contributes most to their competence. However, recent works show the attention-based module in transformers can be replaced by spatial MLPs and the resulted models still perform quite well. Based on this observation, we hypothesize that the general architecture of the transformers, instead of the specific token mixer module, is more essential to the model's performance. To verify this, we deliberately replace the attention module in transformers with an embarrassingly simple spatial pooling operator to conduct only the most basic token mixing. Surprisingly, we observe that the derived model, termed as PoolFormer, achieves competitive performance on multiple computer vision tasks. For example, on ImageNet-1K, PoolFormer achieves 82.1% top-1 accuracy, surpassing well-tuned vision transformer/MLP-like baselines DeiT-B/ResMLP-B24 by 0.3%/1.1% accuracy with 35%/52% fewer parameters and 48%/60% fewer MACs. The effectiveness of PoolFormer verifies our hypothesis and urges us to initiate the concept of "MetaFormer", a general architecture abstracted from transformers without specifying the token mixer. Based on the extensive experiments, we argue that MetaFormer is the key player in achieving superior results for recent transformer and MLP-like models on vision tasks. This work calls for more future research dedicated to improving MetaFormer instead of focusing on the token mixer modules. Additionally, our proposed PoolFormer could serve as a starting baseline for future MetaFormer architecture design.*
The figure below illustrates the architecture of PoolFormer. Taken from the [original paper](https://arxiv.org/abs/2111.11418).
<img width="600" src="https://user-images.githubusercontent.com/15921929/142746124-1ab7635d-2536-4a0e-ad43-b4fe2c5a525d.png"/>
This model was contributed by [heytanay](https://huggingface.co/heytanay). The original code can be found [here](https://github.com/sail-sg/poolformer).
## Usage tips
- PoolFormer has a hierarchical architecture, where instead of Attention, a simple Average Pooling layer is present. All checkpoints of the model can be found on the [hub](https://huggingface.co/models?other=poolformer).
- One can use [`PoolFormerImageProcessor`] to prepare images for the model.
- As most models, PoolFormer comes in different sizes, the details of which can be found in the table below.
| **Model variant** | **Depths** | **Hidden sizes** | **Params (M)** | **ImageNet-1k Top 1** |
| :---------------: | ------------- | ------------------- | :------------: | :-------------------: |
| s12 | [2, 2, 6, 2] | [64, 128, 320, 512] | 12 | 77.2 |
| s24 | [4, 4, 12, 4] | [64, 128, 320, 512] | 21 | 80.3 |
| s36 | [6, 6, 18, 6] | [64, 128, 320, 512] | 31 | 81.4 |
| m36 | [6, 6, 18, 6] | [96, 192, 384, 768] | 56 | 82.1 |
| m48 | [8, 8, 24, 8] | [96, 192, 384, 768] | 73 | 82.5 |
## Resources
A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with PoolFormer.
<PipelineTag pipeline="image-classification"/>
- [`PoolFormerForImageClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/image-classification) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/image_classification.ipynb).
- See also: [Image classification task guide](../tasks/image_classification)
If you're interested in submitting a resource to be included here, please feel free to open a Pull Request and we'll review it! The resource should ideally demonstrate something new instead of duplicating an existing resource.
## PoolFormerConfig
[[autodoc]] PoolFormerConfig
## PoolFormerFeatureExtractor
[[autodoc]] PoolFormerFeatureExtractor
- __call__
## PoolFormerImageProcessor
[[autodoc]] PoolFormerImageProcessor
- preprocess
## PoolFormerModel
[[autodoc]] PoolFormerModel
- forward
## PoolFormerForImageClassification
[[autodoc]] PoolFormerForImageClassification
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/poolformer.md |
!---
Copyright 2021 The Google Flax Team Authors and HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-->
# Token classification examples
Fine-tuning the library models for token classification task such as Named Entity Recognition (NER), Parts-of-speech tagging (POS) or phrase extraction (CHUNKS). The main script run_flax_ner.py leverages the 🤗 Datasets library. You can easily customize it to your needs if you need extra processing on your datasets.
It will either run on a datasets hosted on our hub or with your own text files for training and validation, you might just need to add some tweaks in the data preprocessing.
The following example fine-tunes BERT on CoNLL-2003:
```bash
python run_flax_ner.py \
--model_name_or_path bert-base-cased \
--dataset_name conll2003 \
--max_seq_length 128 \
--learning_rate 2e-5 \
--num_train_epochs 3 \
--per_device_train_batch_size 4 \
--output_dir ./bert-ner-conll2003 \
--eval_steps 300 \
--push_to_hub
```
Using the command above, the script will train for 3 epochs and run eval after each epoch.
Metrics and hyperparameters are stored in Tensorflow event files in `--output_dir`.
You can see the results by running `tensorboard` in that directory:
```bash
$ tensorboard --logdir .
```
or directly on the hub under *Training metrics*.
sample Metrics - [tfhub.dev](https://tensorboard.dev/experiment/u52qsBIpQSKEEXEJd2LVYA) | huggingface/transformers/blob/main/examples/flax/token-classification/README.md |
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Fully Sharded Data Parallel
[Fully Sharded Data Parallel (FSDP)](https://pytorch.org/blog/introducing-pytorch-fully-sharded-data-parallel-api/) is a data parallel method that shards a model's parameters, gradients and optimizer states across the number of available GPUs (also called workers or *rank*). Unlike [DistributedDataParallel (DDP)](https://pytorch.org/docs/stable/generated/torch.nn.parallel.DistributedDataParallel.html), FSDP reduces memory-usage because a model is replicated on each GPU. This improves GPU memory-efficiency and allows you to train much larger models on fewer GPUs. FSDP is integrated with the Accelerate, a library for easily managing training in distributed environments, which means it is available for use from the [`Trainer`] class.
Before you start, make sure Accelerate is installed and at least PyTorch 2.1.0 or newer.
```bash
pip install accelerate
```
## FSDP configuration
To start, run the [`accelerate config`](https://huggingface.co/docs/accelerate/package_reference/cli#accelerate-config) command to create a configuration file for your training environment. Accelerate uses this configuration file to automatically setup the correct training environment based on your selected training options in `accelerate config`.
```bash
accelerate config
```
When you run `accelerate config`, you'll be prompted with a series of options to configure your training environment. This section covers some of the most important FSDP options. To learn more about the other available FSDP options, take a look at the [fsdp_config](https://huggingface.co/docs/transformers/main_classes/trainer#transformers.TrainingArguments.fsdp_config) parameters.
### Sharding strategy
FSDP offers a number of sharding strategies to select from:
* `FULL_SHARD` - shards model parameters, gradients and optimizer states across workers; select `1` for this option
* `SHARD_GRAD_OP`- shard gradients and optimizer states across workers; select `2` for this option
* `NO_SHARD` - don't shard anything (this is equivalent to DDP); select `3` for this option
* `HYBRID_SHARD` - shard model parameters, gradients and optimizer states within each worker where each worker also has a full copy; select `4` for this option
* `HYBRID_SHARD_ZERO2` - shard gradients and optimizer states within each worker where each worker also has a full copy; select `5` for this option
This is enabled by the `fsdp_sharding_strategy` flag.
### CPU offload
You could also offload parameters and gradients when they are not in use to the CPU to save even more GPU memory and help you fit large models where even FSDP may not be sufficient. This is enabled by setting `fsdp_offload_params: true` when running `accelerate config`.
### Wrapping policy
FSDP is applied by wrapping each layer in the network. The wrapping is usually applied in a nested way where the full weights are discarded after each forward pass to save memory for use in the next layer. The *auto wrapping* policy is the simplest way to implement this and you don't need to change any code. You should select `fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP` to wrap a Transformer layer and `fsdp_transformer_layer_cls_to_wrap` to specify which layer to wrap (for example `BertLayer`).
Otherwise, you can choose a size-based wrapping policy where FSDP is applied to a layer if it exceeds a certain number of parameters. This is enabled by setting `fsdp_wrap_policy: SIZE_BASED_WRAP` and `min_num_param` to the desired size threshold.
### Checkpointing
Intermediate checkpoints should be saved with `fsdp_state_dict_type: SHARDED_STATE_DICT` because saving the full state dict with CPU offloading on rank 0 takes a lot of time and often results in `NCCL Timeout` errors due to indefinite hanging during broadcasting. You can resume training with the sharded state dicts with the [`~accelerate.Accelerator.load_state`]` method.
```py
# directory containing checkpoints
accelerator.load_state("ckpt")
```
However, when training ends, you want to save the full state dict because sharded state dict is only compatible with FSDP.
```py
if trainer.is_fsdp_enabled:
trainer.accelerator.state.fsdp_plugin.set_state_dict_type("FULL_STATE_DICT")
trainer.save_model(script_args.output_dir)
```
### TPU
[PyTorch XLA](https://pytorch.org/xla/release/2.1/index.html) supports FSDP training for TPUs and it can be enabled by modifying the FSDP configuration file generated by `accelerate config`. In addition to the sharding strategies and wrapping options specified above, you can add the parameters shown below to the file.
```yaml
xla: True # must be set to True to enable PyTorch/XLA
xla_fsdp_settings: # XLA-specific FSDP parameters
xla_fsdp_grad_ckpt: True # use gradient checkpointing
```
The [`xla_fsdp_settings`](https://github.com/pytorch/xla/blob/2e6e183e0724818f137c8135b34ef273dea33318/torch_xla/distributed/fsdp/xla_fully_sharded_data_parallel.py#L128) allow you to configure additional XLA-specific parameters for FSDP.
## Launch training
An example FSDP configuration file may look like:
```yaml
compute_environment: LOCAL_MACHINE
debug: false
distributed_type: FSDP
downcast_bf16: 'no'
fsdp_config:
fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
fsdp_backward_prefetch_policy: BACKWARD_PRE
fsdp_cpu_ram_efficient_loading: true
fsdp_forward_prefetch: false
fsdp_offload_params: true
fsdp_sharding_strategy: 1
fsdp_state_dict_type: SHARDED_STATE_DICT
fsdp_sync_module_states: true
fsdp_transformer_layer_cls_to_wrap: BertLayer
fsdp_use_orig_params: true
machine_rank: 0
main_training_function: main
mixed_precision: bf16
num_machines: 1
num_processes: 2
rdzv_backend: static
same_network: true
tpu_env: []
tpu_use_cluster: false
tpu_use_sudo: false
use_cpu: false
```
To launch training, run the [`accelerate launch`](https://huggingface.co/docs/accelerate/package_reference/cli#accelerate-launch) command and it'll automatically use the configuration file you previously created with `accelerate config`.
```bash
accelerate launch my-trainer-script.py
```
```bash
accelerate launch --fsdp="full shard" --fsdp_config="path/to/fsdp_config/ my-trainer-script.py
```
## Next steps
FSDP can be a powerful tool for training really large models and you have access to more than one GPU or TPU. By sharding the model parameters, optimizer and gradient states, and even offloading them to the CPU when they're inactive, FSDP can reduce the high cost of large-scale training. If you're interested in learning more, the following may be helpful:
* Follow along with the more in-depth Accelerate guide for [FSDP](https://huggingface.co/docs/accelerate/usage_guides/fsdp).
* Read the [Introducing PyTorch Fully Sharded Data Parallel (FSDP) API](https://pytorch.org/blog/introducing-pytorch-fully-sharded-data-parallel-api/) blog post.
* Read the [Scaling PyTorch models on Cloud TPUs with FSDP](https://pytorch.org/blog/scaling-pytorch-models-on-cloud-tpus-with-fsdp/) blog post.
| huggingface/transformers/blob/main/docs/source/en/fsdp.md |
Testing new Hugging Face Deep Learning Container.
This document explains the testing strategy for releasing the new Hugging Face Deep Learning Container. AWS maintains 14 days of currency with framework releases. Besides framework releases, AWS release train is bi-weekly on Monday. Code cutoff date for any changes is the Wednesday before release-Monday.
## Test Case 1: Releasing a New Version (Minor/Major) of 🤗 Transformers
### Requirements: Test should run on Release Candidate for new `transformers` release to validate the new release is compatible with the DLCs. To run these tests you need credentials for the HF SageMaker AWS Account. You can ask @philschmid or @n1t0 to get access.
### Run Tests:
Before we can run the tests we need to adjust the `requirements.txt` for PyTorch under `/tests/sagemaker/scripts/pytorch` and for TensorFlow under `/tests/sagemaker/scripts/pytorch`. We adjust the branch to the new RC-tag.
```
git+https://github.com/huggingface/[email protected] # install main or adjust ist with vX.X.X for installing version specific-transforms
```
After we adjusted the `requirements.txt` we can run Amazon SageMaker tests with:
```bash
AWS_PROFILE=<enter-your-profile> make test-sagemaker
```
These tests take around 10-15 minutes to finish. Preferably make a screenshot of the successfully ran tests.
### After Transformers Release:
After we have released the Release Candidate we need to create a PR at the [Deep Learning Container Repository](https://github.com/aws/deep-learning-containers).
**Creating the update PR:**
1. Update the two latest `buildspec.yaml` config for [PyTorch](https://github.com/aws/deep-learning-containers/tree/master/huggingface/pytorch) and [TensorFlow](https://github.com/aws/deep-learning-containers/tree/master/huggingface/tensorflow). The two latest `buildspec.yaml` are the `buildspec.yaml` without a version tag and the one with the highest framework version, e.g. `buildspec-1-7-1.yml` and not `buildspec-1-6.yml`.
To update the `buildspec.yaml` we need to adjust either the `transformers_version` or the `datasets_version` or both. Example for upgrading to `transformers 4.5.0` and `datasets 1.6.0`.
```yaml
account_id: &ACCOUNT_ID <set-$ACCOUNT_ID-in-environment>
region: ®ION <set-$REGION-in-environment>
base_framework: &BASE_FRAMEWORK pytorch
framework: &FRAMEWORK !join [ "huggingface_", *BASE_FRAMEWORK]
version: &VERSION 1.6.0
short_version: &SHORT_VERSION 1.6
repository_info:
training_repository: &TRAINING_REPOSITORY
image_type: &TRAINING_IMAGE_TYPE training
root: !join [ "huggingface/", *BASE_FRAMEWORK, "/", *TRAINING_IMAGE_TYPE ]
repository_name: &REPOSITORY_NAME !join ["pr", "-", "huggingface", "-", *BASE_FRAMEWORK, "-", *TRAINING_IMAGE_TYPE]
repository: &REPOSITORY !join [ *ACCOUNT_ID, .dkr.ecr., *REGION, .amazonaws.com/,
*REPOSITORY_NAME ]
images:
BuildHuggingFacePytorchGpuPy37Cu110TrainingDockerImage:
<<: *TRAINING_REPOSITORY
build: &HUGGINGFACE_PYTORCH_GPU_TRAINING_PY3 false
image_size_baseline: &IMAGE_SIZE_BASELINE 15000
device_type: &DEVICE_TYPE gpu
python_version: &DOCKER_PYTHON_VERSION py3
tag_python_version: &TAG_PYTHON_VERSION py36
cuda_version: &CUDA_VERSION cu110
os_version: &OS_VERSION ubuntu18.04
transformers_version: &TRANSFORMERS_VERSION 4.5.0 # this was adjusted from 4.4.2 to 4.5.0
datasets_version: &DATASETS_VERSION 1.6.0 # this was adjusted from 1.5.0 to 1.6.0
tag: !join [ *VERSION, '-', 'transformers', *TRANSFORMERS_VERSION, '-', *DEVICE_TYPE, '-', *TAG_PYTHON_VERSION, '-',
*CUDA_VERSION, '-', *OS_VERSION ]
docker_file: !join [ docker/, *SHORT_VERSION, /, *DOCKER_PYTHON_VERSION, /,
*CUDA_VERSION, /Dockerfile., *DEVICE_TYPE ]
```
2. In the PR comment describe what test, we ran and with which package versions. Here you can copy the table from [Current Tests](#current-tests).
2. In the PR comment describe what test we ran and with which framework versions. Here you can copy the table from [Current Tests](#current-tests). You can take a look at this [PR](https://github.com/aws/deep-learning-containers/pull/1016), which information are needed.
## Test Case 2: Releasing a New AWS Framework DLC
## Execute Tests
### Requirements:
AWS is going to release new DLCs for PyTorch and/or TensorFlow. The Tests should run on the new framework versions with current `transformers` release to validate the new framework release is compatible with the `transformers` version. To run these tests you need credentials for the HF SageMaker AWS Account. You can ask @philschmid or @n1t0 to get access. AWS will notify us with a new issue in the repository pointing to their framework upgrade PR.
### Run Tests:
Before we can run the tests we need to adjust the `requirements.txt` for Pytorch under `/tests/sagemaker/scripts/pytorch` and for Tensorflow under `/tests/sagemaker/scripts/pytorch`. We add the new framework version to it.
```
torch==1.8.1 # for pytorch
tensorflow-gpu==2.5.0 # for tensorflow
```
After we adjusted the `requirements.txt` we can run Amazon SageMaker tests with.
```bash
AWS_PROFILE=<enter-your-profile> make test-sagemaker
```
These tests take around 10-15 minutes to finish. Preferably make a screenshot of the successfully ran tests.
### After successful Tests:
After we have successfully run tests for the new framework version we need to create a PR at the [Deep Learning Container Repository](https://github.com/aws/deep-learning-containers).
**Creating the update PR:**
1. Create a new `buildspec.yaml` config for [PyTorch](https://github.com/aws/deep-learning-containers/tree/master/huggingface/pytorch) and [TensorFlow](https://github.com/aws/deep-learning-containers/tree/master/huggingface/tensorflow) and rename the old `buildspec.yaml` to `buildespec-x.x.x`, where `x.x.x` is the base framework version, e.g. if pytorch 1.6.0 is the latest version in `buildspec.yaml` the file should be renamed to `buildspec-yaml-1-6.yaml`.
To create the new `buildspec.yaml` we need to adjust the `version` and the `short_version`. Example for upgrading to `pytorch 1.7.1`.
```yaml
account_id: &ACCOUNT_ID <set-$ACCOUNT_ID-in-environment>
region: ®ION <set-$REGION-in-environment>
base_framework: &BASE_FRAMEWORK pytorch
framework: &FRAMEWORK !join [ "huggingface_", *BASE_FRAMEWORK]
version: &VERSION 1.7.1 # this was adjusted from 1.6.0 to 1.7.1
short_version: &SHORT_VERSION 1.7 # this was adjusted from 1.6 to 1.7
repository_info:
training_repository: &TRAINING_REPOSITORY
image_type: &TRAINING_IMAGE_TYPE training
root: !join [ "huggingface/", *BASE_FRAMEWORK, "/", *TRAINING_IMAGE_TYPE ]
repository_name: &REPOSITORY_NAME !join ["pr", "-", "huggingface", "-", *BASE_FRAMEWORK, "-", *TRAINING_IMAGE_TYPE]
repository: &REPOSITORY !join [ *ACCOUNT_ID, .dkr.ecr., *REGION, .amazonaws.com/,
*REPOSITORY_NAME ]
images:
BuildHuggingFacePytorchGpuPy37Cu110TrainingDockerImage:
<<: *TRAINING_REPOSITORY
build: &HUGGINGFACE_PYTORCH_GPU_TRAINING_PY3 false
image_size_baseline: &IMAGE_SIZE_BASELINE 15000
device_type: &DEVICE_TYPE gpu
python_version: &DOCKER_PYTHON_VERSION py3
tag_python_version: &TAG_PYTHON_VERSION py36
cuda_version: &CUDA_VERSION cu110
os_version: &OS_VERSION ubuntu18.04
transformers_version: &TRANSFORMERS_VERSION 4.4.2
datasets_version: &DATASETS_VERSION 1.5.0
tag: !join [ *VERSION, '-', 'transformers', *TRANSFORMERS_VERSION, '-', *DEVICE_TYPE, '-', *TAG_PYTHON_VERSION, '-',
*CUDA_VERSION, '-', *OS_VERSION ]
docker_file: !join [ docker/, *SHORT_VERSION, /, *DOCKER_PYTHON_VERSION, /,
*CUDA_VERSION, /Dockerfile., *DEVICE_TYPE ]
```
2. In the PR comment describe what test we ran and with which framework versions. Here you can copy the table from [Current Tests](#current-tests). You can take a look at this [PR](https://github.com/aws/deep-learning-containers/pull/1025), which information are needed.
## Current Tests
| ID | Description | Platform | #GPUS | Collected & evaluated metrics |
|-------------------------------------|-------------------------------------------------------------------|-----------------------------|-------|------------------------------------------|
| pytorch-transfromers-test-single | test bert finetuning using BERT fromtransformerlib+PT | SageMaker createTrainingJob | 1 | train_runtime, eval_accuracy & eval_loss |
| pytorch-transfromers-test-2-ddp | test bert finetuning using BERT from transformer lib+ PT DPP | SageMaker createTrainingJob | 16 | train_runtime, eval_accuracy & eval_loss |
| pytorch-transfromers-test-2-smd | test bert finetuning using BERT from transformer lib+ PT SM DDP | SageMaker createTrainingJob | 16 | train_runtime, eval_accuracy & eval_loss |
| pytorch-transfromers-test-1-smp | test roberta finetuning using BERT from transformer lib+ PT SM MP | SageMaker createTrainingJob | 8 | train_runtime, eval_accuracy & eval_loss |
| tensorflow-transfromers-test-single | Test bert finetuning using BERT from transformer lib+TF | SageMaker createTrainingJob | 1 | train_runtime, eval_accuracy & eval_loss |
| tensorflow-transfromers-test-2-smd | test bert finetuning using BERT from transformer lib+ TF SM DDP | SageMaker createTrainingJob | 16 | train_runtime, eval_accuracy & eval_loss |
| huggingface/transformers/blob/main/tests/sagemaker/README.md |
!--Copyright 2021 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Wav2Vec2Phoneme
## Overview
The Wav2Vec2Phoneme model was proposed in [Simple and Effective Zero-shot Cross-lingual Phoneme Recognition (Xu et al.,
2021](https://arxiv.org/abs/2109.11680) by Qiantong Xu, Alexei Baevski, Michael Auli.
The abstract from the paper is the following:
*Recent progress in self-training, self-supervised pretraining and unsupervised learning enabled well performing speech
recognition systems without any labeled data. However, in many cases there is labeled data available for related
languages which is not utilized by these methods. This paper extends previous work on zero-shot cross-lingual transfer
learning by fine-tuning a multilingually pretrained wav2vec 2.0 model to transcribe unseen languages. This is done by
mapping phonemes of the training languages to the target language using articulatory features. Experiments show that
this simple method significantly outperforms prior work which introduced task-specific architectures and used only part
of a monolingually pretrained model.*
Relevant checkpoints can be found under https://huggingface.co/models?other=phoneme-recognition.
This model was contributed by [patrickvonplaten](https://huggingface.co/patrickvonplaten)
The original code can be found [here](https://github.com/pytorch/fairseq/tree/master/fairseq/models/wav2vec).
## Usage tips
- Wav2Vec2Phoneme uses the exact same architecture as Wav2Vec2
- Wav2Vec2Phoneme is a speech model that accepts a float array corresponding to the raw waveform of the speech signal.
- Wav2Vec2Phoneme model was trained using connectionist temporal classification (CTC) so the model output has to be
decoded using [`Wav2Vec2PhonemeCTCTokenizer`].
- Wav2Vec2Phoneme can be fine-tuned on multiple language at once and decode unseen languages in a single forward pass
to a sequence of phonemes
- By default, the model outputs a sequence of phonemes. In order to transform the phonemes to a sequence of words one
should make use of a dictionary and language model.
<Tip>
Wav2Vec2Phoneme's architecture is based on the Wav2Vec2 model, for API reference, check out [`Wav2Vec2`](wav2vec2)'s documentation page
except for the tokenizer.
</Tip>
## Wav2Vec2PhonemeCTCTokenizer
[[autodoc]] Wav2Vec2PhonemeCTCTokenizer
- __call__
- batch_decode
- decode
- phonemize
| huggingface/transformers/blob/main/docs/source/en/model_doc/wav2vec2_phoneme.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Configuration
The base class [`PretrainedConfig`] implements the common methods for loading/saving a configuration
either from a local file or directory, or from a pretrained model configuration provided by the library (downloaded
from HuggingFace's AWS S3 repository).
Each derived config class implements model specific attributes. Common attributes present in all config classes are:
`hidden_size`, `num_attention_heads`, and `num_hidden_layers`. Text models further implement:
`vocab_size`.
## PretrainedConfig
[[autodoc]] PretrainedConfig
- push_to_hub
- all
| huggingface/transformers/blob/main/docs/source/en/main_classes/configuration.md |
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Falcon
## Overview
Falcon is a class of causal decoder-only models built by [TII](https://www.tii.ae/). The largest Falcon checkpoints
have been trained on >=1T tokens of text, with a particular emphasis on the [RefinedWeb](https://arxiv.org/abs/2306.01116)
corpus. They are made available under the Apache 2.0 license.
Falcon's architecture is modern and optimized for inference, with multi-query attention and support for efficient
attention variants like `FlashAttention`. Both 'base' models trained only as causal language models as well as
'instruct' models that have received further fine-tuning are available.
Falcon models are (as of 2023) some of the largest and most powerful open-source language models,
and consistently rank highly in the [OpenLLM leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard).
## Converting custom checkpoints
<Tip>
Falcon models were initially added to the Hugging Face Hub as custom code checkpoints. However, Falcon is now fully
supported in the Transformers library. If you fine-tuned a model from a custom code checkpoint, we recommend converting
your checkpoint to the new in-library format, as this should give significant improvements to stability and
performance, especially for generation, as well as removing the need to use `trust_remote_code=True`!
</Tip>
You can convert custom code checkpoints to full Transformers checkpoints using the `convert_custom_code_checkpoint.py`
script located in the
[Falcon model directory](https://github.com/huggingface/transformers/tree/main/src/transformers/models/falcon)
of the Transformers library. To use this script, simply call it with
`python convert_custom_code_checkpoint.py --checkpoint_dir my_model`. This will convert your checkpoint in-place, and
you can immediately load it from the directory afterwards with e.g. `from_pretrained()`. If your model hasn't been
uploaded to the Hub, we recommend making a backup before attempting the conversion, just in case!
## FalconConfig
[[autodoc]] FalconConfig
- all
## FalconModel
[[autodoc]] FalconModel
- forward
## FalconForCausalLM
[[autodoc]] FalconForCausalLM
- forward
## FalconForSequenceClassification
[[autodoc]] FalconForSequenceClassification
- forward
## FalconForTokenClassification
[[autodoc]] FalconForTokenClassification
- forward
## FalconForQuestionAnswering
[[autodoc]] FalconForQuestionAnswering
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/falcon.md |
!--Copyright 2022 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# GLPN
<Tip>
This is a recently introduced model so the API hasn't been tested extensively. There may be some bugs or slight
breaking changes to fix it in the future. If you see something strange, file a [Github Issue](https://github.com/huggingface/transformers/issues/new?assignees=&labels=&template=bug-report.md&title).
</Tip>
## Overview
The GLPN model was proposed in [Global-Local Path Networks for Monocular Depth Estimation with Vertical CutDepth](https://arxiv.org/abs/2201.07436) by Doyeon Kim, Woonghyun Ga, Pyungwhan Ahn, Donggyu Joo, Sehwan Chun, Junmo Kim.
GLPN combines [SegFormer](segformer)'s hierarchical mix-Transformer with a lightweight decoder for monocular depth estimation. The proposed decoder shows better performance than the previously proposed decoders, with considerably
less computational complexity.
The abstract from the paper is the following:
*Depth estimation from a single image is an important task that can be applied to various fields in computer vision, and has grown rapidly with the development of convolutional neural networks. In this paper, we propose a novel structure and training strategy for monocular depth estimation to further improve the prediction accuracy of the network. We deploy a hierarchical transformer encoder to capture and convey the global context, and design a lightweight yet powerful decoder to generate an estimated depth map while considering local connectivity. By constructing connected paths between multi-scale local features and the global decoding stream with our proposed selective feature fusion module, the network can integrate both representations and recover fine details. In addition, the proposed decoder shows better performance than the previously proposed decoders, with considerably less computational complexity. Furthermore, we improve the depth-specific augmentation method by utilizing an important observation in depth estimation to enhance the model. Our network achieves state-of-the-art performance over the challenging depth dataset NYU Depth V2. Extensive experiments have been conducted to validate and show the effectiveness of the proposed approach. Finally, our model shows better generalisation ability and robustness than other comparative models.*
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/glpn_architecture.jpg"
alt="drawing" width="600"/>
<small> Summary of the approach. Taken from the <a href="https://arxiv.org/abs/2201.07436" target="_blank">original paper</a>. </small>
This model was contributed by [nielsr](https://huggingface.co/nielsr). The original code can be found [here](https://github.com/vinvino02/GLPDepth).
## Resources
A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with GLPN.
- Demo notebooks for [`GLPNForDepthEstimation`] can be found [here](https://github.com/NielsRogge/Transformers-Tutorials/tree/master/GLPN).
- [Monocular depth estimation task guide](../tasks/monocular_depth_estimation)
## GLPNConfig
[[autodoc]] GLPNConfig
## GLPNFeatureExtractor
[[autodoc]] GLPNFeatureExtractor
- __call__
## GLPNImageProcessor
[[autodoc]] GLPNImageProcessor
- preprocess
## GLPNModel
[[autodoc]] GLPNModel
- forward
## GLPNForDepthEstimation
[[autodoc]] GLPNForDepthEstimation
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/glpn.md |
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# ALIGN
## Overview
The ALIGN model was proposed in [Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision](https://arxiv.org/abs/2102.05918) by Chao Jia, Yinfei Yang, Ye Xia, Yi-Ting Chen, Zarana Parekh, Hieu Pham, Quoc V. Le, Yunhsuan Sung, Zhen Li, Tom Duerig. ALIGN is a multi-modal vision and language model. It can be used for image-text similarity and for zero-shot image classification. ALIGN features a dual-encoder architecture with [EfficientNet](efficientnet) as its vision encoder and [BERT](bert) as its text encoder, and learns to align visual and text representations with contrastive learning. Unlike previous work, ALIGN leverages a massive noisy dataset and shows that the scale of the corpus can be used to achieve SOTA representations with a simple recipe.
The abstract from the paper is the following:
*Pre-trained representations are becoming crucial for many NLP and perception tasks. While representation learning in NLP has transitioned to training on raw text without human annotations, visual and vision-language representations still rely heavily on curated training datasets that are expensive or require expert knowledge. For vision applications, representations are mostly learned using datasets with explicit class labels such as ImageNet or OpenImages. For vision-language, popular datasets like Conceptual Captions, MSCOCO, or CLIP all involve a non-trivial data collection (and cleaning) process. This costly curation process limits the size of datasets and hence hinders the scaling of trained models. In this paper, we leverage a noisy dataset of over one billion image alt-text pairs, obtained without expensive filtering or post-processing steps in the Conceptual Captions dataset. A simple dual-encoder architecture learns to align visual and language representations of the image and text pairs using a contrastive loss. We show that the scale of our corpus can make up for its noise and leads to state-of-the-art representations even with such a simple learning scheme. Our visual representation achieves strong performance when transferred to classification tasks such as ImageNet and VTAB. The aligned visual and language representations enables zero-shot image classification and also set new state-of-the-art results on Flickr30K and MSCOCO image-text retrieval benchmarks, even when compared with more sophisticated cross-attention models. The representations also enable cross-modality search with complex text and text + image queries.*
This model was contributed by [Alara Dirik](https://huggingface.co/adirik).
The original code is not released, this implementation is based on the Kakao Brain implementation based on the original paper.
## Usage example
ALIGN uses EfficientNet to get visual features and BERT to get the text features. Both the text and visual features are then projected to a latent space with identical dimension. The dot product between the projected image and text features is then used as a similarity score.
[`AlignProcessor`] wraps [`EfficientNetImageProcessor`] and [`BertTokenizer`] into a single instance to both encode the text and preprocess the images. The following example shows how to get the image-text similarity scores using [`AlignProcessor`] and [`AlignModel`].
```python
import requests
import torch
from PIL import Image
from transformers import AlignProcessor, AlignModel
processor = AlignProcessor.from_pretrained("kakaobrain/align-base")
model = AlignModel.from_pretrained("kakaobrain/align-base")
url = "http://images.cocodataset.org/val2017/000000039769.jpg"
image = Image.open(requests.get(url, stream=True).raw)
candidate_labels = ["an image of a cat", "an image of a dog"]
inputs = processor(text=candidate_labels, images=image, return_tensors="pt")
with torch.no_grad():
outputs = model(**inputs)
# this is the image-text similarity score
logits_per_image = outputs.logits_per_image
# we can take the softmax to get the label probabilities
probs = logits_per_image.softmax(dim=1)
print(probs)
```
## Resources
A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with ALIGN.
- A blog post on [ALIGN and the COYO-700M dataset](https://huggingface.co/blog/vit-align).
- A zero-shot image classification [demo](https://huggingface.co/spaces/adirik/ALIGN-zero-shot-image-classification).
- [Model card](https://huggingface.co/kakaobrain/align-base) of `kakaobrain/align-base` model.
If you're interested in submitting a resource to be included here, please feel free to open a Pull Request and we will review it. The resource should ideally demonstrate something new instead of duplicating an existing resource.
## AlignConfig
[[autodoc]] AlignConfig
- from_text_vision_configs
## AlignTextConfig
[[autodoc]] AlignTextConfig
## AlignVisionConfig
[[autodoc]] AlignVisionConfig
## AlignProcessor
[[autodoc]] AlignProcessor
## AlignModel
[[autodoc]] AlignModel
- forward
- get_text_features
- get_image_features
## AlignTextModel
[[autodoc]] AlignTextModel
- forward
## AlignVisionModel
[[autodoc]] AlignVisionModel
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/align.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Philosophy
🤗 Transformers is an opinionated library built for:
- machine learning researchers and educators seeking to use, study or extend large-scale Transformers models.
- hands-on practitioners who want to fine-tune those models or serve them in production, or both.
- engineers who just want to download a pretrained model and use it to solve a given machine learning task.
The library was designed with two strong goals in mind:
1. Be as easy and fast to use as possible:
- We strongly limited the number of user-facing abstractions to learn, in fact, there are almost no abstractions,
just three standard classes required to use each model: [configuration](main_classes/configuration),
[models](main_classes/model), and a preprocessing class ([tokenizer](main_classes/tokenizer) for NLP, [image processor](main_classes/image_processor) for vision, [feature extractor](main_classes/feature_extractor) for audio, and [processor](main_classes/processors) for multimodal inputs).
- All of these classes can be initialized in a simple and unified way from pretrained instances by using a common
`from_pretrained()` method which downloads (if needed), caches and
loads the related class instance and associated data (configurations' hyperparameters, tokenizers' vocabulary,
and models' weights) from a pretrained checkpoint provided on [Hugging Face Hub](https://huggingface.co/models) or your own saved checkpoint.
- On top of those three base classes, the library provides two APIs: [`pipeline`] for quickly
using a model for inference on a given task and [`Trainer`] to quickly train or fine-tune a PyTorch model (all TensorFlow models are compatible with `Keras.fit`).
- As a consequence, this library is NOT a modular toolbox of building blocks for neural nets. If you want to
extend or build upon the library, just use regular Python, PyTorch, TensorFlow, Keras modules and inherit from the base
classes of the library to reuse functionalities like model loading and saving. If you'd like to learn more about our coding philosophy for models, check out our [Repeat Yourself](https://huggingface.co/blog/transformers-design-philosophy) blog post.
2. Provide state-of-the-art models with performances as close as possible to the original models:
- We provide at least one example for each architecture which reproduces a result provided by the official authors
of said architecture.
- The code is usually as close to the original code base as possible which means some PyTorch code may be not as
*pytorchic* as it could be as a result of being converted TensorFlow code and vice versa.
A few other goals:
- Expose the models' internals as consistently as possible:
- We give access, using a single API, to the full hidden-states and attention weights.
- The preprocessing classes and base model APIs are standardized to easily switch between models.
- Incorporate a subjective selection of promising tools for fine-tuning and investigating these models:
- A simple and consistent way to add new tokens to the vocabulary and embeddings for fine-tuning.
- Simple ways to mask and prune Transformer heads.
- Easily switch between PyTorch, TensorFlow 2.0 and Flax, allowing training with one framework and inference with another.
## Main concepts
The library is built around three types of classes for each model:
- **Model classes** can be PyTorch models ([torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module)), Keras models ([tf.keras.Model](https://www.tensorflow.org/api_docs/python/tf/keras/Model)) or JAX/Flax models ([flax.linen.Module](https://flax.readthedocs.io/en/latest/api_reference/flax.linen/module.html)) that work with the pretrained weights provided in the library.
- **Configuration classes** store the hyperparameters required to build a model (such as the number of layers and hidden size). You don't always need to instantiate these yourself. In particular, if you are using a pretrained model without any modification, creating the model will automatically take care of instantiating the configuration (which is part of the model).
- **Preprocessing classes** convert the raw data into a format accepted by the model. A [tokenizer](main_classes/tokenizer) stores the vocabulary for each model and provide methods for encoding and decoding strings in a list of token embedding indices to be fed to a model. [Image processors](main_classes/image_processor) preprocess vision inputs, [feature extractors](main_classes/feature_extractor) preprocess audio inputs, and a [processor](main_classes/processors) handles multimodal inputs.
All these classes can be instantiated from pretrained instances, saved locally, and shared on the Hub with three methods:
- `from_pretrained()` lets you instantiate a model, configuration, and preprocessing class from a pretrained version either
provided by the library itself (the supported models can be found on the [Model Hub](https://huggingface.co/models)) or
stored locally (or on a server) by the user.
- `save_pretrained()` lets you save a model, configuration, and preprocessing class locally so that it can be reloaded using
`from_pretrained()`.
- `push_to_hub()` lets you share a model, configuration, and a preprocessing class to the Hub, so it is easily accessible to everyone.
| huggingface/transformers/blob/main/docs/source/en/philosophy.md |
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Preprocess
[[open-in-colab]]
Before you can train a model on a dataset, it needs to be preprocessed into the expected model input format. Whether your data is text, images, or audio, they need to be converted and assembled into batches of tensors. 🤗 Transformers provides a set of preprocessing classes to help prepare your data for the model. In this tutorial, you'll learn that for:
* Text, use a [Tokenizer](./main_classes/tokenizer) to convert text into a sequence of tokens, create a numerical representation of the tokens, and assemble them into tensors.
* Speech and audio, use a [Feature extractor](./main_classes/feature_extractor) to extract sequential features from audio waveforms and convert them into tensors.
* Image inputs use a [ImageProcessor](./main_classes/image) to convert images into tensors.
* Multimodal inputs, use a [Processor](./main_classes/processors) to combine a tokenizer and a feature extractor or image processor.
<Tip>
`AutoProcessor` **always** works and automatically chooses the correct class for the model you're using, whether you're using a tokenizer, image processor, feature extractor or processor.
</Tip>
Before you begin, install 🤗 Datasets so you can load some datasets to experiment with:
```bash
pip install datasets
```
## Natural Language Processing
<Youtube id="Yffk5aydLzg"/>
The main tool for preprocessing textual data is a [tokenizer](main_classes/tokenizer). A tokenizer splits text into *tokens* according to a set of rules. The tokens are converted into numbers and then tensors, which become the model inputs. Any additional inputs required by the model are added by the tokenizer.
<Tip>
If you plan on using a pretrained model, it's important to use the associated pretrained tokenizer. This ensures the text is split the same way as the pretraining corpus, and uses the same corresponding tokens-to-index (usually referred to as the *vocab*) during pretraining.
</Tip>
Get started by loading a pretrained tokenizer with the [`AutoTokenizer.from_pretrained`] method. This downloads the *vocab* a model was pretrained with:
```py
>>> from transformers import AutoTokenizer
>>> tokenizer = AutoTokenizer.from_pretrained("bert-base-cased")
```
Then pass your text to the tokenizer:
```py
>>> encoded_input = tokenizer("Do not meddle in the affairs of wizards, for they are subtle and quick to anger.")
>>> print(encoded_input)
{'input_ids': [101, 2079, 2025, 19960, 10362, 1999, 1996, 3821, 1997, 16657, 1010, 2005, 2027, 2024, 11259, 1998, 4248, 2000, 4963, 1012, 102],
'token_type_ids': [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
'attention_mask': [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]}
```
The tokenizer returns a dictionary with three important items:
* [input_ids](glossary#input-ids) are the indices corresponding to each token in the sentence.
* [attention_mask](glossary#attention-mask) indicates whether a token should be attended to or not.
* [token_type_ids](glossary#token-type-ids) identifies which sequence a token belongs to when there is more than one sequence.
Return your input by decoding the `input_ids`:
```py
>>> tokenizer.decode(encoded_input["input_ids"])
'[CLS] Do not meddle in the affairs of wizards, for they are subtle and quick to anger. [SEP]'
```
As you can see, the tokenizer added two special tokens - `CLS` and `SEP` (classifier and separator) - to the sentence. Not all models need
special tokens, but if they do, the tokenizer automatically adds them for you.
If there are several sentences you want to preprocess, pass them as a list to the tokenizer:
```py
>>> batch_sentences = [
... "But what about second breakfast?",
... "Don't think he knows about second breakfast, Pip.",
... "What about elevensies?",
... ]
>>> encoded_inputs = tokenizer(batch_sentences)
>>> print(encoded_inputs)
{'input_ids': [[101, 1252, 1184, 1164, 1248, 6462, 136, 102],
[101, 1790, 112, 189, 1341, 1119, 3520, 1164, 1248, 6462, 117, 21902, 1643, 119, 102],
[101, 1327, 1164, 5450, 23434, 136, 102]],
'token_type_ids': [[0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0]],
'attention_mask': [[1, 1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1]]}
```
### Pad
Sentences aren't always the same length which can be an issue because tensors, the model inputs, need to have a uniform shape. Padding is a strategy for ensuring tensors are rectangular by adding a special *padding token* to shorter sentences.
Set the `padding` parameter to `True` to pad the shorter sequences in the batch to match the longest sequence:
```py
>>> batch_sentences = [
... "But what about second breakfast?",
... "Don't think he knows about second breakfast, Pip.",
... "What about elevensies?",
... ]
>>> encoded_input = tokenizer(batch_sentences, padding=True)
>>> print(encoded_input)
{'input_ids': [[101, 1252, 1184, 1164, 1248, 6462, 136, 102, 0, 0, 0, 0, 0, 0, 0],
[101, 1790, 112, 189, 1341, 1119, 3520, 1164, 1248, 6462, 117, 21902, 1643, 119, 102],
[101, 1327, 1164, 5450, 23434, 136, 102, 0, 0, 0, 0, 0, 0, 0, 0]],
'token_type_ids': [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]],
'attention_mask': [[1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0],
[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0]]}
```
The first and third sentences are now padded with `0`'s because they are shorter.
### Truncation
On the other end of the spectrum, sometimes a sequence may be too long for a model to handle. In this case, you'll need to truncate the sequence to a shorter length.
Set the `truncation` parameter to `True` to truncate a sequence to the maximum length accepted by the model:
```py
>>> batch_sentences = [
... "But what about second breakfast?",
... "Don't think he knows about second breakfast, Pip.",
... "What about elevensies?",
... ]
>>> encoded_input = tokenizer(batch_sentences, padding=True, truncation=True)
>>> print(encoded_input)
{'input_ids': [[101, 1252, 1184, 1164, 1248, 6462, 136, 102, 0, 0, 0, 0, 0, 0, 0],
[101, 1790, 112, 189, 1341, 1119, 3520, 1164, 1248, 6462, 117, 21902, 1643, 119, 102],
[101, 1327, 1164, 5450, 23434, 136, 102, 0, 0, 0, 0, 0, 0, 0, 0]],
'token_type_ids': [[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]],
'attention_mask': [[1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0],
[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0]]}
```
<Tip>
Check out the [Padding and truncation](./pad_truncation) concept guide to learn more different padding and truncation arguments.
</Tip>
### Build tensors
Finally, you want the tokenizer to return the actual tensors that get fed to the model.
Set the `return_tensors` parameter to either `pt` for PyTorch, or `tf` for TensorFlow:
<frameworkcontent>
<pt>
```py
>>> batch_sentences = [
... "But what about second breakfast?",
... "Don't think he knows about second breakfast, Pip.",
... "What about elevensies?",
... ]
>>> encoded_input = tokenizer(batch_sentences, padding=True, truncation=True, return_tensors="pt")
>>> print(encoded_input)
{'input_ids': tensor([[101, 1252, 1184, 1164, 1248, 6462, 136, 102, 0, 0, 0, 0, 0, 0, 0],
[101, 1790, 112, 189, 1341, 1119, 3520, 1164, 1248, 6462, 117, 21902, 1643, 119, 102],
[101, 1327, 1164, 5450, 23434, 136, 102, 0, 0, 0, 0, 0, 0, 0, 0]]),
'token_type_ids': tensor([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]),
'attention_mask': tensor([[1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0],
[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0]])}
```
</pt>
<tf>
```py
>>> batch_sentences = [
... "But what about second breakfast?",
... "Don't think he knows about second breakfast, Pip.",
... "What about elevensies?",
... ]
>>> encoded_input = tokenizer(batch_sentences, padding=True, truncation=True, return_tensors="tf")
>>> print(encoded_input)
{'input_ids': <tf.Tensor: shape=(2, 9), dtype=int32, numpy=
array([[101, 1252, 1184, 1164, 1248, 6462, 136, 102, 0, 0, 0, 0, 0, 0, 0],
[101, 1790, 112, 189, 1341, 1119, 3520, 1164, 1248, 6462, 117, 21902, 1643, 119, 102],
[101, 1327, 1164, 5450, 23434, 136, 102, 0, 0, 0, 0, 0, 0, 0, 0]],
dtype=int32)>,
'token_type_ids': <tf.Tensor: shape=(2, 9), dtype=int32, numpy=
array([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]], dtype=int32)>,
'attention_mask': <tf.Tensor: shape=(2, 9), dtype=int32, numpy=
array([[1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0],
[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
[1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0]], dtype=int32)>}
```
</tf>
</frameworkcontent>
## Audio
For audio tasks, you'll need a [feature extractor](main_classes/feature_extractor) to prepare your dataset for the model. The feature extractor is designed to extract features from raw audio data, and convert them into tensors.
Load the [MInDS-14](https://huggingface.co/datasets/PolyAI/minds14) dataset (see the 🤗 [Datasets tutorial](https://huggingface.co/docs/datasets/load_hub) for more details on how to load a dataset) to see how you can use a feature extractor with audio datasets:
```py
>>> from datasets import load_dataset, Audio
>>> dataset = load_dataset("PolyAI/minds14", name="en-US", split="train")
```
Access the first element of the `audio` column to take a look at the input. Calling the `audio` column automatically loads and resamples the audio file:
```py
>>> dataset[0]["audio"]
{'array': array([ 0. , 0.00024414, -0.00024414, ..., -0.00024414,
0. , 0. ], dtype=float32),
'path': '/root/.cache/huggingface/datasets/downloads/extracted/f14948e0e84be638dd7943ac36518a4cf3324e8b7aa331c5ab11541518e9368c/en-US~JOINT_ACCOUNT/602ba55abb1e6d0fbce92065.wav',
'sampling_rate': 8000}
```
This returns three items:
* `array` is the speech signal loaded - and potentially resampled - as a 1D array.
* `path` points to the location of the audio file.
* `sampling_rate` refers to how many data points in the speech signal are measured per second.
For this tutorial, you'll use the [Wav2Vec2](https://huggingface.co/facebook/wav2vec2-base) model. Take a look at the model card, and you'll learn Wav2Vec2 is pretrained on 16kHz sampled speech audio. It is important your audio data's sampling rate matches the sampling rate of the dataset used to pretrain the model. If your data's sampling rate isn't the same, then you need to resample your data.
1. Use 🤗 Datasets' [`~datasets.Dataset.cast_column`] method to upsample the sampling rate to 16kHz:
```py
>>> dataset = dataset.cast_column("audio", Audio(sampling_rate=16_000))
```
2. Call the `audio` column again to resample the audio file:
```py
>>> dataset[0]["audio"]
{'array': array([ 2.3443763e-05, 2.1729663e-04, 2.2145823e-04, ...,
3.8356509e-05, -7.3497440e-06, -2.1754686e-05], dtype=float32),
'path': '/root/.cache/huggingface/datasets/downloads/extracted/f14948e0e84be638dd7943ac36518a4cf3324e8b7aa331c5ab11541518e9368c/en-US~JOINT_ACCOUNT/602ba55abb1e6d0fbce92065.wav',
'sampling_rate': 16000}
```
Next, load a feature extractor to normalize and pad the input. When padding textual data, a `0` is added for shorter sequences. The same idea applies to audio data. The feature extractor adds a `0` - interpreted as silence - to `array`.
Load the feature extractor with [`AutoFeatureExtractor.from_pretrained`]:
```py
>>> from transformers import AutoFeatureExtractor
>>> feature_extractor = AutoFeatureExtractor.from_pretrained("facebook/wav2vec2-base")
```
Pass the audio `array` to the feature extractor. We also recommend adding the `sampling_rate` argument in the feature extractor in order to better debug any silent errors that may occur.
```py
>>> audio_input = [dataset[0]["audio"]["array"]]
>>> feature_extractor(audio_input, sampling_rate=16000)
{'input_values': [array([ 3.8106556e-04, 2.7506407e-03, 2.8015103e-03, ...,
5.6335266e-04, 4.6588284e-06, -1.7142107e-04], dtype=float32)]}
```
Just like the tokenizer, you can apply padding or truncation to handle variable sequences in a batch. Take a look at the sequence length of these two audio samples:
```py
>>> dataset[0]["audio"]["array"].shape
(173398,)
>>> dataset[1]["audio"]["array"].shape
(106496,)
```
Create a function to preprocess the dataset so the audio samples are the same lengths. Specify a maximum sample length, and the feature extractor will either pad or truncate the sequences to match it:
```py
>>> def preprocess_function(examples):
... audio_arrays = [x["array"] for x in examples["audio"]]
... inputs = feature_extractor(
... audio_arrays,
... sampling_rate=16000,
... padding=True,
... max_length=100000,
... truncation=True,
... )
... return inputs
```
Apply the `preprocess_function` to the first few examples in the dataset:
```py
>>> processed_dataset = preprocess_function(dataset[:5])
```
The sample lengths are now the same and match the specified maximum length. You can pass your processed dataset to the model now!
```py
>>> processed_dataset["input_values"][0].shape
(100000,)
>>> processed_dataset["input_values"][1].shape
(100000,)
```
## Computer vision
For computer vision tasks, you'll need an [image processor](main_classes/image_processor) to prepare your dataset for the model.
Image preprocessing consists of several steps that convert images into the input expected by the model. These steps
include but are not limited to resizing, normalizing, color channel correction, and converting images to tensors.
<Tip>
Image preprocessing often follows some form of image augmentation. Both image preprocessing and image augmentation
transform image data, but they serve different purposes:
* Image augmentation alters images in a way that can help prevent overfitting and increase the robustness of the model. You can get creative in how you augment your data - adjust brightness and colors, crop, rotate, resize, zoom, etc. However, be mindful not to change the meaning of the images with your augmentations.
* Image preprocessing guarantees that the images match the model’s expected input format. When fine-tuning a computer vision model, images must be preprocessed exactly as when the model was initially trained.
You can use any library you like for image augmentation. For image preprocessing, use the `ImageProcessor` associated with the model.
</Tip>
Load the [food101](https://huggingface.co/datasets/food101) dataset (see the 🤗 [Datasets tutorial](https://huggingface.co/docs/datasets/load_hub) for more details on how to load a dataset) to see how you can use an image processor with computer vision datasets:
<Tip>
Use 🤗 Datasets `split` parameter to only load a small sample from the training split since the dataset is quite large!
</Tip>
```py
>>> from datasets import load_dataset
>>> dataset = load_dataset("food101", split="train[:100]")
```
Next, take a look at the image with 🤗 Datasets [`Image`](https://huggingface.co/docs/datasets/package_reference/main_classes?highlight=image#datasets.Image) feature:
```py
>>> dataset[0]["image"]
```
<div class="flex justify-center">
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/vision-preprocess-tutorial.png"/>
</div>
Load the image processor with [`AutoImageProcessor.from_pretrained`]:
```py
>>> from transformers import AutoImageProcessor
>>> image_processor = AutoImageProcessor.from_pretrained("google/vit-base-patch16-224")
```
First, let's add some image augmentation. You can use any library you prefer, but in this tutorial, we'll use torchvision's [`transforms`](https://pytorch.org/vision/stable/transforms.html) module. If you're interested in using another data augmentation library, learn how in the [Albumentations](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/image_classification_albumentations.ipynb) or [Kornia notebooks](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/image_classification_kornia.ipynb).
1. Here we use [`Compose`](https://pytorch.org/vision/master/generated/torchvision.transforms.Compose.html) to chain together a couple of
transforms - [`RandomResizedCrop`](https://pytorch.org/vision/main/generated/torchvision.transforms.RandomResizedCrop.html) and [`ColorJitter`](https://pytorch.org/vision/main/generated/torchvision.transforms.ColorJitter.html).
Note that for resizing, we can get the image size requirements from the `image_processor`. For some models, an exact height and
width are expected, for others only the `shortest_edge` is defined.
```py
>>> from torchvision.transforms import RandomResizedCrop, ColorJitter, Compose
>>> size = (
... image_processor.size["shortest_edge"]
... if "shortest_edge" in image_processor.size
... else (image_processor.size["height"], image_processor.size["width"])
... )
>>> _transforms = Compose([RandomResizedCrop(size), ColorJitter(brightness=0.5, hue=0.5)])
```
2. The model accepts [`pixel_values`](model_doc/visionencoderdecoder#transformers.VisionEncoderDecoderModel.forward.pixel_values)
as its input. `ImageProcessor` can take care of normalizing the images, and generating appropriate tensors.
Create a function that combines image augmentation and image preprocessing for a batch of images and generates `pixel_values`:
```py
>>> def transforms(examples):
... images = [_transforms(img.convert("RGB")) for img in examples["image"]]
... examples["pixel_values"] = image_processor(images, do_resize=False, return_tensors="pt")["pixel_values"]
... return examples
```
<Tip>
In the example above we set `do_resize=False` because we have already resized the images in the image augmentation transformation,
and leveraged the `size` attribute from the appropriate `image_processor`. If you do not resize images during image augmentation,
leave this parameter out. By default, `ImageProcessor` will handle the resizing.
If you wish to normalize images as a part of the augmentation transformation, use the `image_processor.image_mean`,
and `image_processor.image_std` values.
</Tip>
3. Then use 🤗 Datasets[`~datasets.Dataset.set_transform`] to apply the transforms on the fly:
```py
>>> dataset.set_transform(transforms)
```
4. Now when you access the image, you'll notice the image processor has added `pixel_values`. You can pass your processed dataset to the model now!
```py
>>> dataset[0].keys()
```
Here is what the image looks like after the transforms are applied. The image has been randomly cropped and it's color properties are different.
```py
>>> import numpy as np
>>> import matplotlib.pyplot as plt
>>> img = dataset[0]["pixel_values"]
>>> plt.imshow(img.permute(1, 2, 0))
```
<div class="flex justify-center">
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/preprocessed_image.png"/>
</div>
<Tip>
For tasks like object detection, semantic segmentation, instance segmentation, and panoptic segmentation, `ImageProcessor`
offers post processing methods. These methods convert model's raw outputs into meaningful predictions such as bounding boxes,
or segmentation maps.
</Tip>
### Pad
In some cases, for instance, when fine-tuning [DETR](./model_doc/detr), the model applies scale augmentation at training
time. This may cause images to be different sizes in a batch. You can use [`DetrImageProcessor.pad`]
from [`DetrImageProcessor`] and define a custom `collate_fn` to batch images together.
```py
>>> def collate_fn(batch):
... pixel_values = [item["pixel_values"] for item in batch]
... encoding = image_processor.pad(pixel_values, return_tensors="pt")
... labels = [item["labels"] for item in batch]
... batch = {}
... batch["pixel_values"] = encoding["pixel_values"]
... batch["pixel_mask"] = encoding["pixel_mask"]
... batch["labels"] = labels
... return batch
```
## Multimodal
For tasks involving multimodal inputs, you'll need a [processor](main_classes/processors) to prepare your dataset for the model. A processor couples together two processing objects such as as tokenizer and feature extractor.
Load the [LJ Speech](https://huggingface.co/datasets/lj_speech) dataset (see the 🤗 [Datasets tutorial](https://huggingface.co/docs/datasets/load_hub) for more details on how to load a dataset) to see how you can use a processor for automatic speech recognition (ASR):
```py
>>> from datasets import load_dataset
>>> lj_speech = load_dataset("lj_speech", split="train")
```
For ASR, you're mainly focused on `audio` and `text` so you can remove the other columns:
```py
>>> lj_speech = lj_speech.map(remove_columns=["file", "id", "normalized_text"])
```
Now take a look at the `audio` and `text` columns:
```py
>>> lj_speech[0]["audio"]
{'array': array([-7.3242188e-04, -7.6293945e-04, -6.4086914e-04, ...,
7.3242188e-04, 2.1362305e-04, 6.1035156e-05], dtype=float32),
'path': '/root/.cache/huggingface/datasets/downloads/extracted/917ece08c95cf0c4115e45294e3cd0dee724a1165b7fc11798369308a465bd26/LJSpeech-1.1/wavs/LJ001-0001.wav',
'sampling_rate': 22050}
>>> lj_speech[0]["text"]
'Printing, in the only sense with which we are at present concerned, differs from most if not from all the arts and crafts represented in the Exhibition'
```
Remember you should always [resample](preprocessing#audio) your audio dataset's sampling rate to match the sampling rate of the dataset used to pretrain a model!
```py
>>> lj_speech = lj_speech.cast_column("audio", Audio(sampling_rate=16_000))
```
Load a processor with [`AutoProcessor.from_pretrained`]:
```py
>>> from transformers import AutoProcessor
>>> processor = AutoProcessor.from_pretrained("facebook/wav2vec2-base-960h")
```
1. Create a function to process the audio data contained in `array` to `input_values`, and tokenize `text` to `labels`. These are the inputs to the model:
```py
>>> def prepare_dataset(example):
... audio = example["audio"]
... example.update(processor(audio=audio["array"], text=example["text"], sampling_rate=16000))
... return example
```
2. Apply the `prepare_dataset` function to a sample:
```py
>>> prepare_dataset(lj_speech[0])
```
The processor has now added `input_values` and `labels`, and the sampling rate has also been correctly downsampled to 16kHz. You can pass your processed dataset to the model now!
| huggingface/transformers/blob/main/docs/source/en/preprocessing.md |
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# LLM prompting guide
[[open-in-colab]]
Large Language Models such as Falcon, LLaMA, etc. are pretrained transformer models initially trained to predict the
next token given some input text. They typically have billions of parameters and have been trained on trillions of
tokens for an extended period of time. As a result, these models become quite powerful and versatile, and you can use
them to solve multiple NLP tasks out of the box by instructing the models with natural language prompts.
Designing such prompts to ensure the optimal output is often called "prompt engineering". Prompt engineering is an
iterative process that requires a fair amount of experimentation. Natural languages are much more flexible and expressive
than programming languages, however, they can also introduce some ambiguity. At the same time, prompts in natural language
are quite sensitive to changes. Even minor modifications in prompts can lead to wildly different outputs.
While there is no exact recipe for creating prompts to match all cases, researchers have worked out a number of best
practices that help to achieve optimal results more consistently.
This guide covers the prompt engineering best practices to help you craft better LLM prompts and solve various NLP tasks.
You'll learn:
- [Basics of prompting](#basic-prompts)
- [Best practices of LLM prompting](#best-practices-of-llm-prompting)
- [Advanced prompting techniques: few-shot prompting and chain-of-thought](#advanced-prompting-techniques)
- [When to fine-tune instead of prompting](#prompting-vs-fine-tuning)
<Tip>
Prompt engineering is only a part of the LLM output optimization process. Another essential component is choosing the
optimal text generation strategy. You can customize how your LLM selects each of the subsequent tokens when generating
the text without modifying any of the trainable parameters. By tweaking the text generation parameters, you can reduce
repetition in the generated text and make it more coherent and human-sounding.
Text generation strategies and parameters are out of scope for this guide, but you can learn more about these topics in
the following guides:
* [Generation with LLMs](../llm_tutorial)
* [Text generation strategies](../generation_strategies)
</Tip>
## Basics of prompting
### Types of models
The majority of modern LLMs are decoder-only transformers. Some examples include: [LLaMA](../model_doc/llama),
[Llama2](../model_doc/llama2), [Falcon](../model_doc/falcon), [GPT2](../model_doc/gpt2). However, you may encounter
encoder-decoder transformer LLMs as well, for instance, [Flan-T5](../model_doc/flan-t5) and [BART](../model_doc/bart).
Encoder-decoder-style models are typically used in generative tasks where the output **heavily** relies on the input, for
example, in translation and summarization. The decoder-only models are used for all other types of generative tasks.
When using a pipeline to generate text with an LLM, it's important to know what type of LLM you are using, because
they use different pipelines.
Run inference with decoder-only models with the `text-generation` pipeline:
```python
>>> from transformers import pipeline
>>> import torch
>>> torch.manual_seed(0) # doctest: +IGNORE_RESULT
>>> generator = pipeline('text-generation', model = 'gpt2')
>>> prompt = "Hello, I'm a language model"
>>> generator(prompt, max_length = 30)
[{'generated_text': "Hello, I'm a language model expert, so I'm a big believer in the concept that I know very well and then I try to look into"}]
```
To run inference with an encoder-decoder, use the `text2text-generation` pipeline:
```python
>>> text2text_generator = pipeline("text2text-generation", model = 'google/flan-t5-base')
>>> prompt = "Translate from English to French: I'm very happy to see you"
>>> text2text_generator(prompt)
[{'generated_text': 'Je suis très heureuse de vous rencontrer.'}]
```
### Base vs instruct/chat models
Most of the recent LLM checkpoints available on 🤗 Hub come in two versions: base and instruct (or chat). For example,
[`tiiuae/falcon-7b`](https://huggingface.co/tiiuae/falcon-7b) and [`tiiuae/falcon-7b-instruct`](https://huggingface.co/tiiuae/falcon-7b-instruct).
Base models are excellent at completing the text when given an initial prompt, however, they are not ideal for NLP tasks
where they need to follow instructions, or for conversational use. This is where the instruct (chat) versions come in.
These checkpoints are the result of further fine-tuning of the pre-trained base versions on instructions and conversational data.
This additional fine-tuning makes them a better choice for many NLP tasks.
Let's illustrate some simple prompts that you can use with [`tiiuae/falcon-7b-instruct`](https://huggingface.co/tiiuae/falcon-7b-instruct)
to solve some common NLP tasks.
### NLP tasks
First, let's set up the environment:
```bash
pip install -q transformers accelerate
```
Next, let's load the model with the appropriate pipeline (`"text-generation"`):
```python
>>> from transformers import pipeline, AutoTokenizer
>>> import torch
>>> torch.manual_seed(0) # doctest: +IGNORE_RESULT
>>> model = "tiiuae/falcon-7b-instruct"
>>> tokenizer = AutoTokenizer.from_pretrained(model)
>>> pipe = pipeline(
... "text-generation",
... model=model,
... tokenizer=tokenizer,
... torch_dtype=torch.bfloat16,
... device_map="auto",
... )
```
<Tip>
Note that Falcon models were trained using the `bfloat16` datatype, so we recommend you use the same. This requires a recent
version of CUDA and works best on modern cards.
</Tip>
Now that we have the model loaded via the pipeline, let's explore how you can use prompts to solve NLP tasks.
#### Text classification
One of the most common forms of text classification is sentiment analysis, which assigns a label like "positive", "negative",
or "neutral" to a sequence of text. Let's write a prompt that instructs the model to classify a given text (a movie review).
We'll start by giving the instruction, and then specifying the text to classify. Note that instead of leaving it at that, we're
also adding the beginning of the response - `"Sentiment: "`:
```python
>>> torch.manual_seed(0) # doctest: +IGNORE_RESULT
>>> prompt = """Classify the text into neutral, negative or positive.
... Text: This movie is definitely one of my favorite movies of its kind. The interaction between respectable and morally strong characters is an ode to chivalry and the honor code amongst thieves and policemen.
... Sentiment:
... """
>>> sequences = pipe(
... prompt,
... max_new_tokens=10,
... )
>>> for seq in sequences:
... print(f"Result: {seq['generated_text']}")
Result: Classify the text into neutral, negative or positive.
Text: This movie is definitely one of my favorite movies of its kind. The interaction between respectable and morally strong characters is an ode to chivalry and the honor code amongst thieves and policemen.
Sentiment:
Positive
```
As a result, the output contains a classification label from the list we have provided in the instructions, and it is a correct one!
<Tip>
You may notice that in addition to the prompt, we pass a `max_new_tokens` parameter. It controls the number of tokens the
model shall generate, and it is one of the many text generation parameters that you can learn about
in [Text generation strategies](../generation_strategies) guide.
</Tip>
#### Named Entity Recognition
Named Entity Recognition (NER) is a task of finding named entities in a piece of text, such as a person, location, or organization.
Let's modify the instructions in the prompt to make the LLM perform this task. Here, let's also set `return_full_text = False`
so that output doesn't contain the prompt:
```python
>>> torch.manual_seed(1) # doctest: +IGNORE_RESULT
>>> prompt = """Return a list of named entities in the text.
... Text: The Golden State Warriors are an American professional basketball team based in San Francisco.
... Named entities:
... """
>>> sequences = pipe(
... prompt,
... max_new_tokens=15,
... return_full_text = False,
... )
>>> for seq in sequences:
... print(f"{seq['generated_text']}")
- Golden State Warriors
- San Francisco
```
As you can see, the model correctly identified two named entities from the given text.
#### Translation
Another task LLMs can perform is translation. You can choose to use encoder-decoder models for this task, however, here,
for the simplicity of the examples, we'll keep using Falcon-7b-instruct, which does a decent job. Once again, here's how
you can write a basic prompt to instruct a model to translate a piece of text from English to Italian:
```python
>>> torch.manual_seed(2) # doctest: +IGNORE_RESULT
>>> prompt = """Translate the English text to Italian.
... Text: Sometimes, I've believed as many as six impossible things before breakfast.
... Translation:
... """
>>> sequences = pipe(
... prompt,
... max_new_tokens=20,
... do_sample=True,
... top_k=10,
... return_full_text = False,
... )
>>> for seq in sequences:
... print(f"{seq['generated_text']}")
A volte, ho creduto a sei impossibili cose prima di colazione.
```
Here we've added a `do_sample=True` and `top_k=10` to allow the model to be a bit more flexible when generating output.
#### Text summarization
Similar to the translation, text summarization is another generative task where the output **heavily** relies on the input,
and encoder-decoder models can be a better choice. However, decoder-style models can be used for this task as well.
Previously, we have placed the instructions at the very beginning of the prompt. However, the very end of the prompt can
also be a suitable location for instructions. Typically, it's better to place the instruction on one of the extreme ends.
```python
>>> torch.manual_seed(3) # doctest: +IGNORE_RESULT
>>> prompt = """Permaculture is a design process mimicking the diversity, functionality and resilience of natural ecosystems. The principles and practices are drawn from traditional ecological knowledge of indigenous cultures combined with modern scientific understanding and technological innovations. Permaculture design provides a framework helping individuals and communities develop innovative, creative and effective strategies for meeting basic needs while preparing for and mitigating the projected impacts of climate change.
... Write a summary of the above text.
... Summary:
... """
>>> sequences = pipe(
... prompt,
... max_new_tokens=30,
... do_sample=True,
... top_k=10,
... return_full_text = False,
... )
>>> for seq in sequences:
... print(f"{seq['generated_text']}")
Permaculture is an ecological design mimicking natural ecosystems to meet basic needs and prepare for climate change. It is based on traditional knowledge and scientific understanding.
```
#### Question answering
For question answering task we can structure the prompt into the following logical components: instructions, context, question, and
the leading word or phrase (`"Answer:"`) to nudge the model to start generating the answer:
```python
>>> torch.manual_seed(4) # doctest: +IGNORE_RESULT
>>> prompt = """Answer the question using the context below.
... Context: Gazpacho is a cold soup and drink made of raw, blended vegetables. Most gazpacho includes stale bread, tomato, cucumbers, onion, bell peppers, garlic, olive oil, wine vinegar, water, and salt. Northern recipes often include cumin and/or pimentón (smoked sweet paprika). Traditionally, gazpacho was made by pounding the vegetables in a mortar with a pestle; this more laborious method is still sometimes used as it helps keep the gazpacho cool and avoids the foam and silky consistency of smoothie versions made in blenders or food processors.
... Question: What modern tool is used to make gazpacho?
... Answer:
... """
>>> sequences = pipe(
... prompt,
... max_new_tokens=10,
... do_sample=True,
... top_k=10,
... return_full_text = False,
... )
>>> for seq in sequences:
... print(f"Result: {seq['generated_text']}")
Result: Modern tools are used, such as immersion blenders
```
#### Reasoning
Reasoning is one of the most difficult tasks for LLMs, and achieving good results often requires applying advanced prompting techniques, like
[Chain-of-though](#chain-of-thought).
Let's try if we can make a model reason about a simple arithmetics task with a basic prompt:
```python
>>> torch.manual_seed(5) # doctest: +IGNORE_RESULT
>>> prompt = """There are 5 groups of students in the class. Each group has 4 students. How many students are there in the class?"""
>>> sequences = pipe(
... prompt,
... max_new_tokens=30,
... do_sample=True,
... top_k=10,
... return_full_text = False,
... )
>>> for seq in sequences:
... print(f"Result: {seq['generated_text']}")
Result:
There are a total of 5 groups, so there are 5 x 4=20 students in the class.
```
Correct! Let's increase the complexity a little and see if we can still get away with a basic prompt:
```python
>>> torch.manual_seed(6) # doctest: +IGNORE_RESULT
>>> prompt = """I baked 15 muffins. I ate 2 muffins and gave 5 muffins to a neighbor. My partner then bought 6 more muffins and ate 2. How many muffins do we now have?"""
>>> sequences = pipe(
... prompt,
... max_new_tokens=10,
... do_sample=True,
... top_k=10,
... return_full_text = False,
... )
>>> for seq in sequences:
... print(f"Result: {seq['generated_text']}")
Result:
The total number of muffins now is 21
```
This is a wrong answer, it should be 12. In this case, this can be due to the prompt being too basic, or due to the choice
of model, after all we've picked the smallest version of Falcon. Reasoning is difficult for models of all sizes, but larger
models are likely to perform better.
## Best practices of LLM prompting
In this section of the guide we have compiled a list of best practices that tend to improve the prompt results:
* When choosing the model to work with, the latest and most capable models are likely to perform better.
* Start with a simple and short prompt, and iterate from there.
* Put the instructions at the beginning of the prompt, or at the very end. When working with large context, models apply various optimizations to prevent Attention complexity from scaling quadratically. This may make a model more attentive to the beginning or end of a prompt than the middle.
* Clearly separate instructions from the text they apply to - more on this in the next section.
* Be specific and descriptive about the task and the desired outcome - its format, length, style, language, etc.
* Avoid ambiguous descriptions and instructions.
* Favor instructions that say "what to do" instead of those that say "what not to do".
* "Lead" the output in the right direction by writing the first word (or even begin the first sentence for the model).
* Use advanced techniques like [Few-shot prompting](#few-shot-prompting) and [Chain-of-thought](#chain-of-thought)
* Test your prompts with different models to assess their robustness.
* Version and track the performance of your prompts.
## Advanced prompting techniques
### Few-shot prompting
The basic prompts in the sections above are the examples of "zero-shot" prompts, meaning, the model has been given
instructions and context, but no examples with solutions. LLMs that have been fine-tuned on instruction datasets, generally
perform well on such "zero-shot" tasks. However, you may find that your task has more complexity or nuance, and, perhaps,
you have some requirements for the output that the model doesn't catch on just from the instructions. In this case, you can
try the technique called few-shot prompting.
In few-shot prompting, we provide examples in the prompt giving the model more context to improve the performance.
The examples condition the model to generate the output following the patterns in the examples.
Here's an example:
```python
>>> torch.manual_seed(0) # doctest: +IGNORE_RESULT
>>> prompt = """Text: The first human went into space and orbited the Earth on April 12, 1961.
... Date: 04/12/1961
... Text: The first-ever televised presidential debate in the United States took place on September 28, 1960, between presidential candidates John F. Kennedy and Richard Nixon.
... Date:"""
>>> sequences = pipe(
... prompt,
... max_new_tokens=8,
... do_sample=True,
... top_k=10,
... )
>>> for seq in sequences:
... print(f"Result: {seq['generated_text']}")
Result: Text: The first human went into space and orbited the Earth on April 12, 1961.
Date: 04/12/1961
Text: The first-ever televised presidential debate in the United States took place on September 28, 1960, between presidential candidates John F. Kennedy and Richard Nixon.
Date: 09/28/1960
```
In the above code snippet we used a single example to demonstrate the desired output to the model, so this can be called a
"one-shot" prompting. However, depending on the task complexity you may need to use more than one example.
Limitations of the few-shot prompting technique:
- While LLMs can pick up on the patterns in the examples, these technique doesn't work well on complex reasoning tasks
- Few-shot prompting requires creating lengthy prompts. Prompts with large number of tokens can increase computation and latency. There's also a limit to the length of the prompts.
- Sometimes when given a number of examples, models can learn patterns that you didn't intend them to learn, e.g. that the third movie review is always negative.
### Chain-of-thought
Chain-of-thought (CoT) prompting is a technique that nudges a model to produce intermediate reasoning steps thus improving
the results on complex reasoning tasks.
There are two ways of steering a model to producing the reasoning steps:
- few-shot prompting by illustrating examples with detailed answers to questions, showing the model how to work through a problem.
- by instructing the model to reason by adding phrases like "Let's think step by step" or "Take a deep breath and work through the problem step by step."
If we apply the CoT technique to the muffins example from the [reasoning section](#reasoning) and use a larger model,
such as (`tiiuae/falcon-180B-chat`) which you can play with in the [HuggingChat](https://huggingface.co/chat/),
we'll get a significant improvement on the reasoning result:
```text
Let's go through this step-by-step:
1. You start with 15 muffins.
2. You eat 2 muffins, leaving you with 13 muffins.
3. You give 5 muffins to your neighbor, leaving you with 8 muffins.
4. Your partner buys 6 more muffins, bringing the total number of muffins to 14.
5. Your partner eats 2 muffins, leaving you with 12 muffins.
Therefore, you now have 12 muffins.
```
## Prompting vs fine-tuning
You can achieve great results by optimizing your prompts, however, you may still ponder whether fine-tuning a model
would work better for your case. Here are some scenarios when fine-tuning a smaller model may be a preferred option:
- Your domain is wildly different from what LLMs were pre-trained on and extensive prompt optimization did not yield sufficient results.
- You need your model to work well in a low-resource language.
- You need the model to be trained on sensitive data that is under strict regulations.
- You have to use a small model due to cost, privacy, infrastructure or other limitations.
In all of the above examples, you will need to make sure that you either already have or can easily obtain a large enough
domain-specific dataset at a reasonable cost to fine-tune a model. You will also need to have enough time and resources
to fine-tune a model.
If the above examples are not the case for you, optimizing prompts can prove to be more beneficial.
| huggingface/transformers/blob/main/docs/source/en/tasks/prompting.md |
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# XLA Integration for TensorFlow Models
[[open-in-colab]]
Accelerated Linear Algebra, dubbed XLA, is a compiler for accelerating the runtime of TensorFlow Models. From the [official documentation](https://www.tensorflow.org/xla):
XLA (Accelerated Linear Algebra) is a domain-specific compiler for linear algebra that can accelerate TensorFlow models with potentially no source code changes.
Using XLA in TensorFlow is simple – it comes packaged inside the `tensorflow` library, and it can be triggered with the `jit_compile` argument in any graph-creating function such as [`tf.function`](https://www.tensorflow.org/guide/intro_to_graphs). When using Keras methods like `fit()` and `predict()`, you can enable XLA simply by passing the `jit_compile` argument to `model.compile()`. However, XLA is not limited to these methods - it can also be used to accelerate any arbitrary `tf.function`.
Several TensorFlow methods in 🤗 Transformers have been rewritten to be XLA-compatible, including text generation for models such as [GPT2](https://huggingface.co/docs/transformers/model_doc/gpt2), [T5](https://huggingface.co/docs/transformers/model_doc/t5) and [OPT](https://huggingface.co/docs/transformers/model_doc/opt), as well as speech processing for models such as [Whisper](https://huggingface.co/docs/transformers/model_doc/whisper).
While the exact amount of speed-up is very much model-dependent, for TensorFlow text generation models inside 🤗 Transformers, we noticed a speed-up of ~100x. This document will explain how you can use XLA for these models to get the maximum amount of performance. We’ll also provide links to additional resources if you’re interested to learn more about the benchmarks and our design philosophy behind the XLA integration.
## Running TF functions with XLA
Let us consider the following model in TensorFlow:
```py
import tensorflow as tf
model = tf.keras.Sequential(
[tf.keras.layers.Dense(10, input_shape=(10,), activation="relu"), tf.keras.layers.Dense(5, activation="softmax")]
)
```
The above model accepts inputs having a dimension of `(10, )`. We can use the model for running a forward pass like so:
```py
# Generate random inputs for the model.
batch_size = 16
input_vector_dim = 10
random_inputs = tf.random.normal((batch_size, input_vector_dim))
# Run a forward pass.
_ = model(random_inputs)
```
In order to run the forward pass with an XLA-compiled function, we’d need to do:
```py
xla_fn = tf.function(model, jit_compile=True)
_ = xla_fn(random_inputs)
```
The default `call()` function of the `model` is used for compiling the XLA graph. But if there’s any other model function you want to compile into XLA that’s also possible with:
```py
my_xla_fn = tf.function(model.my_xla_fn, jit_compile=True)
```
## Running a TF text generation model with XLA from 🤗 Transformers
To enable XLA-accelerated generation within 🤗 Transformers, you need to have a recent version of `transformers` installed. You can install it by running:
```bash
pip install transformers --upgrade
```
And then you can run the following code:
```py
import tensorflow as tf
from transformers import AutoTokenizer, TFAutoModelForCausalLM
# Will error if the minimal version of Transformers is not installed.
from transformers.utils import check_min_version
check_min_version("4.21.0")
tokenizer = AutoTokenizer.from_pretrained("gpt2", padding_side="left", pad_token="</s>")
model = TFAutoModelForCausalLM.from_pretrained("gpt2")
input_string = ["TensorFlow is"]
# One line to create an XLA generation function
xla_generate = tf.function(model.generate, jit_compile=True)
tokenized_input = tokenizer(input_string, return_tensors="tf")
generated_tokens = xla_generate(**tokenized_input, num_beams=2)
decoded_text = tokenizer.decode(generated_tokens[0], skip_special_tokens=True)
print(f"Generated -- {decoded_text}")
# Generated -- TensorFlow is an open-source, open-source, distributed-source application # framework for the
```
As you can notice, enabling XLA on `generate()` is just a single line of code. The rest of the code remains unchanged. However, there are a couple of gotchas in the above code snippet that are specific to XLA. You need to be aware of those to realize the speed-ups that XLA can bring in. We discuss these in the following section.
## Gotchas to be aware of
When you are executing an XLA-enabled function (like `xla_generate()` above) for the first time, it will internally try to infer the computation graph, which is time-consuming. This process is known as [“tracing”](https://www.tensorflow.org/guide/intro_to_graphs#when_is_a_function_tracing).
You might notice that the generation time is not fast. Successive calls of `xla_generate()` (or any other XLA-enabled function) won’t have to infer the computation graph, given the inputs to the function follow the same shape with which the computation graph was initially built. While this is not a problem for modalities with fixed input shapes (e.g., images), you must pay attention if you are working with variable input shape modalities (e.g., text).
To ensure `xla_generate()` always operates with the same input shapes, you can specify the `padding` arguments when calling the tokenizer.
```py
import tensorflow as tf
from transformers import AutoTokenizer, TFAutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("gpt2", padding_side="left", pad_token="</s>")
model = TFAutoModelForCausalLM.from_pretrained("gpt2")
input_string = ["TensorFlow is"]
xla_generate = tf.function(model.generate, jit_compile=True)
# Here, we call the tokenizer with padding options.
tokenized_input = tokenizer(input_string, pad_to_multiple_of=8, padding=True, return_tensors="tf")
generated_tokens = xla_generate(**tokenized_input, num_beams=2)
decoded_text = tokenizer.decode(generated_tokens[0], skip_special_tokens=True)
print(f"Generated -- {decoded_text}")
```
This way, you can ensure that the inputs to `xla_generate()` will always receive inputs with the shape it was traced with and thus leading to speed-ups in the generation time. You can verify this with the code below:
```py
import time
import tensorflow as tf
from transformers import AutoTokenizer, TFAutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("gpt2", padding_side="left", pad_token="</s>")
model = TFAutoModelForCausalLM.from_pretrained("gpt2")
xla_generate = tf.function(model.generate, jit_compile=True)
for input_string in ["TensorFlow is", "TensorFlow is a", "TFLite is a"]:
tokenized_input = tokenizer(input_string, pad_to_multiple_of=8, padding=True, return_tensors="tf")
start = time.time_ns()
generated_tokens = xla_generate(**tokenized_input, num_beams=2)
end = time.time_ns()
print(f"Execution time -- {(end - start) / 1e6:.1f} ms\n")
```
On a Tesla T4 GPU, you can expect the outputs like so:
```bash
Execution time -- 30819.6 ms
Execution time -- 79.0 ms
Execution time -- 78.9 ms
```
The first call to `xla_generate()` is time-consuming because of tracing, but the successive calls are orders of magnitude faster. Keep in mind that any change in the generation options at any point with trigger re-tracing and thus leading to slow-downs in the generation time.
We didn’t cover all the text generation options 🤗 Transformers provides in this document. We encourage you to read the documentation for advanced use cases.
## Additional Resources
Here, we leave you with some additional resources if you want to delve deeper into XLA in 🤗 Transformers and in general.
* [This Colab Notebook](https://colab.research.google.com/github/huggingface/blog/blob/main/notebooks/91_tf_xla_generate.ipynb) provides an interactive demonstration if you want to fiddle with the XLA-compatible encoder-decoder (like [T5](https://huggingface.co/docs/transformers/model_doc/t5)) and decoder-only (like [GPT2](https://huggingface.co/docs/transformers/model_doc/gpt2)) text generation models.
* [This blog post](https://huggingface.co/blog/tf-xla-generate) provides an overview of the comparison benchmarks for XLA-compatible models along with a friendly introduction to XLA in TensorFlow.
* [This blog post](https://blog.tensorflow.org/2022/11/how-hugging-face-improved-text-generation-performance-with-xla.html) discusses our design philosophy behind adding XLA support to the TensorFlow models in 🤗 Transformers.
* Recommended posts for learning more about XLA and TensorFlow graphs in general:
* [XLA: Optimizing Compiler for Machine Learning](https://www.tensorflow.org/xla)
* [Introduction to graphs and tf.function](https://www.tensorflow.org/guide/intro_to_graphs)
* [Better performance with tf.function](https://www.tensorflow.org/guide/function) | huggingface/transformers/blob/main/docs/source/en/tf_xla.md |
!---
Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-->
# Examples
This folder contains actively maintained examples of use of 🤗 Transformers using the PyTorch backend, organized by ML task.
## The Big Table of Tasks
Here is the list of all our examples:
- with information on whether they are **built on top of `Trainer`** (if not, they still work, they might
just lack some features),
- whether or not they have a version using the [🤗 Accelerate](https://github.com/huggingface/accelerate) library.
- whether or not they leverage the [🤗 Datasets](https://github.com/huggingface/datasets) library.
- links to **Colab notebooks** to walk through the scripts and run them easily,
<!--
Coming soon!
- links to **Cloud deployments** to be able to deploy large-scale trainings in the Cloud with little to no setup.
-->
| Task | Example datasets | Trainer support | 🤗 Accelerate | 🤗 Datasets | Colab
|---|---|:---:|:---:|:---:|:---:|
| [**`language-modeling`**](https://github.com/huggingface/transformers/tree/main/examples/pytorch/language-modeling) | [WikiText-2](https://huggingface.co/datasets/wikitext) | ✅ | ✅ | ✅ | [](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/language_modeling.ipynb)
| [**`multiple-choice`**](https://github.com/huggingface/transformers/tree/main/examples/pytorch/multiple-choice) | [SWAG](https://huggingface.co/datasets/swag) | ✅ | ✅ | ✅ | [](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/multiple_choice.ipynb)
| [**`question-answering`**](https://github.com/huggingface/transformers/tree/main/examples/pytorch/question-answering) | [SQuAD](https://huggingface.co/datasets/squad) | ✅ | ✅ | ✅ | [](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/question_answering.ipynb)
| [**`summarization`**](https://github.com/huggingface/transformers/tree/main/examples/pytorch/summarization) | [XSum](https://huggingface.co/datasets/xsum) | ✅ | ✅ | ✅ | [](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/summarization.ipynb)
| [**`text-classification`**](https://github.com/huggingface/transformers/tree/main/examples/pytorch/text-classification) | [GLUE](https://huggingface.co/datasets/glue) | ✅ | ✅ | ✅ | [](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/text_classification.ipynb)
| [**`text-generation`**](https://github.com/huggingface/transformers/tree/main/examples/pytorch/text-generation) | - | n/a | - | - | [](https://colab.research.google.com/github/huggingface/blog/blob/main/notebooks/02_how_to_generate.ipynb)
| [**`token-classification`**](https://github.com/huggingface/transformers/tree/main/examples/pytorch/token-classification) | [CoNLL NER](https://huggingface.co/datasets/conll2003) | ✅ |✅ | ✅ | [](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/token_classification.ipynb)
| [**`translation`**](https://github.com/huggingface/transformers/tree/main/examples/pytorch/translation) | [WMT](https://huggingface.co/datasets/wmt17) | ✅ | ✅ |✅ | [](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/translation.ipynb)
| [**`speech-recognition`**](https://github.com/huggingface/transformers/tree/main/examples/pytorch/speech-recognition) | [TIMIT](https://huggingface.co/datasets/timit_asr) | ✅ | - |✅ | [](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/speech_recognition.ipynb)
| [**`multi-lingual speech-recognition`**](https://github.com/huggingface/transformers/tree/main/examples/pytorch/speech-recognition) | [Common Voice](https://huggingface.co/datasets/common_voice) | ✅ | - |✅ | [](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/multi_lingual_speech_recognition.ipynb)
| [**`audio-classification`**](https://github.com/huggingface/transformers/tree/main/examples/pytorch/audio-classification) | [SUPERB KS](https://huggingface.co/datasets/superb) | ✅ | - |✅ | [](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/audio_classification.ipynb)
| [**`image-pretraining`**](https://github.com/huggingface/transformers/tree/main/examples/pytorch/image-pretraining) | [ImageNet-1k](https://huggingface.co/datasets/imagenet-1k) | ✅ | - |✅ | /
| [**`image-classification`**](https://github.com/huggingface/transformers/tree/main/examples/pytorch/image-classification) | [CIFAR-10](https://huggingface.co/datasets/cifar10) | ✅ | ✅ |✅ | [](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/image_classification.ipynb)
| [**`semantic-segmentation`**](https://github.com/huggingface/transformers/tree/main/examples/pytorch/semantic-segmentation) | [SCENE_PARSE_150](https://huggingface.co/datasets/scene_parse_150) | ✅ | ✅ |✅ | [](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/semantic_segmentation.ipynb)
## Running quick tests
Most examples are equipped with a mechanism to truncate the number of dataset samples to the desired length. This is useful for debugging purposes, for example to quickly check that all stages of the programs can complete, before running the same setup on the full dataset which may take hours to complete.
For example here is how to truncate all three splits to just 50 samples each:
```
examples/pytorch/token-classification/run_ner.py \
--max_train_samples 50 \
--max_eval_samples 50 \
--max_predict_samples 50 \
[...]
```
Most example scripts should have the first two command line arguments and some have the third one. You can quickly check if a given example supports any of these by passing a `-h` option, e.g.:
```
examples/pytorch/token-classification/run_ner.py -h
```
## Resuming training
You can resume training from a previous checkpoint like this:
1. Pass `--output_dir previous_output_dir` without `--overwrite_output_dir` to resume training from the latest checkpoint in `output_dir` (what you would use if the training was interrupted, for instance).
2. Pass `--resume_from_checkpoint path_to_a_specific_checkpoint` to resume training from that checkpoint folder.
Should you want to turn an example into a notebook where you'd no longer have access to the command
line, 🤗 Trainer supports resuming from a checkpoint via `trainer.train(resume_from_checkpoint)`.
1. If `resume_from_checkpoint` is `True` it will look for the last checkpoint in the value of `output_dir` passed via `TrainingArguments`.
2. If `resume_from_checkpoint` is a path to a specific checkpoint it will use that saved checkpoint folder to resume the training from.
### Upload the trained/fine-tuned model to the Hub
All the example scripts support automatic upload of your final model to the [Model Hub](https://huggingface.co/models) by adding a `--push_to_hub` argument. It will then create a repository with your username slash the name of the folder you are using as `output_dir`. For instance, `"sgugger/test-mrpc"` if your username is `sgugger` and you are working in the folder `~/tmp/test-mrpc`.
To specify a given repository name, use the `--hub_model_id` argument. You will need to specify the whole repository name (including your username), for instance `--hub_model_id sgugger/finetuned-bert-mrpc`. To upload to an organization you are a member of, just use the name of that organization instead of your username: `--hub_model_id huggingface/finetuned-bert-mrpc`.
A few notes on this integration:
- you will need to be logged in to the Hugging Face website locally for it to work, the easiest way to achieve this is to run `huggingface-cli login` and then type your username and password when prompted. You can also pass along your authentication token with the `--hub_token` argument.
- the `output_dir` you pick will either need to be a new folder or a local clone of the distant repository you are using.
## Distributed training and mixed precision
All the PyTorch scripts mentioned above work out of the box with distributed training and mixed precision, thanks to
the [Trainer API](https://huggingface.co/transformers/main_classes/trainer.html). To launch one of them on _n_ GPUs,
use the following command:
```bash
torchrun \
--nproc_per_node number_of_gpu_you_have path_to_script.py \
--all_arguments_of_the_script
```
As an example, here is how you would fine-tune the BERT large model (with whole word masking) on the text
classification MNLI task using the `run_glue` script, with 8 GPUs:
```bash
torchrun \
--nproc_per_node 8 pytorch/text-classification/run_glue.py \
--model_name_or_path bert-large-uncased-whole-word-masking \
--task_name mnli \
--do_train \
--do_eval \
--max_seq_length 128 \
--per_device_train_batch_size 8 \
--learning_rate 2e-5 \
--num_train_epochs 3.0 \
--output_dir /tmp/mnli_output/
```
If you have a GPU with mixed precision capabilities (architecture Pascal or more recent), you can use mixed precision
training with PyTorch 1.6.0 or latest, or by installing the [Apex](https://github.com/NVIDIA/apex) library for previous
versions. Just add the flag `--fp16` to your command launching one of the scripts mentioned above!
Using mixed precision training usually results in 2x-speedup for training with the same final results (as shown in
[this table](https://github.com/huggingface/transformers/tree/main/examples/text-classification#mixed-precision-training)
for text classification).
## Running on TPUs
When using Tensorflow, TPUs are supported out of the box as a `tf.distribute.Strategy`.
When using PyTorch, we support TPUs thanks to `pytorch/xla`. For more context and information on how to setup your TPU environment refer to Google's documentation and to the
very detailed [pytorch/xla README](https://github.com/pytorch/xla/blob/master/README.md).
In this repo, we provide a very simple launcher script named
[xla_spawn.py](https://github.com/huggingface/transformers/tree/main/examples/pytorch/xla_spawn.py) that lets you run our
example scripts on multiple TPU cores without any boilerplate. Just pass a `--num_cores` flag to this script, then your
regular training script with its arguments (this is similar to the `torch.distributed.launch` helper for
`torch.distributed`):
```bash
python xla_spawn.py --num_cores num_tpu_you_have \
path_to_script.py \
--all_arguments_of_the_script
```
As an example, here is how you would fine-tune the BERT large model (with whole word masking) on the text
classification MNLI task using the `run_glue` script, with 8 TPUs (from this folder):
```bash
python xla_spawn.py --num_cores 8 \
text-classification/run_glue.py \
--model_name_or_path bert-large-uncased-whole-word-masking \
--task_name mnli \
--do_train \
--do_eval \
--max_seq_length 128 \
--per_device_train_batch_size 8 \
--learning_rate 2e-5 \
--num_train_epochs 3.0 \
--output_dir /tmp/mnli_output/
```
## Using Accelerate
Most PyTorch example scripts have a version using the [🤗 Accelerate](https://github.com/huggingface/accelerate) library
that exposes the training loop so it's easy for you to customize or tweak them to your needs. They all require you to
install `accelerate` with the latest development version
```bash
pip install git+https://github.com/huggingface/accelerate
```
Then you can easily launch any of the scripts by running
```bash
accelerate config
```
and reply to the questions asked. Then
```bash
accelerate test
```
that will check everything is ready for training. Finally, you can launch training with
```bash
accelerate launch path_to_script.py --args_to_script
```
## Logging & Experiment tracking
You can easily log and monitor your runs code. The following are currently supported:
* [TensorBoard](https://www.tensorflow.org/tensorboard)
* [Weights & Biases](https://docs.wandb.ai/integrations/huggingface)
* [Comet ML](https://www.comet.ml/docs/python-sdk/huggingface/)
* [Neptune](https://docs.neptune.ai/integrations-and-supported-tools/model-training/hugging-face)
* [ClearML](https://clear.ml/docs/latest/docs/getting_started/ds/ds_first_steps)
* [DVCLive](https://dvc.org/doc/dvclive/ml-frameworks/huggingface)
### Weights & Biases
To use Weights & Biases, install the wandb package with:
```bash
pip install wandb
```
Then log in the command line:
```bash
wandb login
```
If you are in Jupyter or Colab, you should login with:
```python
import wandb
wandb.login()
```
To enable logging to W&B, include `"wandb"` in the `report_to` of your `TrainingArguments` or script. Or just pass along `--report_to_all` if you have `wandb` installed.
Whenever you use `Trainer` or `TFTrainer` classes, your losses, evaluation metrics, model topology and gradients (for `Trainer` only) will automatically be logged.
Advanced configuration is possible by setting environment variables:
| Environment Variable | Value |
|---|---|
| WANDB_LOG_MODEL | Log the model as artifact (log the model as artifact at the end of training) (`false` by default) |
| WANDB_WATCH | one of `gradients` (default) to log histograms of gradients, `all` to log histograms of both gradients and parameters, or `false` for no histogram logging |
| WANDB_PROJECT | Organize runs by project |
Set run names with `run_name` argument present in scripts or as part of `TrainingArguments`.
Additional configuration options are available through generic [wandb environment variables](https://docs.wandb.com/library/environment-variables).
Refer to related [documentation & examples](https://docs.wandb.ai/integrations/huggingface).
### Comet.ml
To use `comet_ml`, install the Python package with:
```bash
pip install comet_ml
```
or if in a Conda environment:
```bash
conda install -c comet_ml -c anaconda -c conda-forge comet_ml
```
### Neptune
First, install the Neptune client library. You can do it with either `pip` or `conda`:
`pip`:
```bash
pip install neptune
```
`conda`:
```bash
conda install -c conda-forge neptune
```
Next, in your model training script, import `NeptuneCallback`:
```python
from transformers.integrations import NeptuneCallback
```
To enable Neptune logging, in your `TrainingArguments`, set the `report_to` argument to `"neptune"`:
```python
training_args = TrainingArguments(
"quick-training-distilbert-mrpc",
evaluation_strategy="steps",
eval_steps=20,
report_to="neptune",
)
trainer = Trainer(
model,
training_args,
...
)
```
**Note:** This method requires saving your Neptune credentials as environment variables (see the bottom of the section).
Alternatively, for more logging options, create a Neptune callback:
```python
neptune_callback = NeptuneCallback()
```
To add more detail to the tracked run, you can supply optional arguments to `NeptuneCallback`.
Some examples:
```python
neptune_callback = NeptuneCallback(
name = "DistilBERT",
description = "DistilBERT fine-tuned on GLUE/MRPC",
tags = ["args-callback", "fine-tune", "MRPC"], # tags help you manage runs in Neptune
base_namespace="callback", # the default is "finetuning"
log_checkpoints = "best", # other options are "last", "same", and None
capture_hardware_metrics = False, # additional keyword arguments for a Neptune run
)
```
Pass the callback to the Trainer:
```python
training_args = TrainingArguments(..., report_to=None)
trainer = Trainer(
model,
training_args,
...
callbacks=[neptune_callback],
)
```
Now, when you start the training with `trainer.train()`, your metadata will be logged in Neptune.
**Note:** Although you can pass your **Neptune API token** and **project name** as arguments when creating the callback, the recommended way is to save them as environment variables:
| Environment variable | Value |
| :------------------- | :--------------------------------------------------- |
| `NEPTUNE_API_TOKEN` | Your Neptune API token. To find and copy it, click your Neptune avatar and select **Get your API token**. |
| `NEPTUNE_PROJECT` | The full name of your Neptune project (`workspace-name/project-name`). To find and copy it, head to **project settings** → **Properties**. |
For detailed instructions and examples, see the [Neptune docs](https://docs.neptune.ai/integrations/transformers/).
### ClearML
To use ClearML, install the clearml package with:
```bash
pip install clearml
```
Then [create new credentials]() from the ClearML Server. You can get a free hosted server [here]() or [self-host your own]()!
After creating your new credentials, you can either copy the local snippet which you can paste after running:
```bash
clearml-init
```
Or you can copy the jupyter snippet if you are in Jupyter or Colab:
```python
%env CLEARML_WEB_HOST=https://app.clear.ml
%env CLEARML_API_HOST=https://api.clear.ml
%env CLEARML_FILES_HOST=https://files.clear.ml
%env CLEARML_API_ACCESS_KEY=***
%env CLEARML_API_SECRET_KEY=***
```
To enable logging to ClearML, include `"clearml"` in the `report_to` of your `TrainingArguments` or script. Or just pass along `--report_to all` if you have `clearml` already installed.
Advanced configuration is possible by setting environment variables:
| Environment Variable | Value |
|---|---|
| CLEARML_PROJECT | Name of the project in ClearML. (default: `"HuggingFace Transformers"`) |
| CLEARML_TASK | Name of the task in ClearML. (default: `"Trainer"`) |
Additional configuration options are available through generic [clearml environment variables](https://clear.ml/docs/latest/docs/configs/env_vars).
| huggingface/transformers/blob/main/examples/pytorch/README.md |
!---
Copyright 2022 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Custom hardware for training
The hardware you use to run model training and inference can have a big effect on performance. For a deep dive into GPUs make sure to check out Tim Dettmer's excellent [blog post](https://timdettmers.com/2020/09/07/which-gpu-for-deep-learning/).
Let's have a look at some practical advice for GPU setups.
## GPU
When you train bigger models you have essentially three options:
- bigger GPUs
- more GPUs
- more CPU and NVMe (offloaded to by [DeepSpeed-Infinity](main_classes/deepspeed#nvme-support))
Let's start at the case where you have a single GPU.
### Power and Cooling
If you bought an expensive high end GPU make sure you give it the correct power and sufficient cooling.
**Power**:
Some high end consumer GPU cards have 2 and sometimes 3 PCI-E 8-Pin power sockets. Make sure you have as many independent 12V PCI-E 8-Pin cables plugged into the card as there are sockets. Do not use the 2 splits at one end of the same cable (also known as pigtail cable). That is if you have 2 sockets on the GPU, you want 2 PCI-E 8-Pin cables going from your PSU to the card and not one that has 2 PCI-E 8-Pin connectors at the end! You won't get the full performance out of your card otherwise.
Each PCI-E 8-Pin power cable needs to be plugged into a 12V rail on the PSU side and can supply up to 150W of power.
Some other cards may use a PCI-E 12-Pin connectors, and these can deliver up to 500-600W of power.
Low end cards may use 6-Pin connectors, which supply up to 75W of power.
Additionally you want the high-end PSU that has stable voltage. Some lower quality ones may not give the card the stable voltage it needs to function at its peak.
And of course the PSU needs to have enough unused Watts to power the card.
**Cooling**:
When a GPU gets overheated it will start throttling down and will not deliver full performance and it can even shutdown if it gets too hot.
It's hard to tell the exact best temperature to strive for when a GPU is heavily loaded, but probably anything under +80C is good, but lower is better - perhaps 70-75C is an excellent range to be in. The throttling down is likely to start at around 84-90C. But other than throttling performance a prolonged very high temperature is likely to reduce the lifespan of a GPU.
Next let's have a look at one of the most important aspects when having multiple GPUs: connectivity.
### Multi-GPU Connectivity
If you use multiple GPUs the way cards are inter-connected can have a huge impact on the total training time. If the GPUs are on the same physical node, you can run:
```
nvidia-smi topo -m
```
and it will tell you how the GPUs are inter-connected. On a machine with dual-GPU and which are connected with NVLink, you will most likely see something like:
```
GPU0 GPU1 CPU Affinity NUMA Affinity
GPU0 X NV2 0-23 N/A
GPU1 NV2 X 0-23 N/A
```
on a different machine w/o NVLink we may see:
```
GPU0 GPU1 CPU Affinity NUMA Affinity
GPU0 X PHB 0-11 N/A
GPU1 PHB X 0-11 N/A
```
The report includes this legend:
```
X = Self
SYS = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI)
NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node
PHB = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU)
PXB = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge)
PIX = Connection traversing at most a single PCIe bridge
NV# = Connection traversing a bonded set of # NVLinks
```
So the first report `NV2` tells us the GPUs are interconnected with 2 NVLinks, and the second report `PHB` we have a typical consumer-level PCIe+Bridge setup.
Check what type of connectivity you have on your setup. Some of these will make the communication between cards faster (e.g. NVLink), others slower (e.g. PHB).
Depending on the type of scalability solution used, the connectivity speed could have a major or a minor impact. If the GPUs need to sync rarely, as in DDP, the impact of a slower connection will be less significant. If the GPUs need to send messages to each other often, as in ZeRO-DP, then faster connectivity becomes super important to achieve faster training.
#### NVlink
[NVLink](https://en.wikipedia.org/wiki/NVLink) is a wire-based serial multi-lane near-range communications link developed by Nvidia.
Each new generation provides a faster bandwidth, e.g. here is a quote from [Nvidia Ampere GA102 GPU Architecture](https://www.nvidia.com/content/dam/en-zz/Solutions/geforce/ampere/pdf/NVIDIA-ampere-GA102-GPU-Architecture-Whitepaper-V1.pdf):
> Third-Generation NVLink®
> GA102 GPUs utilize NVIDIA’s third-generation NVLink interface, which includes four x4 links,
> with each link providing 14.0625 GB/sec bandwidth in each direction between two GPUs. Four
> links provide 56.25 GB/sec bandwidth in each direction, and 112.5 GB/sec total bandwidth
> between two GPUs. Two RTX 3090 GPUs can be connected together for SLI using NVLink.
> (Note that 3-Way and 4-Way SLI configurations are not supported.)
So the higher `X` you get in the report of `NVX` in the output of `nvidia-smi topo -m` the better. The generation will depend on your GPU architecture.
Let's compare the execution of a gpt2 language model training over a small sample of wikitext.
The results are:
| NVlink | Time |
| ----- | ---: |
| Y | 101s |
| N | 131s |
You can see that NVLink completes the training ~23% faster. In the second benchmark we use `NCCL_P2P_DISABLE=1` to tell the GPUs not to use NVLink.
Here is the full benchmark code and outputs:
```bash
# DDP w/ NVLink
rm -r /tmp/test-clm; CUDA_VISIBLE_DEVICES=0,1 torchrun \
--nproc_per_node 2 examples/pytorch/language-modeling/run_clm.py --model_name_or_path gpt2 \
--dataset_name wikitext --dataset_config_name wikitext-2-raw-v1 --do_train \
--output_dir /tmp/test-clm --per_device_train_batch_size 4 --max_steps 200
{'train_runtime': 101.9003, 'train_samples_per_second': 1.963, 'epoch': 0.69}
# DDP w/o NVLink
rm -r /tmp/test-clm; CUDA_VISIBLE_DEVICES=0,1 NCCL_P2P_DISABLE=1 torchrun \
--nproc_per_node 2 examples/pytorch/language-modeling/run_clm.py --model_name_or_path gpt2 \
--dataset_name wikitext --dataset_config_name wikitext-2-raw-v1 --do_train
--output_dir /tmp/test-clm --per_device_train_batch_size 4 --max_steps 200
{'train_runtime': 131.4367, 'train_samples_per_second': 1.522, 'epoch': 0.69}
```
Hardware: 2x TITAN RTX 24GB each + NVlink with 2 NVLinks (`NV2` in `nvidia-smi topo -m`)
Software: `pytorch-1.8-to-be` + `cuda-11.0` / `transformers==4.3.0.dev0`
| huggingface/transformers/blob/main/docs/source/en/perf_hardware.md |
!--Copyright 2022 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# MaskFormer
<Tip>
This is a recently introduced model so the API hasn't been tested extensively. There may be some bugs or slight
breaking changes to fix it in the future. If you see something strange, file a [Github Issue](https://github.com/huggingface/transformers/issues/new?assignees=&labels=&template=bug-report.md&title).
</Tip>
## Overview
The MaskFormer model was proposed in [Per-Pixel Classification is Not All You Need for Semantic Segmentation](https://arxiv.org/abs/2107.06278) by Bowen Cheng, Alexander G. Schwing, Alexander Kirillov. MaskFormer addresses semantic segmentation with a mask classification paradigm instead of performing classic pixel-level classification.
The abstract from the paper is the following:
*Modern approaches typically formulate semantic segmentation as a per-pixel classification task, while instance-level segmentation is handled with an alternative mask classification. Our key insight: mask classification is sufficiently general to solve both semantic- and instance-level segmentation tasks in a unified manner using the exact same model, loss, and training procedure. Following this observation, we propose MaskFormer, a simple mask classification model which predicts a set of binary masks, each associated with a single global class label prediction. Overall, the proposed mask classification-based method simplifies the landscape of effective approaches to semantic and panoptic segmentation tasks and shows excellent empirical results. In particular, we observe that MaskFormer outperforms per-pixel classification baselines when the number of classes is large. Our mask classification-based method outperforms both current state-of-the-art semantic (55.6 mIoU on ADE20K) and panoptic segmentation (52.7 PQ on COCO) models.*
The figure below illustrates the architecture of MaskFormer. Taken from the [original paper](https://arxiv.org/abs/2107.06278).
<img width="600" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/maskformer_architecture.png"/>
This model was contributed by [francesco](https://huggingface.co/francesco). The original code can be found [here](https://github.com/facebookresearch/MaskFormer).
## Usage tips
- MaskFormer's Transformer decoder is identical to the decoder of [DETR](detr). During training, the authors of DETR did find it helpful to use auxiliary losses in the decoder, especially to help the model output the correct number of objects of each class. If you set the parameter `use_auxilary_loss` of [`MaskFormerConfig`] to `True`, then prediction feedforward neural networks and Hungarian losses are added after each decoder layer (with the FFNs sharing parameters).
- If you want to train the model in a distributed environment across multiple nodes, then one should update the
`get_num_masks` function inside in the `MaskFormerLoss` class of `modeling_maskformer.py`. When training on multiple nodes, this should be
set to the average number of target masks across all nodes, as can be seen in the original implementation [here](https://github.com/facebookresearch/MaskFormer/blob/da3e60d85fdeedcb31476b5edd7d328826ce56cc/mask_former/modeling/criterion.py#L169).
- One can use [`MaskFormerImageProcessor`] to prepare images for the model and optional targets for the model.
- To get the final segmentation, depending on the task, you can call [`~MaskFormerImageProcessor.post_process_semantic_segmentation`] or [`~MaskFormerImageProcessor.post_process_panoptic_segmentation`]. Both tasks can be solved using [`MaskFormerForInstanceSegmentation`] output, panoptic segmentation accepts an optional `label_ids_to_fuse` argument to fuse instances of the target object/s (e.g. sky) together.
## Resources
<PipelineTag pipeline="image-segmentation"/>
- All notebooks that illustrate inference as well as fine-tuning on custom data with MaskFormer can be found [here](https://github.com/NielsRogge/Transformers-Tutorials/tree/master/MaskFormer).
## MaskFormer specific outputs
[[autodoc]] models.maskformer.modeling_maskformer.MaskFormerModelOutput
[[autodoc]] models.maskformer.modeling_maskformer.MaskFormerForInstanceSegmentationOutput
## MaskFormerConfig
[[autodoc]] MaskFormerConfig
## MaskFormerImageProcessor
[[autodoc]] MaskFormerImageProcessor
- preprocess
- encode_inputs
- post_process_semantic_segmentation
- post_process_instance_segmentation
- post_process_panoptic_segmentation
## MaskFormerFeatureExtractor
[[autodoc]] MaskFormerFeatureExtractor
- __call__
- encode_inputs
- post_process_semantic_segmentation
- post_process_instance_segmentation
- post_process_panoptic_segmentation
## MaskFormerModel
[[autodoc]] MaskFormerModel
- forward
## MaskFormerForInstanceSegmentation
[[autodoc]] MaskFormerForInstanceSegmentation
- forward | huggingface/transformers/blob/main/docs/source/en/model_doc/maskformer.md |
!--Copyright 2022 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Hybrid Vision Transformer (ViT Hybrid)
## Overview
The hybrid Vision Transformer (ViT) model was proposed in [An Image is Worth 16x16 Words: Transformers for Image Recognition
at Scale](https://arxiv.org/abs/2010.11929) by Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk
Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob
Uszkoreit, Neil Houlsby. It's the first paper that successfully trains a Transformer encoder on ImageNet, attaining
very good results compared to familiar convolutional architectures. ViT hybrid is a slight variant of the [plain Vision Transformer](vit),
by leveraging a convolutional backbone (specifically, [BiT](bit)) whose features are used as initial "tokens" for the Transformer.
The abstract from the paper is the following:
*While the Transformer architecture has become the de-facto standard for natural language processing tasks, its
applications to computer vision remain limited. In vision, attention is either applied in conjunction with
convolutional networks, or used to replace certain components of convolutional networks while keeping their overall
structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to
sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of
data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.),
Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring
substantially fewer computational resources to train.*
This model was contributed by [nielsr](https://huggingface.co/nielsr). The original code (written in JAX) can be
found [here](https://github.com/google-research/vision_transformer).
## Resources
A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with ViT Hybrid.
<PipelineTag pipeline="image-classification"/>
- [`ViTHybridForImageClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/image-classification) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/image_classification.ipynb).
- See also: [Image classification task guide](../tasks/image_classification)
If you're interested in submitting a resource to be included here, please feel free to open a Pull Request and we'll review it! The resource should ideally demonstrate something new instead of duplicating an existing resource.
## ViTHybridConfig
[[autodoc]] ViTHybridConfig
## ViTHybridImageProcessor
[[autodoc]] ViTHybridImageProcessor
- preprocess
## ViTHybridModel
[[autodoc]] ViTHybridModel
- forward
## ViTHybridForImageClassification
[[autodoc]] ViTHybridForImageClassification
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/vit_hybrid.md |
!--Copyright 2022 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# LayoutLMv3
## Overview
The LayoutLMv3 model was proposed in [LayoutLMv3: Pre-training for Document AI with Unified Text and Image Masking](https://arxiv.org/abs/2204.08387) by Yupan Huang, Tengchao Lv, Lei Cui, Yutong Lu, Furu Wei.
LayoutLMv3 simplifies [LayoutLMv2](layoutlmv2) by using patch embeddings (as in [ViT](vit)) instead of leveraging a CNN backbone, and pre-trains the model on 3 objectives: masked language modeling (MLM), masked image modeling (MIM)
and word-patch alignment (WPA).
The abstract from the paper is the following:
*Self-supervised pre-training techniques have achieved remarkable progress in Document AI. Most multimodal pre-trained models use a masked language modeling objective to learn bidirectional representations on the text modality, but they differ in pre-training objectives for the image modality. This discrepancy adds difficulty to multimodal representation learning. In this paper, we propose LayoutLMv3 to pre-train multimodal Transformers for Document AI with unified text and image masking. Additionally, LayoutLMv3 is pre-trained with a word-patch alignment objective to learn cross-modal alignment by predicting whether the corresponding image patch of a text word is masked. The simple unified architecture and training objectives make LayoutLMv3 a general-purpose pre-trained model for both text-centric and image-centric Document AI tasks. Experimental results show that LayoutLMv3 achieves state-of-the-art performance not only in text-centric tasks, including form understanding, receipt understanding, and document visual question answering, but also in image-centric tasks such as document image classification and document layout analysis.*
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/layoutlmv3_architecture.png"
alt="drawing" width="600"/>
<small> LayoutLMv3 architecture. Taken from the <a href="https://arxiv.org/abs/2204.08387">original paper</a>. </small>
This model was contributed by [nielsr](https://huggingface.co/nielsr). The TensorFlow version of this model was added by [chriskoo](https://huggingface.co/chriskoo), [tokec](https://huggingface.co/tokec), and [lre](https://huggingface.co/lre). The original code can be found [here](https://github.com/microsoft/unilm/tree/master/layoutlmv3).
## Usage tips
- In terms of data processing, LayoutLMv3 is identical to its predecessor [LayoutLMv2](layoutlmv2), except that:
- images need to be resized and normalized with channels in regular RGB format. LayoutLMv2 on the other hand normalizes the images internally and expects the channels in BGR format.
- text is tokenized using byte-pair encoding (BPE), as opposed to WordPiece.
Due to these differences in data preprocessing, one can use [`LayoutLMv3Processor`] which internally combines a [`LayoutLMv3ImageProcessor`] (for the image modality) and a [`LayoutLMv3Tokenizer`]/[`LayoutLMv3TokenizerFast`] (for the text modality) to prepare all data for the model.
- Regarding usage of [`LayoutLMv3Processor`], we refer to the [usage guide](layoutlmv2#usage-layoutlmv2processor) of its predecessor.
## Resources
A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with LayoutLMv3. If you're interested in submitting a resource to be included here, please feel free to open a Pull Request and we'll review it! The resource should ideally demonstrate something new instead of duplicating an existing resource.
<Tip>
LayoutLMv3 is nearly identical to LayoutLMv2, so we've also included LayoutLMv2 resources you can adapt for LayoutLMv3 tasks. For these notebooks, take care to use [`LayoutLMv2Processor`] instead when preparing data for the model!
</Tip>
- Demo notebooks for LayoutLMv3 can be found [here](https://github.com/NielsRogge/Transformers-Tutorials/tree/master/LayoutLMv3).
- Demo scripts can be found [here](https://github.com/huggingface/transformers/tree/main/examples/research_projects/layoutlmv3).
<PipelineTag pipeline="text-classification"/>
- [`LayoutLMv2ForSequenceClassification`] is supported by this [notebook](https://colab.research.google.com/github/NielsRogge/Transformers-Tutorials/blob/master/LayoutLMv2/RVL-CDIP/Fine_tuning_LayoutLMv2ForSequenceClassification_on_RVL_CDIP.ipynb).
- [Text classification task guide](../tasks/sequence_classification)
<PipelineTag pipeline="token-classification"/>
- [`LayoutLMv3ForTokenClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/research_projects/layoutlmv3) and [notebook](https://colab.research.google.com/github/NielsRogge/Transformers-Tutorials/blob/master/LayoutLMv3/Fine_tune_LayoutLMv3_on_FUNSD_(HuggingFace_Trainer).ipynb).
- A [notebook](https://colab.research.google.com/github/NielsRogge/Transformers-Tutorials/blob/master/LayoutLMv2/FUNSD/Inference_with_LayoutLMv2ForTokenClassification.ipynb) for how to perform inference with [`LayoutLMv2ForTokenClassification`] and a [notebook](https://colab.research.google.com/github/NielsRogge/Transformers-Tutorials/blob/master/LayoutLMv2/FUNSD/True_inference_with_LayoutLMv2ForTokenClassification_%2B_Gradio_demo.ipynb) for how to perform inference when no labels are available with [`LayoutLMv2ForTokenClassification`].
- A [notebook](https://colab.research.google.com/github/NielsRogge/Transformers-Tutorials/blob/master/LayoutLMv2/FUNSD/Fine_tuning_LayoutLMv2ForTokenClassification_on_FUNSD_using_HuggingFace_Trainer.ipynb) for how to finetune [`LayoutLMv2ForTokenClassification`] with the 🤗 Trainer.
- [Token classification task guide](../tasks/token_classification)
<PipelineTag pipeline="question-answering"/>
- [`LayoutLMv2ForQuestionAnswering`] is supported by this [notebook](https://colab.research.google.com/github/NielsRogge/Transformers-Tutorials/blob/master/LayoutLMv2/DocVQA/Fine_tuning_LayoutLMv2ForQuestionAnswering_on_DocVQA.ipynb).
- [Question answering task guide](../tasks/question_answering)
**Document question answering**
- [Document question answering task guide](../tasks/document_question_answering)
## LayoutLMv3Config
[[autodoc]] LayoutLMv3Config
## LayoutLMv3FeatureExtractor
[[autodoc]] LayoutLMv3FeatureExtractor
- __call__
## LayoutLMv3ImageProcessor
[[autodoc]] LayoutLMv3ImageProcessor
- preprocess
## LayoutLMv3Tokenizer
[[autodoc]] LayoutLMv3Tokenizer
- __call__
- save_vocabulary
## LayoutLMv3TokenizerFast
[[autodoc]] LayoutLMv3TokenizerFast
- __call__
## LayoutLMv3Processor
[[autodoc]] LayoutLMv3Processor
- __call__
<frameworkcontent>
<pt>
## LayoutLMv3Model
[[autodoc]] LayoutLMv3Model
- forward
## LayoutLMv3ForSequenceClassification
[[autodoc]] LayoutLMv3ForSequenceClassification
- forward
## LayoutLMv3ForTokenClassification
[[autodoc]] LayoutLMv3ForTokenClassification
- forward
## LayoutLMv3ForQuestionAnswering
[[autodoc]] LayoutLMv3ForQuestionAnswering
- forward
</pt>
<tf>
## TFLayoutLMv3Model
[[autodoc]] TFLayoutLMv3Model
- call
## TFLayoutLMv3ForSequenceClassification
[[autodoc]] TFLayoutLMv3ForSequenceClassification
- call
## TFLayoutLMv3ForTokenClassification
[[autodoc]] TFLayoutLMv3ForTokenClassification
- call
## TFLayoutLMv3ForQuestionAnswering
[[autodoc]] TFLayoutLMv3ForQuestionAnswering
- call
</tf>
</frameworkcontent>
| huggingface/transformers/blob/main/docs/source/en/model_doc/layoutlmv3.md |
!---
Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-->
# 🤗 Benchmark results
Here, you can find a list of the different benchmark results created by the community.
If you would like to list benchmark results on your favorite models of the [model hub](https://huggingface.co/models) here, please open a Pull Request and add it below.
| Benchmark description | Results | Environment info | Author |
|:----------|:-------------|:-------------|------:|
| PyTorch Benchmark on inference for `bert-base-cased` |[memory](https://github.com/patrickvonplaten/files_to_link_to/blob/master/bert_benchmark/inference_memory.csv) | [env](https://github.com/patrickvonplaten/files_to_link_to/blob/master/bert_benchmark/env.csv) | [Partick von Platen](https://github.com/patrickvonplaten) |
| PyTorch Benchmark on inference for `bert-base-cased` |[time](https://github.com/patrickvonplaten/files_to_link_to/blob/master/bert_benchmark/inference_time.csv) | [env](https://github.com/patrickvonplaten/files_to_link_to/blob/master/bert_benchmark/env.csv) | [Partick von Platen](https://github.com/patrickvonplaten) |
| huggingface/transformers/blob/main/examples/tensorflow/benchmarking/README.md |
!--Copyright 2021 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# XGLM
## Overview
The XGLM model was proposed in [Few-shot Learning with Multilingual Language Models](https://arxiv.org/abs/2112.10668)
by Xi Victoria Lin, Todor Mihaylov, Mikel Artetxe, Tianlu Wang, Shuohui Chen, Daniel Simig, Myle Ott, Naman Goyal,
Shruti Bhosale, Jingfei Du, Ramakanth Pasunuru, Sam Shleifer, Punit Singh Koura, Vishrav Chaudhary, Brian O'Horo,
Jeff Wang, Luke Zettlemoyer, Zornitsa Kozareva, Mona Diab, Veselin Stoyanov, Xian Li.
The abstract from the paper is the following:
*Large-scale autoregressive language models such as GPT-3 are few-shot learners that can perform a wide range of language
tasks without fine-tuning. While these models are known to be able to jointly represent many different languages,
their training data is dominated by English, potentially limiting their cross-lingual generalization.
In this work, we train multilingual autoregressive language models on a balanced corpus covering a diverse set of languages,
and study their few- and zero-shot learning capabilities in a wide range of tasks. Our largest model with 7.5 billion parameters
sets new state of the art in few-shot learning in more than 20 representative languages, outperforming GPT-3 of comparable size
in multilingual commonsense reasoning (with +7.4% absolute accuracy improvement in 0-shot settings and +9.4% in 4-shot settings)
and natural language inference (+5.4% in each of 0-shot and 4-shot settings). On the FLORES-101 machine translation benchmark,
our model outperforms GPT-3 on 171 out of 182 translation directions with 32 training examples, while surpassing the
official supervised baseline in 45 directions. We present a detailed analysis of where the model succeeds and fails,
showing in particular that it enables cross-lingual in-context learning on some tasks, while there is still room for improvement
on surface form robustness and adaptation to tasks that do not have a natural cloze form. Finally, we evaluate our models
in social value tasks such as hate speech detection in five languages and find it has limitations similar to comparable sized GPT-3 models.*
This model was contributed by [Suraj](https://huggingface.co/valhalla). The original code can be found [here](https://github.com/pytorch/fairseq/tree/main/examples/xglm).
## Resources
- [Causal language modeling task guide](../tasks/language_modeling)
## XGLMConfig
[[autodoc]] XGLMConfig
## XGLMTokenizer
[[autodoc]] XGLMTokenizer
- build_inputs_with_special_tokens
- get_special_tokens_mask
- create_token_type_ids_from_sequences
- save_vocabulary
## XGLMTokenizerFast
[[autodoc]] XGLMTokenizerFast
<frameworkcontent>
<pt>
## XGLMModel
[[autodoc]] XGLMModel
- forward
## XGLMForCausalLM
[[autodoc]] XGLMForCausalLM
- forward
</pt>
<tf>
## TFXGLMModel
[[autodoc]] TFXGLMModel
- call
## TFXGLMForCausalLM
[[autodoc]] TFXGLMForCausalLM
- call
</tf>
<jax>
## FlaxXGLMModel
[[autodoc]] FlaxXGLMModel
- __call__
## FlaxXGLMForCausalLM
[[autodoc]] FlaxXGLMForCausalLM
- __call__
</jax>
</frameworkcontent> | huggingface/transformers/blob/main/docs/source/en/model_doc/xglm.md |
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contains specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# MADLAD-400
## Overview
MADLAD-400 models were released in the paper [MADLAD-400: A Multilingual And Document-Level Large Audited Dataset](MADLAD-400: A Multilingual And Document-Level Large Audited Dataset).
The abstract from the paper is the following:
*We introduce MADLAD-400, a manually audited, general domain 3T token monolingual dataset based on CommonCrawl, spanning 419 languages. We discuss
the limitations revealed by self-auditing MADLAD-400, and the role data auditing
had in the dataset creation process. We then train and release a 10.7B-parameter
multilingual machine translation model on 250 billion tokens covering over 450
languages using publicly available data, and find that it is competitive with models
that are significantly larger, and report the results on different domains. In addition, we train a 8B-parameter language model, and assess the results on few-shot
translation. We make the baseline models 1
available to the research community.*
This model was added by [Juarez Bochi](https://huggingface.co/jbochi). The original checkpoints can be found [here](https://github.com/google-research/google-research/tree/master/madlad_400).
This is a machine translation model that supports many low-resource languages, and that is competitive with models that are significantly larger.
One can directly use MADLAD-400 weights without finetuning the model:
```python
>>> from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
>>> model = AutoModelForSeq2SeqLM.from_pretrained("google/madlad400-3b-mt")
>>> tokenizer = AutoTokenizer.from_pretrained("google/madlad400-3b-mt")
>>> inputs = tokenizer("<2pt> I love pizza!", return_tensors="pt")
>>> outputs = model.generate(**inputs)
>>> print(tokenizer.batch_decode(outputs, skip_special_tokens=True))
['Eu amo pizza!']
```
Google has released the following variants:
- [google/madlad400-3b-mt](https://huggingface.co/google/madlad400-3b-mt)
- [google/madlad400-7b-mt](https://huggingface.co/google/madlad400-7b-mt)
- [google/madlad400-7b-mt-bt](https://huggingface.co/google/madlad400-7b-mt-bt)
- [google/madlad400-10b-mt](https://huggingface.co/google/madlad400-10b-mt)
The original checkpoints can be found [here](https://github.com/google-research/google-research/tree/master/madlad_400).
<Tip>
Refer to [T5's documentation page](t5) for all API references, code examples, and notebooks. For more details regarding training and evaluation of the MADLAD-400, refer to the model card.
</Tip>
| huggingface/transformers/blob/main/docs/source/en/model_doc/madlad-400.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# How to create a custom pipeline?
In this guide, we will see how to create a custom pipeline and share it on the [Hub](hf.co/models) or add it to the
🤗 Transformers library.
First and foremost, you need to decide the raw entries the pipeline will be able to take. It can be strings, raw bytes,
dictionaries or whatever seems to be the most likely desired input. Try to keep these inputs as pure Python as possible
as it makes compatibility easier (even through other languages via JSON). Those will be the `inputs` of the
pipeline (`preprocess`).
Then define the `outputs`. Same policy as the `inputs`. The simpler, the better. Those will be the outputs of
`postprocess` method.
Start by inheriting the base class `Pipeline` with the 4 methods needed to implement `preprocess`,
`_forward`, `postprocess`, and `_sanitize_parameters`.
```python
from transformers import Pipeline
class MyPipeline(Pipeline):
def _sanitize_parameters(self, **kwargs):
preprocess_kwargs = {}
if "maybe_arg" in kwargs:
preprocess_kwargs["maybe_arg"] = kwargs["maybe_arg"]
return preprocess_kwargs, {}, {}
def preprocess(self, inputs, maybe_arg=2):
model_input = Tensor(inputs["input_ids"])
return {"model_input": model_input}
def _forward(self, model_inputs):
# model_inputs == {"model_input": model_input}
outputs = self.model(**model_inputs)
# Maybe {"logits": Tensor(...)}
return outputs
def postprocess(self, model_outputs):
best_class = model_outputs["logits"].softmax(-1)
return best_class
```
The structure of this breakdown is to support relatively seamless support for CPU/GPU, while supporting doing
pre/postprocessing on the CPU on different threads
`preprocess` will take the originally defined inputs, and turn them into something feedable to the model. It might
contain more information and is usually a `Dict`.
`_forward` is the implementation detail and is not meant to be called directly. `forward` is the preferred
called method as it contains safeguards to make sure everything is working on the expected device. If anything is
linked to a real model it belongs in the `_forward` method, anything else is in the preprocess/postprocess.
`postprocess` methods will take the output of `_forward` and turn it into the final output that was decided
earlier.
`_sanitize_parameters` exists to allow users to pass any parameters whenever they wish, be it at initialization
time `pipeline(...., maybe_arg=4)` or at call time `pipe = pipeline(...); output = pipe(...., maybe_arg=4)`.
The returns of `_sanitize_parameters` are the 3 dicts of kwargs that will be passed directly to `preprocess`,
`_forward`, and `postprocess`. Don't fill anything if the caller didn't call with any extra parameter. That
allows to keep the default arguments in the function definition which is always more "natural".
A classic example would be a `top_k` argument in the post processing in classification tasks.
```python
>>> pipe = pipeline("my-new-task")
>>> pipe("This is a test")
[{"label": "1-star", "score": 0.8}, {"label": "2-star", "score": 0.1}, {"label": "3-star", "score": 0.05}
{"label": "4-star", "score": 0.025}, {"label": "5-star", "score": 0.025}]
>>> pipe("This is a test", top_k=2)
[{"label": "1-star", "score": 0.8}, {"label": "2-star", "score": 0.1}]
```
In order to achieve that, we'll update our `postprocess` method with a default parameter to `5`. and edit
`_sanitize_parameters` to allow this new parameter.
```python
def postprocess(self, model_outputs, top_k=5):
best_class = model_outputs["logits"].softmax(-1)
# Add logic to handle top_k
return best_class
def _sanitize_parameters(self, **kwargs):
preprocess_kwargs = {}
if "maybe_arg" in kwargs:
preprocess_kwargs["maybe_arg"] = kwargs["maybe_arg"]
postprocess_kwargs = {}
if "top_k" in kwargs:
postprocess_kwargs["top_k"] = kwargs["top_k"]
return preprocess_kwargs, {}, postprocess_kwargs
```
Try to keep the inputs/outputs very simple and ideally JSON-serializable as it makes the pipeline usage very easy
without requiring users to understand new kinds of objects. It's also relatively common to support many different types
of arguments for ease of use (audio files, which can be filenames, URLs or pure bytes)
## Adding it to the list of supported tasks
To register your `new-task` to the list of supported tasks, you have to add it to the `PIPELINE_REGISTRY`:
```python
from transformers.pipelines import PIPELINE_REGISTRY
PIPELINE_REGISTRY.register_pipeline(
"new-task",
pipeline_class=MyPipeline,
pt_model=AutoModelForSequenceClassification,
)
```
You can specify a default model if you want, in which case it should come with a specific revision (which can be the name of a branch or a commit hash, here we took `"abcdef"`) as well as the type:
```python
PIPELINE_REGISTRY.register_pipeline(
"new-task",
pipeline_class=MyPipeline,
pt_model=AutoModelForSequenceClassification,
default={"pt": ("user/awesome_model", "abcdef")},
type="text", # current support type: text, audio, image, multimodal
)
```
## Share your pipeline on the Hub
To share your custom pipeline on the Hub, you just have to save the custom code of your `Pipeline` subclass in a
python file. For instance, let's say we want to use a custom pipeline for sentence pair classification like this:
```py
import numpy as np
from transformers import Pipeline
def softmax(outputs):
maxes = np.max(outputs, axis=-1, keepdims=True)
shifted_exp = np.exp(outputs - maxes)
return shifted_exp / shifted_exp.sum(axis=-1, keepdims=True)
class PairClassificationPipeline(Pipeline):
def _sanitize_parameters(self, **kwargs):
preprocess_kwargs = {}
if "second_text" in kwargs:
preprocess_kwargs["second_text"] = kwargs["second_text"]
return preprocess_kwargs, {}, {}
def preprocess(self, text, second_text=None):
return self.tokenizer(text, text_pair=second_text, return_tensors=self.framework)
def _forward(self, model_inputs):
return self.model(**model_inputs)
def postprocess(self, model_outputs):
logits = model_outputs.logits[0].numpy()
probabilities = softmax(logits)
best_class = np.argmax(probabilities)
label = self.model.config.id2label[best_class]
score = probabilities[best_class].item()
logits = logits.tolist()
return {"label": label, "score": score, "logits": logits}
```
The implementation is framework agnostic, and will work for PyTorch and TensorFlow models. If we have saved this in
a file named `pair_classification.py`, we can then import it and register it like this:
```py
from pair_classification import PairClassificationPipeline
from transformers.pipelines import PIPELINE_REGISTRY
from transformers import AutoModelForSequenceClassification, TFAutoModelForSequenceClassification
PIPELINE_REGISTRY.register_pipeline(
"pair-classification",
pipeline_class=PairClassificationPipeline,
pt_model=AutoModelForSequenceClassification,
tf_model=TFAutoModelForSequenceClassification,
)
```
Once this is done, we can use it with a pretrained model. For instance `sgugger/finetuned-bert-mrpc` has been
fine-tuned on the MRPC dataset, which classifies pairs of sentences as paraphrases or not.
```py
from transformers import pipeline
classifier = pipeline("pair-classification", model="sgugger/finetuned-bert-mrpc")
```
Then we can share it on the Hub by using the `save_pretrained` method in a `Repository`:
```py
from huggingface_hub import Repository
repo = Repository("test-dynamic-pipeline", clone_from="{your_username}/test-dynamic-pipeline")
classifier.save_pretrained("test-dynamic-pipeline")
repo.push_to_hub()
```
This will copy the file where you defined `PairClassificationPipeline` inside the folder `"test-dynamic-pipeline"`,
along with saving the model and tokenizer of the pipeline, before pushing everything into the repository
`{your_username}/test-dynamic-pipeline`. After that, anyone can use it as long as they provide the option
`trust_remote_code=True`:
```py
from transformers import pipeline
classifier = pipeline(model="{your_username}/test-dynamic-pipeline", trust_remote_code=True)
```
## Add the pipeline to 🤗 Transformers
If you want to contribute your pipeline to 🤗 Transformers, you will need to add a new module in the `pipelines` submodule
with the code of your pipeline, then add it to the list of tasks defined in `pipelines/__init__.py`.
Then you will need to add tests. Create a new file `tests/test_pipelines_MY_PIPELINE.py` with examples of the other tests.
The `run_pipeline_test` function will be very generic and run on small random models on every possible
architecture as defined by `model_mapping` and `tf_model_mapping`.
This is very important to test future compatibility, meaning if someone adds a new model for
`XXXForQuestionAnswering` then the pipeline test will attempt to run on it. Because the models are random it's
impossible to check for actual values, that's why there is a helper `ANY` that will simply attempt to match the
output of the pipeline TYPE.
You also *need* to implement 2 (ideally 4) tests.
- `test_small_model_pt` : Define 1 small model for this pipeline (doesn't matter if the results don't make sense)
and test the pipeline outputs. The results should be the same as `test_small_model_tf`.
- `test_small_model_tf` : Define 1 small model for this pipeline (doesn't matter if the results don't make sense)
and test the pipeline outputs. The results should be the same as `test_small_model_pt`.
- `test_large_model_pt` (`optional`): Tests the pipeline on a real pipeline where the results are supposed to
make sense. These tests are slow and should be marked as such. Here the goal is to showcase the pipeline and to make
sure there is no drift in future releases.
- `test_large_model_tf` (`optional`): Tests the pipeline on a real pipeline where the results are supposed to
make sense. These tests are slow and should be marked as such. Here the goal is to showcase the pipeline and to make
sure there is no drift in future releases.
| huggingface/transformers/blob/main/docs/source/en/add_new_pipeline.md |
Training a masked language model end-to-end from scratch on TPUs
In this example, we're going to demonstrate how to train a TensorFlow model from 🤗 Transformers from scratch. If you're interested in some background theory on training Hugging Face models with TensorFlow on TPU, please check out our
[tutorial doc](https://huggingface.co/docs/transformers/main/perf_train_tpu_tf) on this topic!
If you're interested in smaller-scale TPU training from a pre-trained checkpoint, you can also check out the [TPU fine-tuning example](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/tpu_training-tf.ipynb).
This example will demonstrate pre-training language models at the 100M-1B parameter scale, similar to BERT or GPT-2. More concretely, we will show how to train a [RoBERTa](https://huggingface.co/docs/transformers/model_doc/roberta) (base model) from scratch on the [WikiText dataset (v1)](https://huggingface.co/datasets/wikitext).
We've tried to ensure that all the practices we show you here are scalable, though - with relatively few changes, the code could be scaled up to much larger models.
Google's gargantuan [PaLM model](https://arxiv.org/abs/2204.02311), with
over 500B parameters, is a good example of how far you can go with pure TPU training, though gathering the dataset and the budget to train at that scale is not an easy task!
### Table of contents
- [Setting up a TPU-VM](#setting-up-a-tpu-vm)
- [Training a tokenizer](#training-a-tokenizer)
- [Preparing the dataset](#preparing-the-dataset)
- [Training the model](#training-the-model)
- [Inference](#inference)
## Setting up a TPU-VM
Since this example focuses on using TPUs, the first step is to set up access to TPU hardware. For this example, we chose to use a TPU v3-8 VM. Follow [this guide](https://cloud.google.com/tpu/docs/run-calculation-tensorflow) to quickly create a TPU VM with TensorFlow pre-installed.
> 💡 **Note**: You don't need a TPU-enabled hardware for tokenizer training and TFRecord shard preparation.
## Training a tokenizer
To train a language model from scratch, the first step is to tokenize text. In most Hugging Face examples, we begin from a pre-trained model and use its tokenizer. However, in this example, we're going to train a tokenizer from scratch as well. The script for this is `train_unigram.py`. An example command is:
```bash
python train_unigram.py --batch_size 1000 --vocab_size 25000 --export_to_hub
```
The script will automatically load the `train` split of the WikiText dataset and train a [Unigram tokenizer](https://huggingface.co/course/chapter6/7?fw=pt) on it.
> 💡 **Note**: In order for `export_to_hub` to work, you must authenticate yourself with the `huggingface-cli`. Run `huggingface-cli login` and follow the on-screen instructions.
## Preparing the dataset
The next step is to prepare the dataset. This consists of loading a text dataset from the Hugging Face Hub, tokenizing it and grouping it into chunks of a fixed length ready for training. The script for this is `prepare_tfrecord_shards.py`.
The reason we create TFRecord output files from this step is that these files work well with [`tf.data` pipelines](https://www.tensorflow.org/guide/data_performance). This makes them very suitable for scalable TPU training - the dataset can easily be sharded and read in parallel just by tweaking a few parameters in the pipeline. An example command is:
```bash
python prepare_tfrecord_shards.py \
--tokenizer_name_or_path tf-tpu/unigram-tokenizer-wikitext \
--shard_size 5000 \
--split test
--max_length 128 \
--output_dir gs://tf-tpu-training-resources
```
**Notes**:
* While running the above script, you need to specify the `split` accordingly. The example command above will only filter the `test` split of the dataset.
* If you append `gs://` in your `output_dir` the TFRecord shards will be directly serialized to a Google Cloud Storage (GCS) bucket. Ensure that you have already [created the GCS bucket](https://cloud.google.com/storage/docs).
* If you're using a TPU node, you must stream data from a GCS bucket. Otherwise, if you're using a TPU VM,you can store the data locally. You may need to [attach](https://cloud.google.com/tpu/docs/setup-persistent-disk) a persistent storage to the VM.
* Additional CLI arguments are also supported. We encourage you to run `python prepare_tfrecord_shards.py -h` to know more about them.
## Training the model
Once that's done, the model is ready for training. By default, training takes place on TPU, but you can use the `--no_tpu` flag to train on CPU for testing purposes. An example command is:
```bash
python3 run_mlm.py \
--train_dataset gs://tf-tpu-training-resources/train/ \
--eval_dataset gs://tf-tpu-training-resources/validation/ \
--tokenizer tf-tpu/unigram-tokenizer-wikitext \
--output_dir trained_model
```
If you had specified a `hub_model_id` while launching training, then your model will be pushed to a model repository on the Hugging Face Hub. You can find such an example repository here:
[tf-tpu/roberta-base-epochs-500-no-wd](https://huggingface.co/tf-tpu/roberta-base-epochs-500-no-wd).
## Inference
Once the model is trained, you can use 🤗 Pipelines to perform inference:
```python
from transformers import pipeline
model_id = "tf-tpu/roberta-base-epochs-500-no-wd"
unmasker = pipeline("fill-mask", model=model_id, framework="tf")
unmasker("Goal of my life is to [MASK].")
[{'score': 0.1003185287117958,
'token': 52,
'token_str': 'be',
'sequence': 'Goal of my life is to be.'},
{'score': 0.032648514956235886,
'token': 5,
'token_str': '',
'sequence': 'Goal of my life is to .'},
{'score': 0.02152673341333866,
'token': 138,
'token_str': 'work',
'sequence': 'Goal of my life is to work.'},
{'score': 0.019547373056411743,
'token': 984,
'token_str': 'act',
'sequence': 'Goal of my life is to act.'},
{'score': 0.01939118467271328,
'token': 73,
'token_str': 'have',
'sequence': 'Goal of my life is to have.'}]
```
You can also try out inference using the [Inference Widget](https://huggingface.co/tf-tpu/roberta-base-epochs-500-no-wd?text=Goal+of+my+life+is+to+%5BMASK%5D.) from the model page. | huggingface/transformers/blob/main/examples/tensorflow/language-modeling-tpu/README.md |
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Image captioning
[[open-in-colab]]
Image captioning is the task of predicting a caption for a given image. Common real world applications of it include
aiding visually impaired people that can help them navigate through different situations. Therefore, image captioning
helps to improve content accessibility for people by describing images to them.
This guide will show you how to:
* Fine-tune an image captioning model.
* Use the fine-tuned model for inference.
Before you begin, make sure you have all the necessary libraries installed:
```bash
pip install transformers datasets evaluate -q
pip install jiwer -q
```
We encourage you to log in to your Hugging Face account so you can upload and share your model with the community. When prompted, enter your token to log in:
```python
from huggingface_hub import notebook_login
notebook_login()
```
## Load the Pokémon BLIP captions dataset
Use the 🤗 Dataset library to load a dataset that consists of {image-caption} pairs. To create your own image captioning dataset
in PyTorch, you can follow [this notebook](https://github.com/NielsRogge/Transformers-Tutorials/blob/master/GIT/Fine_tune_GIT_on_an_image_captioning_dataset.ipynb).
```python
from datasets import load_dataset
ds = load_dataset("lambdalabs/pokemon-blip-captions")
ds
```
```bash
DatasetDict({
train: Dataset({
features: ['image', 'text'],
num_rows: 833
})
})
```
The dataset has two features, `image` and `text`.
<Tip>
Many image captioning datasets contain multiple captions per image. In those cases, a common strategy is to randomly sample a caption amongst the available ones during training.
</Tip>
Split the dataset’s train split into a train and test set with the [~datasets.Dataset.train_test_split] method:
```python
ds = ds["train"].train_test_split(test_size=0.1)
train_ds = ds["train"]
test_ds = ds["test"]
```
Let's visualize a couple of samples from the training set.
```python
from textwrap import wrap
import matplotlib.pyplot as plt
import numpy as np
def plot_images(images, captions):
plt.figure(figsize=(20, 20))
for i in range(len(images)):
ax = plt.subplot(1, len(images), i + 1)
caption = captions[i]
caption = "\n".join(wrap(caption, 12))
plt.title(caption)
plt.imshow(images[i])
plt.axis("off")
sample_images_to_visualize = [np.array(train_ds[i]["image"]) for i in range(5)]
sample_captions = [train_ds[i]["text"] for i in range(5)]
plot_images(sample_images_to_visualize, sample_captions)
```
<div class="flex justify-center">
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/sample_training_images_image_cap.png" alt="Sample training images"/>
</div>
## Preprocess the dataset
Since the dataset has two modalities (image and text), the pre-processing pipeline will preprocess images and the captions.
To do so, load the processor class associated with the model you are about to fine-tune.
```python
from transformers import AutoProcessor
checkpoint = "microsoft/git-base"
processor = AutoProcessor.from_pretrained(checkpoint)
```
The processor will internally pre-process the image (which includes resizing, and pixel scaling) and tokenize the caption.
```python
def transforms(example_batch):
images = [x for x in example_batch["image"]]
captions = [x for x in example_batch["text"]]
inputs = processor(images=images, text=captions, padding="max_length")
inputs.update({"labels": inputs["input_ids"]})
return inputs
train_ds.set_transform(transforms)
test_ds.set_transform(transforms)
```
With the dataset ready, you can now set up the model for fine-tuning.
## Load a base model
Load the ["microsoft/git-base"](https://huggingface.co/microsoft/git-base) into a [`AutoModelForCausalLM`](https://huggingface.co/docs/transformers/model_doc/auto#transformers.AutoModelForCausalLM) object.
```python
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained(checkpoint)
```
## Evaluate
Image captioning models are typically evaluated with the [Rouge Score](https://huggingface.co/spaces/evaluate-metric/rouge) or [Word Error Rate](https://huggingface.co/spaces/evaluate-metric/wer). For this guide, you will use the Word Error Rate (WER).
We use the 🤗 Evaluate library to do so. For potential limitations and other gotchas of the WER, refer to [this guide](https://huggingface.co/spaces/evaluate-metric/wer).
```python
from evaluate import load
import torch
wer = load("wer")
def compute_metrics(eval_pred):
logits, labels = eval_pred
predicted = logits.argmax(-1)
decoded_labels = processor.batch_decode(labels, skip_special_tokens=True)
decoded_predictions = processor.batch_decode(predicted, skip_special_tokens=True)
wer_score = wer.compute(predictions=decoded_predictions, references=decoded_labels)
return {"wer_score": wer_score}
```
## Train!
Now, you are ready to start fine-tuning the model. You will use the 🤗 [`Trainer`] for this.
First, define the training arguments using [`TrainingArguments`].
```python
from transformers import TrainingArguments, Trainer
model_name = checkpoint.split("/")[1]
training_args = TrainingArguments(
output_dir=f"{model_name}-pokemon",
learning_rate=5e-5,
num_train_epochs=50,
fp16=True,
per_device_train_batch_size=32,
per_device_eval_batch_size=32,
gradient_accumulation_steps=2,
save_total_limit=3,
evaluation_strategy="steps",
eval_steps=50,
save_strategy="steps",
save_steps=50,
logging_steps=50,
remove_unused_columns=False,
push_to_hub=True,
label_names=["labels"],
load_best_model_at_end=True,
)
```
Then pass them along with the datasets and the model to 🤗 Trainer.
```python
trainer = Trainer(
model=model,
args=training_args,
train_dataset=train_ds,
eval_dataset=test_ds,
compute_metrics=compute_metrics,
)
```
To start training, simply call [`~Trainer.train`] on the [`Trainer`] object.
```python
trainer.train()
```
You should see the training loss drop smoothly as training progresses.
Once training is completed, share your model to the Hub with the [`~Trainer.push_to_hub`] method so everyone can use your model:
```python
trainer.push_to_hub()
```
## Inference
Take a sample image from `test_ds` to test the model.
```python
from PIL import Image
import requests
url = "https://huggingface.co/datasets/sayakpaul/sample-datasets/resolve/main/pokemon.png"
image = Image.open(requests.get(url, stream=True).raw)
image
```
<div class="flex justify-center">
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/test_image_image_cap.png" alt="Test image"/>
</div>
Prepare image for the model.
```python
device = "cuda" if torch.cuda.is_available() else "cpu"
inputs = processor(images=image, return_tensors="pt").to(device)
pixel_values = inputs.pixel_values
```
Call [`generate`] and decode the predictions.
```python
generated_ids = model.generate(pixel_values=pixel_values, max_length=50)
generated_caption = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(generated_caption)
```
```bash
a drawing of a pink and blue pokemon
```
Looks like the fine-tuned model generated a pretty good caption!
| huggingface/transformers/blob/main/docs/source/en/tasks/image_captioning.md |
!--Copyright 2022 The HuggingFace Team and Microsoft. All rights reserved.
Licensed under the MIT License; you may not use this file except in compliance with
the License.
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Graphormer
## Overview
The Graphormer model was proposed in [Do Transformers Really Perform Bad for Graph Representation?](https://arxiv.org/abs/2106.05234) by
Chengxuan Ying, Tianle Cai, Shengjie Luo, Shuxin Zheng, Guolin Ke, Di He, Yanming Shen and Tie-Yan Liu. It is a Graph Transformer model, modified to allow computations on graphs instead of text sequences by generating embeddings and features of interest during preprocessing and collation, then using a modified attention.
The abstract from the paper is the following:
*The Transformer architecture has become a dominant choice in many domains, such as natural language processing and computer vision. Yet, it has not achieved competitive performance on popular leaderboards of graph-level prediction compared to mainstream GNN variants. Therefore, it remains a mystery how Transformers could perform well for graph representation learning. In this paper, we solve this mystery by presenting Graphormer, which is built upon the standard Transformer architecture, and could attain excellent results on a broad range of graph representation learning tasks, especially on the recent OGB Large-Scale Challenge. Our key insight to utilizing Transformer in the graph is the necessity of effectively encoding the structural information of a graph into the model. To this end, we propose several simple yet effective structural encoding methods to help Graphormer better model graph-structured data. Besides, we mathematically characterize the expressive power of Graphormer and exhibit that with our ways of encoding the structural information of graphs, many popular GNN variants could be covered as the special cases of Graphormer.*
This model was contributed by [clefourrier](https://huggingface.co/clefourrier). The original code can be found [here](https://github.com/microsoft/Graphormer).
## Usage tips
This model will not work well on large graphs (more than 100 nodes/edges), as it will make the memory explode.
You can reduce the batch size, increase your RAM, or decrease the `UNREACHABLE_NODE_DISTANCE` parameter in algos_graphormer.pyx, but it will be hard to go above 700 nodes/edges.
This model does not use a tokenizer, but instead a special collator during training.
## GraphormerConfig
[[autodoc]] GraphormerConfig
## GraphormerModel
[[autodoc]] GraphormerModel
- forward
## GraphormerForGraphClassification
[[autodoc]] GraphormerForGraphClassification
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/graphormer.md |
!--Copyright 2022 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# YOLOS
## Overview
The YOLOS model was proposed in [You Only Look at One Sequence: Rethinking Transformer in Vision through Object Detection](https://arxiv.org/abs/2106.00666) by Yuxin Fang, Bencheng Liao, Xinggang Wang, Jiemin Fang, Jiyang Qi, Rui Wu, Jianwei Niu, Wenyu Liu.
YOLOS proposes to just leverage the plain [Vision Transformer (ViT)](vit) for object detection, inspired by DETR. It turns out that a base-sized encoder-only Transformer can also achieve 42 AP on COCO, similar to DETR and much more complex frameworks such as Faster R-CNN.
The abstract from the paper is the following:
*Can Transformer perform 2D object- and region-level recognition from a pure sequence-to-sequence perspective with minimal knowledge about the 2D spatial structure? To answer this question, we present You Only Look at One Sequence (YOLOS), a series of object detection models based on the vanilla Vision Transformer with the fewest possible modifications, region priors, as well as inductive biases of the target task. We find that YOLOS pre-trained on the mid-sized ImageNet-1k dataset only can already achieve quite competitive performance on the challenging COCO object detection benchmark, e.g., YOLOS-Base directly adopted from BERT-Base architecture can obtain 42.0 box AP on COCO val. We also discuss the impacts as well as limitations of current pre-train schemes and model scaling strategies for Transformer in vision through YOLOS.*
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/yolos_architecture.png"
alt="drawing" width="600"/>
<small> YOLOS architecture. Taken from the <a href="https://arxiv.org/abs/2106.00666">original paper</a>.</small>
This model was contributed by [nielsr](https://huggingface.co/nielsr). The original code can be found [here](https://github.com/hustvl/YOLOS).
## Resources
A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with YOLOS.
<PipelineTag pipeline="object-detection"/>
- All example notebooks illustrating inference + fine-tuning [`YolosForObjectDetection`] on a custom dataset can be found [here](https://github.com/NielsRogge/Transformers-Tutorials/tree/master/YOLOS).
- See also: [Object detection task guide](../tasks/object_detection)
If you're interested in submitting a resource to be included here, please feel free to open a Pull Request and we'll review it! The resource should ideally demonstrate something new instead of duplicating an existing resource.
<Tip>
Use [`YolosImageProcessor`] for preparing images (and optional targets) for the model. Contrary to [DETR](detr), YOLOS doesn't require a `pixel_mask` to be created.
</Tip>
## YolosConfig
[[autodoc]] YolosConfig
## YolosImageProcessor
[[autodoc]] YolosImageProcessor
- preprocess
- pad
- post_process_object_detection
## YolosFeatureExtractor
[[autodoc]] YolosFeatureExtractor
- __call__
- pad
- post_process_object_detection
## YolosModel
[[autodoc]] YolosModel
- forward
## YolosForObjectDetection
[[autodoc]] YolosForObjectDetection
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/yolos.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# ELECTRA
<div class="flex flex-wrap space-x-1">
<a href="https://huggingface.co/models?filter=electra">
<img alt="Models" src="https://img.shields.io/badge/All_model_pages-electra-blueviolet">
</a>
<a href="https://huggingface.co/spaces/docs-demos/electra_large_discriminator_squad2_512">
<img alt="Spaces" src="https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-blue">
</a>
</div>
## Overview
The ELECTRA model was proposed in the paper [ELECTRA: Pre-training Text Encoders as Discriminators Rather Than
Generators](https://openreview.net/pdf?id=r1xMH1BtvB). ELECTRA is a new pretraining approach which trains two
transformer models: the generator and the discriminator. The generator's role is to replace tokens in a sequence, and
is therefore trained as a masked language model. The discriminator, which is the model we're interested in, tries to
identify which tokens were replaced by the generator in the sequence.
The abstract from the paper is the following:
*Masked language modeling (MLM) pretraining methods such as BERT corrupt the input by replacing some tokens with [MASK]
and then train a model to reconstruct the original tokens. While they produce good results when transferred to
downstream NLP tasks, they generally require large amounts of compute to be effective. As an alternative, we propose a
more sample-efficient pretraining task called replaced token detection. Instead of masking the input, our approach
corrupts it by replacing some tokens with plausible alternatives sampled from a small generator network. Then, instead
of training a model that predicts the original identities of the corrupted tokens, we train a discriminative model that
predicts whether each token in the corrupted input was replaced by a generator sample or not. Thorough experiments
demonstrate this new pretraining task is more efficient than MLM because the task is defined over all input tokens
rather than just the small subset that was masked out. As a result, the contextual representations learned by our
approach substantially outperform the ones learned by BERT given the same model size, data, and compute. The gains are
particularly strong for small models; for example, we train a model on one GPU for 4 days that outperforms GPT (trained
using 30x more compute) on the GLUE natural language understanding benchmark. Our approach also works well at scale,
where it performs comparably to RoBERTa and XLNet while using less than 1/4 of their compute and outperforms them when
using the same amount of compute.*
This model was contributed by [lysandre](https://huggingface.co/lysandre). The original code can be found [here](https://github.com/google-research/electra).
## Usage tips
- ELECTRA is the pretraining approach, therefore there is nearly no changes done to the underlying model: BERT. The
only change is the separation of the embedding size and the hidden size: the embedding size is generally smaller,
while the hidden size is larger. An additional projection layer (linear) is used to project the embeddings from their
embedding size to the hidden size. In the case where the embedding size is the same as the hidden size, no projection
layer is used.
- ELECTRA is a transformer model pretrained with the use of another (small) masked language model. The inputs are corrupted by that language model, which takes an input text that is randomly masked and outputs a text in which ELECTRA has to predict which token is an original and which one has been replaced. Like for GAN training, the small language model is trained for a few steps (but with the original texts as objective, not to fool the ELECTRA model like in a traditional GAN setting) then the ELECTRA model is trained for a few steps.
- The ELECTRA checkpoints saved using [Google Research's implementation](https://github.com/google-research/electra)
contain both the generator and discriminator. The conversion script requires the user to name which model to export
into the correct architecture. Once converted to the HuggingFace format, these checkpoints may be loaded into all
available ELECTRA models, however. This means that the discriminator may be loaded in the
[`ElectraForMaskedLM`] model, and the generator may be loaded in the
[`ElectraForPreTraining`] model (the classification head will be randomly initialized as it
doesn't exist in the generator).
## Resources
- [Text classification task guide](../tasks/sequence_classification)
- [Token classification task guide](../tasks/token_classification)
- [Question answering task guide](../tasks/question_answering)
- [Causal language modeling task guide](../tasks/language_modeling)
- [Masked language modeling task guide](../tasks/masked_language_modeling)
- [Multiple choice task guide](../tasks/multiple_choice)
## ElectraConfig
[[autodoc]] ElectraConfig
## ElectraTokenizer
[[autodoc]] ElectraTokenizer
## ElectraTokenizerFast
[[autodoc]] ElectraTokenizerFast
## Electra specific outputs
[[autodoc]] models.electra.modeling_electra.ElectraForPreTrainingOutput
[[autodoc]] models.electra.modeling_tf_electra.TFElectraForPreTrainingOutput
<frameworkcontent>
<pt>
## ElectraModel
[[autodoc]] ElectraModel
- forward
## ElectraForPreTraining
[[autodoc]] ElectraForPreTraining
- forward
## ElectraForCausalLM
[[autodoc]] ElectraForCausalLM
- forward
## ElectraForMaskedLM
[[autodoc]] ElectraForMaskedLM
- forward
## ElectraForSequenceClassification
[[autodoc]] ElectraForSequenceClassification
- forward
## ElectraForMultipleChoice
[[autodoc]] ElectraForMultipleChoice
- forward
## ElectraForTokenClassification
[[autodoc]] ElectraForTokenClassification
- forward
## ElectraForQuestionAnswering
[[autodoc]] ElectraForQuestionAnswering
- forward
</pt>
<tf>
## TFElectraModel
[[autodoc]] TFElectraModel
- call
## TFElectraForPreTraining
[[autodoc]] TFElectraForPreTraining
- call
## TFElectraForMaskedLM
[[autodoc]] TFElectraForMaskedLM
- call
## TFElectraForSequenceClassification
[[autodoc]] TFElectraForSequenceClassification
- call
## TFElectraForMultipleChoice
[[autodoc]] TFElectraForMultipleChoice
- call
## TFElectraForTokenClassification
[[autodoc]] TFElectraForTokenClassification
- call
## TFElectraForQuestionAnswering
[[autodoc]] TFElectraForQuestionAnswering
- call
</tf>
<jax>
## FlaxElectraModel
[[autodoc]] FlaxElectraModel
- __call__
## FlaxElectraForPreTraining
[[autodoc]] FlaxElectraForPreTraining
- __call__
## FlaxElectraForCausalLM
[[autodoc]] FlaxElectraForCausalLM
- __call__
## FlaxElectraForMaskedLM
[[autodoc]] FlaxElectraForMaskedLM
- __call__
## FlaxElectraForSequenceClassification
[[autodoc]] FlaxElectraForSequenceClassification
- __call__
## FlaxElectraForMultipleChoice
[[autodoc]] FlaxElectraForMultipleChoice
- __call__
## FlaxElectraForTokenClassification
[[autodoc]] FlaxElectraForTokenClassification
- __call__
## FlaxElectraForQuestionAnswering
[[autodoc]] FlaxElectraForQuestionAnswering
- __call__
</jax>
</frameworkcontent>
| huggingface/transformers/blob/main/docs/source/en/model_doc/electra.md |
!--Copyright 2022 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# AltCLIP
## Overview
The AltCLIP model was proposed in [AltCLIP: Altering the Language Encoder in CLIP for Extended Language Capabilities](https://arxiv.org/abs/2211.06679v2) by Zhongzhi Chen, Guang Liu, Bo-Wen Zhang, Fulong Ye, Qinghong Yang, Ledell Wu. AltCLIP
(Altering the Language Encoder in CLIP) is a neural network trained on a variety of image-text and text-text pairs. By switching CLIP's
text encoder with a pretrained multilingual text encoder XLM-R, we could obtain very close performances with CLIP on almost all tasks, and extended original CLIP's capabilities such as multilingual understanding.
The abstract from the paper is the following:
*In this work, we present a conceptually simple and effective method to train a strong bilingual multimodal representation model.
Starting from the pretrained multimodal representation model CLIP released by OpenAI, we switched its text encoder with a pretrained
multilingual text encoder XLM-R, and aligned both languages and image representations by a two-stage training schema consisting of
teacher learning and contrastive learning. We validate our method through evaluations of a wide range of tasks. We set new state-of-the-art
performances on a bunch of tasks including ImageNet-CN, Flicker30k- CN, and COCO-CN. Further, we obtain very close performances with
CLIP on almost all tasks, suggesting that one can simply alter the text encoder in CLIP for extended capabilities such as multilingual understanding.*
This model was contributed by [jongjyh](https://huggingface.co/jongjyh).
## Usage tips and example
The usage of AltCLIP is very similar to the CLIP. the difference between CLIP is the text encoder. Note that we use bidirectional attention instead of casual attention
and we take the [CLS] token in XLM-R to represent text embedding.
AltCLIP is a multi-modal vision and language model. It can be used for image-text similarity and for zero-shot image
classification. AltCLIP uses a ViT like transformer to get visual features and a bidirectional language model to get the text
features. Both the text and visual features are then projected to a latent space with identical dimension. The dot
product between the projected image and text features is then used as a similar score.
To feed images to the Transformer encoder, each image is split into a sequence of fixed-size non-overlapping patches,
which are then linearly embedded. A [CLS] token is added to serve as representation of an entire image. The authors
also add absolute position embeddings, and feed the resulting sequence of vectors to a standard Transformer encoder.
The [`CLIPImageProcessor`] can be used to resize (or rescale) and normalize images for the model.
The [`AltCLIPProcessor`] wraps a [`CLIPImageProcessor`] and a [`XLMRobertaTokenizer`] into a single instance to both
encode the text and prepare the images. The following example shows how to get the image-text similarity scores using
[`AltCLIPProcessor`] and [`AltCLIPModel`].
```python
>>> from PIL import Image
>>> import requests
>>> from transformers import AltCLIPModel, AltCLIPProcessor
>>> model = AltCLIPModel.from_pretrained("BAAI/AltCLIP")
>>> processor = AltCLIPProcessor.from_pretrained("BAAI/AltCLIP")
>>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
>>> image = Image.open(requests.get(url, stream=True).raw)
>>> inputs = processor(text=["a photo of a cat", "a photo of a dog"], images=image, return_tensors="pt", padding=True)
>>> outputs = model(**inputs)
>>> logits_per_image = outputs.logits_per_image # this is the image-text similarity score
>>> probs = logits_per_image.softmax(dim=1) # we can take the softmax to get the label probabilities
```
<Tip>
This model is based on `CLIPModel`, use it like you would use the original [CLIP](clip).
</Tip>
## AltCLIPConfig
[[autodoc]] AltCLIPConfig
- from_text_vision_configs
## AltCLIPTextConfig
[[autodoc]] AltCLIPTextConfig
## AltCLIPVisionConfig
[[autodoc]] AltCLIPVisionConfig
## AltCLIPProcessor
[[autodoc]] AltCLIPProcessor
## AltCLIPModel
[[autodoc]] AltCLIPModel
- forward
- get_text_features
- get_image_features
## AltCLIPTextModel
[[autodoc]] AltCLIPTextModel
- forward
## AltCLIPVisionModel
[[autodoc]] AltCLIPVisionModel
- forward | huggingface/transformers/blob/main/docs/source/en/model_doc/altclip.md |
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# VipLlava
## Overview
The VipLlava model was proposed in [Making Large Multimodal Models Understand Arbitrary Visual Prompts](https://arxiv.org/abs/2312.00784) by Mu Cai, Haotian Liu, Siva Karthik Mustikovela, Gregory P. Meyer, Yuning Chai, Dennis Park, Yong Jae Lee.
VipLlava enhances the training protocol of Llava by marking images and interact with the model using natural cues like a "red bounding box" or "pointed arrow" during training.
The abstract from the paper is the following:
*While existing large vision-language multimodal models focus on whole image understanding, there is a prominent gap in achieving region-specific comprehension. Current approaches that use textual coordinates or spatial encodings often fail to provide a user-friendly interface for visual prompting. To address this challenge, we introduce a novel multimodal model capable of decoding arbitrary visual prompts. This allows users to intuitively mark images and interact with the model using natural cues like a "red bounding box" or "pointed arrow". Our simple design directly overlays visual markers onto the RGB image, eliminating the need for complex region encodings, yet achieves state-of-the-art performance on region-understanding tasks like Visual7W, PointQA, and Visual Commonsense Reasoning benchmark. Furthermore, we present ViP-Bench, a comprehensive benchmark to assess the capability of models in understanding visual prompts across multiple dimensions, enabling future research in this domain. Code, data, and model are publicly available.*
Tips:
- The architecture is similar than llava architecture except that the multi-modal projector takes a set of concatenated vision hidden states and has an additional layernorm layer on that module.
- We advise users to use `padding_side="left"` when computing batched generation as it leads to more accurate results. Simply make sure to call `processor.tokenizer.padding_side = "left"` before generating.
- Note the model has not been explicitly trained to process multiple images in the same prompt, although this is technically possible, you may experience inaccurate results.
- For better results, we recommend users to prompt the model with the correct prompt format:
```bash
A chat between a curious human and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the human's questions.###Human: <image>\n<prompt>###Assistant:
```
For multiple turns conversation:
```bash
A chat between a curious human and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the human's questions.###Human: <image>\n<prompt1>###Assistant: <answer1>###Human: <prompt2>###Assistant:
```
The original code can be found [here](https://github.com/mu-cai/ViP-LLaVA).
This model was contributed by [Younes Belkada](https://huggingface.co/ybelkada)
## VipLlavaConfig
[[autodoc]] VipLlavaConfig
## VipLlavaForConditionalGeneration
[[autodoc]] VipLlavaForConditionalGeneration
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/vipllava.md |
!---
Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-->
# Adding a new model
This folder contains templates to generate new models that fit the current API and pass all tests. It generates
models in both PyTorch, TensorFlow, and Flax and completes the `__init__.py` and auto-modeling files, and creates the
documentation. Their use is described in the [next section](#cookiecutter-templates).
There is also a CLI tool to generate a new model like an existing one called `transformers-cli add-new-model-like`.
Jump to the [Add new model like section](#add-new-model-like-command) to learn how to use it.
## Cookiecutter Templates
Using the `cookiecutter` utility requires to have all the `dev` dependencies installed. Let's first clone the
repository and install it in our environment:
```shell script
git clone https://github.com/huggingface/transformers
cd transformers
pip install -e ".[dev]"
```
Depending on your OS, and since the number of optional dependencies of Transformers is growing, you might get a
failure with this command. If that's the case make sure to install the Deep Learning framework you are working with
(PyTorch, TensorFlow and/or Flax) then do:
```bash
pip install -e ".[quality]"
```
Once the installation is done, you can use the CLI command `add-new-model` to generate your models:
```shell script
transformers-cli add-new-model
```
This should launch the `cookiecutter` package which should prompt you to fill in the configuration.
The `modelname` should be cased according to the plain text casing, i.e., BERT, RoBERTa, DeBERTa.
```
modelname [<ModelNAME>]:
uppercase_modelname [<MODEL_NAME>]:
lowercase_modelname [<model_name>]:
camelcase_modelname [<ModelName>]:
```
Fill in the `authors` with your team members:
```
authors [The HuggingFace Team]:
```
The checkpoint identifier is the checkpoint that will be used in the examples across the files. Put the name you wish,
as it will appear on the modelhub. Do not forget to include the organisation.
```
checkpoint_identifier [organisation/<model_name>-base-cased]:
```
The tokenizer should either be based on BERT if it behaves exactly like the BERT tokenizer, or a standalone otherwise.
```
Select tokenizer_type:
1 - Based on BERT
2 - Standalone
Choose from 1, 2 [1]:
```
<!---
Choose if your model is an encoder-decoder, or an encoder-only architecture.
If your model is an encoder-only architecture, the generated architecture will be based on the BERT model.
If your model is an encoder-decoder architecture, the generated architecture will be based on the BART model. You can,
of course, edit the files once the generation is complete.
```
Select is_encoder_decoder_model:
1 - True
2 - False
Choose from 1, 2 [1]:
```
-->
Once the command has finished, you should have a total of 7 new files spread across the repository:
```
docs/source/model_doc/<model_name>.md
src/transformers/models/<model_name>/configuration_<model_name>.py
src/transformers/models/<model_name>/modeling_<model_name>.py
src/transformers/models/<model_name>/modeling_tf_<model_name>.py
src/transformers/models/<model_name>/tokenization_<model_name>.py
tests/test_modeling_<model_name>.py
tests/test_modeling_tf_<model_name>.py
```
You can run the tests to ensure that they all pass:
```
python -m pytest ./tests/test_*<model_name>*.py
```
Feel free to modify each file to mimic the behavior of your model.
⚠ You should be careful about the classes preceded by the following line:️
```python
# Copied from transformers.[...]
```
This line ensures that the copy does not diverge from the source. If it *should* diverge, because the implementation
is different, this line needs to be deleted. If you don't delete this line and run `make fix-copies`,
your changes will be overwritten.
Once you have edited the files to fit your architecture, simply re-run the tests (and edit them if a change
is needed!) afterwards to make sure everything works as expected.
Once the files are generated and you are happy with your changes, here's a checklist to ensure that your contribution
will be merged quickly:
- You should run the `make fixup` utility to fix the style of the files and to ensure the code quality meets the
library's standards.
- You should complete the documentation file (`docs/source/model_doc/<model_name>.rst`) so that your model may be
usable.
## Add new model like command
Using the `transformers-cli add-new-model-like` command requires to have all the `dev` dependencies installed. Let's
first clone the repository and install it in our environment:
```shell script
git clone https://github.com/huggingface/transformers
cd transformers
pip install -e ".[dev]"
```
Depending on your OS, and since the number of optional dependencies of Transformers is growing, you might get a
failure with this command. If that's the case make sure to install the Deep Learning framework you are working with
(PyTorch, TensorFlow and/or Flax) then do:
```bash
pip install -e ".[quality]"
```
Once the installation is done, you can use the CLI command `add-new-model-like` to generate your models:
```shell script
transformers-cli add-new-model-like
```
This will start a small questionnaire you have to fill.
```
What identifier would you like to use for the model type of this model?
```
You will have to input the model type of the model you want to clone. The model type can be found in several places:
- inside the configuration of any checkpoint of that model
- the name of the documentation page of that model
For instance the doc page of `BigBirdPegasus` is `https://huggingface.co/docs/transformers/model_doc/bigbird_pegasus`
so its model type is `"bigbird_pegasus"`.
If you make a typo, the command will suggest you the closest model types it can find.
Once this is done, the questionnaire will ask you for the new model name and its various casings:
```
What is the name for your new model?
What identifier would you like to use for the model type of this model?
What name would you like to use for the module of this model?
What prefix (camel-cased) would you like to use for the model classes of this model?
What prefix (upper-cased) would you like to use for the constants relative to this model?
```
From your answer to the first question, defaults will be determined for all others. The first name should be written
as you want your model be named in the doc, with no special casing (like RoBERTa) and from there, you can either stick
with the defaults or change the cased versions.
Next will be the name of the config class to use for this model:
```
What will be the name of the config class for this model?
```
Then, you will be asked for a checkpoint identifier:
```
Please give a checkpoint identifier (on the model Hub) for this new model.
```
This is the checkpoint that will be used in the examples across the files and the integration tests. Put the name you
wish, as it will appear on the Model Hub. Do not forget to include the organisation.
Then you will have to say whether your model re-uses the same processing classes as the model you're cloning:
```
Will your new model use the same processing class as Xxx (XxxTokenizer/XxxFeatureExtractor/XxxImageProcessor)
```
Answer yes if you have no intentions to make any change to the class used for preprocessing. It can use different
files (for instance you can reuse the `BertTokenizer` with a new vocab file).
If you answer no, you will have to give the name of the classes
for the new tokenizer/image processor/feature extractor/processor (depending on the model you're cloning).
Next the questionnaire will ask
```
Should we add # Copied from statements when creating the new modeling file?
```
This is the intenal mechanism used in the library to make sure code copied from various modeling files stay consistent.
If you plan to completely rewrite the modeling file, you should answer no, whereas if you just want to tweak one part
of the model, you should answer yes.
Lastly, the questionnaire will inquire about frameworks:
```
Should we add a version of your new model in all the frameworks implemented by Old Model (xxx)?
```
If you answer yes, the new model will have files for all the frameworks implemented by the model you're cloning.
Otherwise, you will get a new question to select the frameworks you want.
Once the command has finished, you will see a new subfolder in the `src/transformers/models/` folder, with the
necessary files (configuration and modeling files for all frameworks requested, and maybe the processing files,
depending on your choices).
You will also see a doc file and tests for your new models. First you should run
```
make style
make fix-copies
```
and then you can start tweaking your model. You should:
- fill the doc file at `docs/source/model_doc/model_name.md`
- tweak the configuration and modeling files to your need
Once you're done, you can run the tests to ensure that they all pass:
```
python -m pytest ./tests/test_*<model_name>*.py
```
⚠ You should be careful about the classes preceded by the following line:️
```python
# Copied from transformers.[...]
```
This line ensures that the copy does not diverge from the source. If it *should* diverge, because the implementation
is different, this line needs to be deleted. If you don't delete this line and run `make fix-copies`,
your changes will be overwritten.
Once you have edited the files to fit your architecture, simply re-run the tests (and edit them if a change
is needed!) afterwards to make sure everything works as expected.
Once the files are generated and you are happy with your changes, here's a checklist to ensure that your contribution
will be merged quickly:
- You should run the `make fixup` utility to fix the style of the files and to ensure the code quality meets the
library's standards.
- You should add your model to the main README then run `make fix-copies`.
| huggingface/transformers/blob/main/templates/adding_a_new_model/README.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Optimization
The `.optimization` module provides:
- an optimizer with weight decay fixed that can be used to fine-tuned models, and
- several schedules in the form of schedule objects that inherit from `_LRSchedule`:
- a gradient accumulation class to accumulate the gradients of multiple batches
## AdamW (PyTorch)
[[autodoc]] AdamW
## AdaFactor (PyTorch)
[[autodoc]] Adafactor
## AdamWeightDecay (TensorFlow)
[[autodoc]] AdamWeightDecay
[[autodoc]] create_optimizer
## Schedules
### Learning Rate Schedules (Pytorch)
[[autodoc]] SchedulerType
[[autodoc]] get_scheduler
[[autodoc]] get_constant_schedule
[[autodoc]] get_constant_schedule_with_warmup
<img alt="" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/warmup_constant_schedule.png"/>
[[autodoc]] get_cosine_schedule_with_warmup
<img alt="" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/warmup_cosine_schedule.png"/>
[[autodoc]] get_cosine_with_hard_restarts_schedule_with_warmup
<img alt="" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/warmup_cosine_hard_restarts_schedule.png"/>
[[autodoc]] get_linear_schedule_with_warmup
<img alt="" src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/warmup_linear_schedule.png"/>
[[autodoc]] get_polynomial_decay_schedule_with_warmup
[[autodoc]] get_inverse_sqrt_schedule
### Warmup (TensorFlow)
[[autodoc]] WarmUp
## Gradient Strategies
### GradientAccumulator (TensorFlow)
[[autodoc]] GradientAccumulator
| huggingface/transformers/blob/main/docs/source/en/main_classes/optimizer_schedules.md |
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# KOSMOS-2
## Overview
The KOSMOS-2 model was proposed in [Kosmos-2: Grounding Multimodal Large Language Models to the World](https://arxiv.org/abs/2306.14824) by Zhiliang Peng, Wenhui Wang, Li Dong, Yaru Hao, Shaohan Huang, Shuming Ma, Furu Wei.
KOSMOS-2 is a Transformer-based causal language model and is trained using the next-word prediction task on a web-scale
dataset of grounded image-text pairs [GRIT](https://huggingface.co/datasets/zzliang/GRIT). The spatial coordinates of
the bounding boxes in the dataset are converted to a sequence of location tokens, which are appended to their respective
entity text spans (for example, `a snowman` followed by `<patch_index_0044><patch_index_0863>`). The data format is
similar to “hyperlinks” that connect the object regions in an image to their text span in the corresponding caption.
The abstract from the paper is the following:
*We introduce Kosmos-2, a Multimodal Large Language Model (MLLM), enabling new capabilities of perceiving object descriptions (e.g., bounding boxes) and grounding text to the visual world. Specifically, we represent refer expressions as links in Markdown, i.e., ``[text span](bounding boxes)'', where object descriptions are sequences of location tokens. Together with multimodal corpora, we construct large-scale data of grounded image-text pairs (called GrIT) to train the model. In addition to the existing capabilities of MLLMs (e.g., perceiving general modalities, following instructions, and performing in-context learning), Kosmos-2 integrates the grounding capability into downstream applications. We evaluate Kosmos-2 on a wide range of tasks, including (i) multimodal grounding, such as referring expression comprehension, and phrase grounding, (ii) multimodal referring, such as referring expression generation, (iii) perception-language tasks, and (iv) language understanding and generation. This work lays out the foundation for the development of Embodiment AI and sheds light on the big convergence of language, multimodal perception, action, and world modeling, which is a key step toward artificial general intelligence. Code and pretrained models are available at https://aka.ms/kosmos-2.*
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/kosmos_2_overview.jpg"
alt="drawing" width="600"/>
<small> Overview of tasks that KOSMOS-2 can handle. Taken from the <a href="https://arxiv.org/abs/2306.14824">original paper</a>. </small>
## Example
```python
>>> from PIL import Image
>>> import requests
>>> from transformers import AutoProcessor, Kosmos2ForConditionalGeneration
>>> model = Kosmos2ForConditionalGeneration.from_pretrained("microsoft/kosmos-2-patch14-224")
>>> processor = AutoProcessor.from_pretrained("microsoft/kosmos-2-patch14-224")
>>> url = "https://huggingface.co/microsoft/kosmos-2-patch14-224/resolve/main/snowman.jpg"
>>> image = Image.open(requests.get(url, stream=True).raw)
>>> prompt = "<grounding> An image of"
>>> inputs = processor(text=prompt, images=image, return_tensors="pt")
>>> generated_ids = model.generate(
... pixel_values=inputs["pixel_values"],
... input_ids=inputs["input_ids"],
... attention_mask=inputs["attention_mask"],
... image_embeds=None,
... image_embeds_position_mask=inputs["image_embeds_position_mask"],
... use_cache=True,
... max_new_tokens=64,
... )
>>> generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
>>> processed_text = processor.post_process_generation(generated_text, cleanup_and_extract=False)
>>> processed_text
'<grounding> An image of<phrase> a snowman</phrase><object><patch_index_0044><patch_index_0863></object> warming himself by<phrase> a fire</phrase><object><patch_index_0005><patch_index_0911></object>.'
>>> caption, entities = processor.post_process_generation(generated_text)
>>> caption
'An image of a snowman warming himself by a fire.'
>>> entities
[('a snowman', (12, 21), [(0.390625, 0.046875, 0.984375, 0.828125)]), ('a fire', (41, 47), [(0.171875, 0.015625, 0.484375, 0.890625)])]
```
This model was contributed by [Yih-Dar SHIEH](https://huggingface.co/ydshieh). The original code can be found [here](https://github.com/microsoft/unilm/tree/master/kosmos-2).
## Kosmos2Config
[[autodoc]] Kosmos2Config
## Kosmos2ImageProcessor
## Kosmos2Processor
[[autodoc]] Kosmos2Processor
- __call__
## Kosmos2Model
[[autodoc]] Kosmos2Model
- forward
## Kosmos2ForConditionalGeneration
[[autodoc]] Kosmos2ForConditionalGeneration
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/kosmos-2.md |
!--Copyright 2021 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# LayoutLMV2
## Overview
The LayoutLMV2 model was proposed in [LayoutLMv2: Multi-modal Pre-training for Visually-Rich Document Understanding](https://arxiv.org/abs/2012.14740) by Yang Xu, Yiheng Xu, Tengchao Lv, Lei Cui, Furu Wei, Guoxin Wang, Yijuan Lu,
Dinei Florencio, Cha Zhang, Wanxiang Che, Min Zhang, Lidong Zhou. LayoutLMV2 improves [LayoutLM](layoutlm) to obtain
state-of-the-art results across several document image understanding benchmarks:
- information extraction from scanned documents: the [FUNSD](https://guillaumejaume.github.io/FUNSD/) dataset (a
collection of 199 annotated forms comprising more than 30,000 words), the [CORD](https://github.com/clovaai/cord)
dataset (a collection of 800 receipts for training, 100 for validation and 100 for testing), the [SROIE](https://rrc.cvc.uab.es/?ch=13) dataset (a collection of 626 receipts for training and 347 receipts for testing)
and the [Kleister-NDA](https://github.com/applicaai/kleister-nda) dataset (a collection of non-disclosure
agreements from the EDGAR database, including 254 documents for training, 83 documents for validation, and 203
documents for testing).
- document image classification: the [RVL-CDIP](https://www.cs.cmu.edu/~aharley/rvl-cdip/) dataset (a collection of
400,000 images belonging to one of 16 classes).
- document visual question answering: the [DocVQA](https://arxiv.org/abs/2007.00398) dataset (a collection of 50,000
questions defined on 12,000+ document images).
The abstract from the paper is the following:
*Pre-training of text and layout has proved effective in a variety of visually-rich document understanding tasks due to
its effective model architecture and the advantage of large-scale unlabeled scanned/digital-born documents. In this
paper, we present LayoutLMv2 by pre-training text, layout and image in a multi-modal framework, where new model
architectures and pre-training tasks are leveraged. Specifically, LayoutLMv2 not only uses the existing masked
visual-language modeling task but also the new text-image alignment and text-image matching tasks in the pre-training
stage, where cross-modality interaction is better learned. Meanwhile, it also integrates a spatial-aware self-attention
mechanism into the Transformer architecture, so that the model can fully understand the relative positional
relationship among different text blocks. Experiment results show that LayoutLMv2 outperforms strong baselines and
achieves new state-of-the-art results on a wide variety of downstream visually-rich document understanding tasks,
including FUNSD (0.7895 -> 0.8420), CORD (0.9493 -> 0.9601), SROIE (0.9524 -> 0.9781), Kleister-NDA (0.834 -> 0.852),
RVL-CDIP (0.9443 -> 0.9564), and DocVQA (0.7295 -> 0.8672). The pre-trained LayoutLMv2 model is publicly available at
this https URL.*
LayoutLMv2 depends on `detectron2`, `torchvision` and `tesseract`. Run the
following to install them:
```
python -m pip install 'git+https://github.com/facebookresearch/detectron2.git'
python -m pip install torchvision tesseract
```
(If you are developing for LayoutLMv2, note that passing the doctests also requires the installation of these packages.)
## Usage tips
- The main difference between LayoutLMv1 and LayoutLMv2 is that the latter incorporates visual embeddings during
pre-training (while LayoutLMv1 only adds visual embeddings during fine-tuning).
- LayoutLMv2 adds both a relative 1D attention bias as well as a spatial 2D attention bias to the attention scores in
the self-attention layers. Details can be found on page 5 of the [paper](https://arxiv.org/abs/2012.14740).
- Demo notebooks on how to use the LayoutLMv2 model on RVL-CDIP, FUNSD, DocVQA, CORD can be found [here](https://github.com/NielsRogge/Transformers-Tutorials).
- LayoutLMv2 uses Facebook AI's [Detectron2](https://github.com/facebookresearch/detectron2/) package for its visual
backbone. See [this link](https://detectron2.readthedocs.io/en/latest/tutorials/install.html) for installation
instructions.
- In addition to `input_ids`, [`~LayoutLMv2Model.forward`] expects 2 additional inputs, namely
`image` and `bbox`. The `image` input corresponds to the original document image in which the text
tokens occur. The model expects each document image to be of size 224x224. This means that if you have a batch of
document images, `image` should be a tensor of shape (batch_size, 3, 224, 224). This can be either a
`torch.Tensor` or a `Detectron2.structures.ImageList`. You don't need to normalize the channels, as this is
done by the model. Important to note is that the visual backbone expects BGR channels instead of RGB, as all models
in Detectron2 are pre-trained using the BGR format. The `bbox` input are the bounding boxes (i.e. 2D-positions)
of the input text tokens. This is identical to [`LayoutLMModel`]. These can be obtained using an
external OCR engine such as Google's [Tesseract](https://github.com/tesseract-ocr/tesseract) (there's a [Python
wrapper](https://pypi.org/project/pytesseract/) available). Each bounding box should be in (x0, y0, x1, y1)
format, where (x0, y0) corresponds to the position of the upper left corner in the bounding box, and (x1, y1)
represents the position of the lower right corner. Note that one first needs to normalize the bounding boxes to be on
a 0-1000 scale. To normalize, you can use the following function:
```python
def normalize_bbox(bbox, width, height):
return [
int(1000 * (bbox[0] / width)),
int(1000 * (bbox[1] / height)),
int(1000 * (bbox[2] / width)),
int(1000 * (bbox[3] / height)),
]
```
Here, `width` and `height` correspond to the width and height of the original document in which the token
occurs (before resizing the image). Those can be obtained using the Python Image Library (PIL) library for example, as
follows:
```python
from PIL import Image
image = Image.open(
"name_of_your_document - can be a png, jpg, etc. of your documents (PDFs must be converted to images)."
)
width, height = image.size
```
However, this model includes a brand new [`~transformers.LayoutLMv2Processor`] which can be used to directly
prepare data for the model (including applying OCR under the hood). More information can be found in the "Usage"
section below.
- Internally, [`~transformers.LayoutLMv2Model`] will send the `image` input through its visual backbone to
obtain a lower-resolution feature map, whose shape is equal to the `image_feature_pool_shape` attribute of
[`~transformers.LayoutLMv2Config`]. This feature map is then flattened to obtain a sequence of image tokens. As
the size of the feature map is 7x7 by default, one obtains 49 image tokens. These are then concatenated with the text
tokens, and send through the Transformer encoder. This means that the last hidden states of the model will have a
length of 512 + 49 = 561, if you pad the text tokens up to the max length. More generally, the last hidden states
will have a shape of `seq_length` + `image_feature_pool_shape[0]` *
`config.image_feature_pool_shape[1]`.
- When calling [`~transformers.LayoutLMv2Model.from_pretrained`], a warning will be printed with a long list of
parameter names that are not initialized. This is not a problem, as these parameters are batch normalization
statistics, which are going to have values when fine-tuning on a custom dataset.
- If you want to train the model in a distributed environment, make sure to call [`synchronize_batch_norm`] on the
model in order to properly synchronize the batch normalization layers of the visual backbone.
In addition, there's LayoutXLM, which is a multilingual version of LayoutLMv2. More information can be found on
[LayoutXLM's documentation page](layoutxlm).
## Resources
A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with LayoutLMv2. If you're interested in submitting a resource to be included here, please feel free to open a Pull Request and we'll review it! The resource should ideally demonstrate something new instead of duplicating an existing resource.
<PipelineTag pipeline="text-classification"/>
- A notebook on how to [finetune LayoutLMv2 for text-classification on RVL-CDIP dataset](https://colab.research.google.com/github/NielsRogge/Transformers-Tutorials/blob/master/LayoutLMv2/RVL-CDIP/Fine_tuning_LayoutLMv2ForSequenceClassification_on_RVL_CDIP.ipynb).
- See also: [Text classification task guide](../tasks/sequence_classification)
<PipelineTag pipeline="question-answering"/>
- A notebook on how to [finetune LayoutLMv2 for question-answering on DocVQA dataset](https://colab.research.google.com/github/NielsRogge/Transformers-Tutorials/blob/master/LayoutLMv2/DocVQA/Fine_tuning_LayoutLMv2ForQuestionAnswering_on_DocVQA.ipynb).
- See also: [Question answering task guide](../tasks/question_answering)
- See also: [Document question answering task guide](../tasks/document_question_answering)
<PipelineTag pipeline="token-classification"/>
- A notebook on how to [finetune LayoutLMv2 for token-classification on CORD dataset](https://colab.research.google.com/github/NielsRogge/Transformers-Tutorials/blob/master/LayoutLMv2/CORD/Fine_tuning_LayoutLMv2ForTokenClassification_on_CORD.ipynb).
- A notebook on how to [finetune LayoutLMv2 for token-classification on FUNSD dataset](https://colab.research.google.com/github/NielsRogge/Transformers-Tutorials/blob/master/LayoutLMv2/FUNSD/Fine_tuning_LayoutLMv2ForTokenClassification_on_FUNSD_using_HuggingFace_Trainer.ipynb).
- See also: [Token classification task guide](../tasks/token_classification)
## Usage: LayoutLMv2Processor
The easiest way to prepare data for the model is to use [`LayoutLMv2Processor`], which internally
combines a image processor ([`LayoutLMv2ImageProcessor`]) and a tokenizer
([`LayoutLMv2Tokenizer`] or [`LayoutLMv2TokenizerFast`]). The image processor
handles the image modality, while the tokenizer handles the text modality. A processor combines both, which is ideal
for a multi-modal model like LayoutLMv2. Note that you can still use both separately, if you only want to handle one
modality.
```python
from transformers import LayoutLMv2ImageProcessor, LayoutLMv2TokenizerFast, LayoutLMv2Processor
image_processor = LayoutLMv2ImageProcessor() # apply_ocr is set to True by default
tokenizer = LayoutLMv2TokenizerFast.from_pretrained("microsoft/layoutlmv2-base-uncased")
processor = LayoutLMv2Processor(image_processor, tokenizer)
```
In short, one can provide a document image (and possibly additional data) to [`LayoutLMv2Processor`],
and it will create the inputs expected by the model. Internally, the processor first uses
[`LayoutLMv2ImageProcessor`] to apply OCR on the image to get a list of words and normalized
bounding boxes, as well to resize the image to a given size in order to get the `image` input. The words and
normalized bounding boxes are then provided to [`LayoutLMv2Tokenizer`] or
[`LayoutLMv2TokenizerFast`], which converts them to token-level `input_ids`,
`attention_mask`, `token_type_ids`, `bbox`. Optionally, one can provide word labels to the processor,
which are turned into token-level `labels`.
[`LayoutLMv2Processor`] uses [PyTesseract](https://pypi.org/project/pytesseract/), a Python
wrapper around Google's Tesseract OCR engine, under the hood. Note that you can still use your own OCR engine of
choice, and provide the words and normalized boxes yourself. This requires initializing
[`LayoutLMv2ImageProcessor`] with `apply_ocr` set to `False`.
In total, there are 5 use cases that are supported by the processor. Below, we list them all. Note that each of these
use cases work for both batched and non-batched inputs (we illustrate them for non-batched inputs).
**Use case 1: document image classification (training, inference) + token classification (inference), apply_ocr =
True**
This is the simplest case, in which the processor (actually the image processor) will perform OCR on the image to get
the words and normalized bounding boxes.
```python
from transformers import LayoutLMv2Processor
from PIL import Image
processor = LayoutLMv2Processor.from_pretrained("microsoft/layoutlmv2-base-uncased")
image = Image.open(
"name_of_your_document - can be a png, jpg, etc. of your documents (PDFs must be converted to images)."
).convert("RGB")
encoding = processor(
image, return_tensors="pt"
) # you can also add all tokenizer parameters here such as padding, truncation
print(encoding.keys())
# dict_keys(['input_ids', 'token_type_ids', 'attention_mask', 'bbox', 'image'])
```
**Use case 2: document image classification (training, inference) + token classification (inference), apply_ocr=False**
In case one wants to do OCR themselves, one can initialize the image processor with `apply_ocr` set to
`False`. In that case, one should provide the words and corresponding (normalized) bounding boxes themselves to
the processor.
```python
from transformers import LayoutLMv2Processor
from PIL import Image
processor = LayoutLMv2Processor.from_pretrained("microsoft/layoutlmv2-base-uncased", revision="no_ocr")
image = Image.open(
"name_of_your_document - can be a png, jpg, etc. of your documents (PDFs must be converted to images)."
).convert("RGB")
words = ["hello", "world"]
boxes = [[1, 2, 3, 4], [5, 6, 7, 8]] # make sure to normalize your bounding boxes
encoding = processor(image, words, boxes=boxes, return_tensors="pt")
print(encoding.keys())
# dict_keys(['input_ids', 'token_type_ids', 'attention_mask', 'bbox', 'image'])
```
**Use case 3: token classification (training), apply_ocr=False**
For token classification tasks (such as FUNSD, CORD, SROIE, Kleister-NDA), one can also provide the corresponding word
labels in order to train a model. The processor will then convert these into token-level `labels`. By default, it
will only label the first wordpiece of a word, and label the remaining wordpieces with -100, which is the
`ignore_index` of PyTorch's CrossEntropyLoss. In case you want all wordpieces of a word to be labeled, you can
initialize the tokenizer with `only_label_first_subword` set to `False`.
```python
from transformers import LayoutLMv2Processor
from PIL import Image
processor = LayoutLMv2Processor.from_pretrained("microsoft/layoutlmv2-base-uncased", revision="no_ocr")
image = Image.open(
"name_of_your_document - can be a png, jpg, etc. of your documents (PDFs must be converted to images)."
).convert("RGB")
words = ["hello", "world"]
boxes = [[1, 2, 3, 4], [5, 6, 7, 8]] # make sure to normalize your bounding boxes
word_labels = [1, 2]
encoding = processor(image, words, boxes=boxes, word_labels=word_labels, return_tensors="pt")
print(encoding.keys())
# dict_keys(['input_ids', 'token_type_ids', 'attention_mask', 'bbox', 'labels', 'image'])
```
**Use case 4: visual question answering (inference), apply_ocr=True**
For visual question answering tasks (such as DocVQA), you can provide a question to the processor. By default, the
processor will apply OCR on the image, and create [CLS] question tokens [SEP] word tokens [SEP].
```python
from transformers import LayoutLMv2Processor
from PIL import Image
processor = LayoutLMv2Processor.from_pretrained("microsoft/layoutlmv2-base-uncased")
image = Image.open(
"name_of_your_document - can be a png, jpg, etc. of your documents (PDFs must be converted to images)."
).convert("RGB")
question = "What's his name?"
encoding = processor(image, question, return_tensors="pt")
print(encoding.keys())
# dict_keys(['input_ids', 'token_type_ids', 'attention_mask', 'bbox', 'image'])
```
**Use case 5: visual question answering (inference), apply_ocr=False**
For visual question answering tasks (such as DocVQA), you can provide a question to the processor. If you want to
perform OCR yourself, you can provide your own words and (normalized) bounding boxes to the processor.
```python
from transformers import LayoutLMv2Processor
from PIL import Image
processor = LayoutLMv2Processor.from_pretrained("microsoft/layoutlmv2-base-uncased", revision="no_ocr")
image = Image.open(
"name_of_your_document - can be a png, jpg, etc. of your documents (PDFs must be converted to images)."
).convert("RGB")
question = "What's his name?"
words = ["hello", "world"]
boxes = [[1, 2, 3, 4], [5, 6, 7, 8]] # make sure to normalize your bounding boxes
encoding = processor(image, question, words, boxes=boxes, return_tensors="pt")
print(encoding.keys())
# dict_keys(['input_ids', 'token_type_ids', 'attention_mask', 'bbox', 'image'])
```
## LayoutLMv2Config
[[autodoc]] LayoutLMv2Config
## LayoutLMv2FeatureExtractor
[[autodoc]] LayoutLMv2FeatureExtractor
- __call__
## LayoutLMv2ImageProcessor
[[autodoc]] LayoutLMv2ImageProcessor
- preprocess
## LayoutLMv2Tokenizer
[[autodoc]] LayoutLMv2Tokenizer
- __call__
- save_vocabulary
## LayoutLMv2TokenizerFast
[[autodoc]] LayoutLMv2TokenizerFast
- __call__
## LayoutLMv2Processor
[[autodoc]] LayoutLMv2Processor
- __call__
## LayoutLMv2Model
[[autodoc]] LayoutLMv2Model
- forward
## LayoutLMv2ForSequenceClassification
[[autodoc]] LayoutLMv2ForSequenceClassification
## LayoutLMv2ForTokenClassification
[[autodoc]] LayoutLMv2ForTokenClassification
## LayoutLMv2ForQuestionAnswering
[[autodoc]] LayoutLMv2ForQuestionAnswering
| huggingface/transformers/blob/main/docs/source/en/model_doc/layoutlmv2.md |
!---
Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-->
# Token classification
## PyTorch version
Fine-tuning the library models for token classification task such as Named Entity Recognition (NER), Parts-of-speech
tagging (POS) or phrase extraction (CHUNKS). The main scrip `run_ner.py` leverages the 🤗 Datasets library and the Trainer API. You can easily
customize it to your needs if you need extra processing on your datasets.
It will either run on a datasets hosted on our [hub](https://huggingface.co/datasets) or with your own text files for
training and validation, you might just need to add some tweaks in the data preprocessing.
The following example fine-tunes BERT on CoNLL-2003:
```bash
python run_ner.py \
--model_name_or_path bert-base-uncased \
--dataset_name conll2003 \
--output_dir /tmp/test-ner \
--do_train \
--do_eval
```
or just can just run the bash script `run.sh`.
To run on your own training and validation files, use the following command:
```bash
python run_ner.py \
--model_name_or_path bert-base-uncased \
--train_file path_to_train_file \
--validation_file path_to_validation_file \
--output_dir /tmp/test-ner \
--do_train \
--do_eval
```
**Note:** This script only works with models that have a fast tokenizer (backed by the 🤗 Tokenizers library) as it
uses special features of those tokenizers. You can check if your favorite model has a fast tokenizer in
[this table](https://huggingface.co/transformers/index.html#supported-frameworks), if it doesn't you can still use the old version
of the script.
> If your model classification head dimensions do not fit the number of labels in the dataset, you can specify `--ignore_mismatched_sizes` to adapt it.
## Old version of the script
You can find the old version of the PyTorch script [here](https://github.com/huggingface/transformers/blob/main/examples/legacy/token-classification/run_ner.py).
## Pytorch version, no Trainer
Based on the script [run_ner_no_trainer.py](https://github.com/huggingface/transformers/blob/main/examples/pytorch/token-classification/run_ner_no_trainer.py).
Like `run_ner.py`, this script allows you to fine-tune any of the models on the [hub](https://huggingface.co/models) on a
token classification task, either NER, POS or CHUNKS tasks or your own data in a csv or a JSON file. The main difference is that this
script exposes the bare training loop, to allow you to quickly experiment and add any customization you would like.
It offers less options than the script with `Trainer` (for instance you can easily change the options for the optimizer
or the dataloaders directly in the script) but still run in a distributed setup, on TPU and supports mixed precision by
the mean of the [🤗 `Accelerate`](https://github.com/huggingface/accelerate) library. You can use the script normally
after installing it:
```bash
pip install git+https://github.com/huggingface/accelerate
```
then
```bash
export TASK_NAME=ner
python run_ner_no_trainer.py \
--model_name_or_path bert-base-cased \
--dataset_name conll2003 \
--task_name $TASK_NAME \
--max_length 128 \
--per_device_train_batch_size 32 \
--learning_rate 2e-5 \
--num_train_epochs 3 \
--output_dir /tmp/$TASK_NAME/
```
You can then use your usual launchers to run in it in a distributed environment, but the easiest way is to run
```bash
accelerate config
```
and reply to the questions asked. Then
```bash
accelerate test
```
that will check everything is ready for training. Finally, you can launch training with
```bash
export TASK_NAME=ner
accelerate launch run_ner_no_trainer.py \
--model_name_or_path bert-base-cased \
--dataset_name conll2003 \
--task_name $TASK_NAME \
--max_length 128 \
--per_device_train_batch_size 32 \
--learning_rate 2e-5 \
--num_train_epochs 3 \
--output_dir /tmp/$TASK_NAME/
```
This command is the same and will work for:
- a CPU-only setup
- a setup with one GPU
- a distributed training with several GPUs (single or multi node)
- a training on TPUs
Note that this library is in alpha release so your feedback is more than welcome if you encounter any problem using it.
| huggingface/transformers/blob/main/examples/pytorch/token-classification/README.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# BARThez
## Overview
The BARThez model was proposed in [BARThez: a Skilled Pretrained French Sequence-to-Sequence Model](https://arxiv.org/abs/2010.12321) by Moussa Kamal Eddine, Antoine J.-P. Tixier, Michalis Vazirgiannis on 23 Oct,
2020.
The abstract of the paper:
*Inductive transfer learning, enabled by self-supervised learning, have taken the entire Natural Language Processing
(NLP) field by storm, with models such as BERT and BART setting new state of the art on countless natural language
understanding tasks. While there are some notable exceptions, most of the available models and research have been
conducted for the English language. In this work, we introduce BARThez, the first BART model for the French language
(to the best of our knowledge). BARThez was pretrained on a very large monolingual French corpus from past research
that we adapted to suit BART's perturbation schemes. Unlike already existing BERT-based French language models such as
CamemBERT and FlauBERT, BARThez is particularly well-suited for generative tasks, since not only its encoder but also
its decoder is pretrained. In addition to discriminative tasks from the FLUE benchmark, we evaluate BARThez on a novel
summarization dataset, OrangeSum, that we release with this paper. We also continue the pretraining of an already
pretrained multilingual BART on BARThez's corpus, and we show that the resulting model, which we call mBARTHez,
provides a significant boost over vanilla BARThez, and is on par with or outperforms CamemBERT and FlauBERT.*
This model was contributed by [moussakam](https://huggingface.co/moussakam). The Authors' code can be found [here](https://github.com/moussaKam/BARThez).
<Tip>
BARThez implementation is the same as BART, except for tokenization. Refer to [BART documentation](bart) for information on
configuration classes and their parameters. BARThez-specific tokenizers are documented below.
</Tip>
## Resources
- BARThez can be fine-tuned on sequence-to-sequence tasks in a similar way as BART, check:
[examples/pytorch/summarization/](https://github.com/huggingface/transformers/tree/main/examples/pytorch/summarization/README.md).
## BarthezTokenizer
[[autodoc]] BarthezTokenizer
## BarthezTokenizerFast
[[autodoc]] BarthezTokenizerFast
| huggingface/transformers/blob/main/docs/source/en/model_doc/barthez.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# HerBERT
## Overview
The HerBERT model was proposed in [KLEJ: Comprehensive Benchmark for Polish Language Understanding](https://www.aclweb.org/anthology/2020.acl-main.111.pdf) by Piotr Rybak, Robert Mroczkowski, Janusz Tracz, and
Ireneusz Gawlik. It is a BERT-based Language Model trained on Polish Corpora using only MLM objective with dynamic
masking of whole words.
The abstract from the paper is the following:
*In recent years, a series of Transformer-based models unlocked major improvements in general natural language
understanding (NLU) tasks. Such a fast pace of research would not be possible without general NLU benchmarks, which
allow for a fair comparison of the proposed methods. However, such benchmarks are available only for a handful of
languages. To alleviate this issue, we introduce a comprehensive multi-task benchmark for the Polish language
understanding, accompanied by an online leaderboard. It consists of a diverse set of tasks, adopted from existing
datasets for named entity recognition, question-answering, textual entailment, and others. We also introduce a new
sentiment analysis task for the e-commerce domain, named Allegro Reviews (AR). To ensure a common evaluation scheme and
promote models that generalize to different NLU tasks, the benchmark includes datasets from varying domains and
applications. Additionally, we release HerBERT, a Transformer-based model trained specifically for the Polish language,
which has the best average performance and obtains the best results for three out of nine tasks. Finally, we provide an
extensive evaluation, including several standard baselines and recently proposed, multilingual Transformer-based
models.*
This model was contributed by [rmroczkowski](https://huggingface.co/rmroczkowski). The original code can be found
[here](https://github.com/allegro/HerBERT).
## Usage example
```python
>>> from transformers import HerbertTokenizer, RobertaModel
>>> tokenizer = HerbertTokenizer.from_pretrained("allegro/herbert-klej-cased-tokenizer-v1")
>>> model = RobertaModel.from_pretrained("allegro/herbert-klej-cased-v1")
>>> encoded_input = tokenizer.encode("Kto ma lepszą sztukę, ma lepszy rząd – to jasne.", return_tensors="pt")
>>> outputs = model(encoded_input)
>>> # HerBERT can also be loaded using AutoTokenizer and AutoModel:
>>> import torch
>>> from transformers import AutoModel, AutoTokenizer
>>> tokenizer = AutoTokenizer.from_pretrained("allegro/herbert-klej-cased-tokenizer-v1")
>>> model = AutoModel.from_pretrained("allegro/herbert-klej-cased-v1")
```
<Tip>
Herbert implementation is the same as `BERT` except for the tokenization method. Refer to [BERT documentation](bert)
for API reference and examples.
</Tip>
## HerbertTokenizer
[[autodoc]] HerbertTokenizer
## HerbertTokenizerFast
[[autodoc]] HerbertTokenizerFast
| huggingface/transformers/blob/main/docs/source/en/model_doc/herbert.md |
!--Copyright 2021 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# SEW
## Overview
SEW (Squeezed and Efficient Wav2Vec) was proposed in [Performance-Efficiency Trade-offs in Unsupervised Pre-training
for Speech Recognition](https://arxiv.org/abs/2109.06870) by Felix Wu, Kwangyoun Kim, Jing Pan, Kyu Han, Kilian Q.
Weinberger, Yoav Artzi.
The abstract from the paper is the following:
*This paper is a study of performance-efficiency trade-offs in pre-trained models for automatic speech recognition
(ASR). We focus on wav2vec 2.0, and formalize several architecture designs that influence both the model performance
and its efficiency. Putting together all our observations, we introduce SEW (Squeezed and Efficient Wav2vec), a
pre-trained model architecture with significant improvements along both performance and efficiency dimensions across a
variety of training setups. For example, under the 100h-960h semi-supervised setup on LibriSpeech, SEW achieves a 1.9x
inference speedup compared to wav2vec 2.0, with a 13.5% relative reduction in word error rate. With a similar inference
time, SEW reduces word error rate by 25-50% across different model sizes.*
This model was contributed by [anton-l](https://huggingface.co/anton-l).
## Usage tips
- SEW is a speech model that accepts a float array corresponding to the raw waveform of the speech signal.
- SEWForCTC is fine-tuned using connectionist temporal classification (CTC) so the model output has to be decoded using
[`Wav2Vec2CTCTokenizer`].
## Resources
- [Audio classification task guide](../tasks/audio_classification)
- [Automatic speech recognition task guide](../tasks/asr)
## SEWConfig
[[autodoc]] SEWConfig
## SEWModel
[[autodoc]] SEWModel
- forward
## SEWForCTC
[[autodoc]] SEWForCTC
- forward
## SEWForSequenceClassification
[[autodoc]] SEWForSequenceClassification
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/sew.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# BERTology
There is a growing field of study concerned with investigating the inner working of large-scale transformers like BERT
(that some call "BERTology"). Some good examples of this field are:
- BERT Rediscovers the Classical NLP Pipeline by Ian Tenney, Dipanjan Das, Ellie Pavlick:
https://arxiv.org/abs/1905.05950
- Are Sixteen Heads Really Better than One? by Paul Michel, Omer Levy, Graham Neubig: https://arxiv.org/abs/1905.10650
- What Does BERT Look At? An Analysis of BERT's Attention by Kevin Clark, Urvashi Khandelwal, Omer Levy, Christopher D.
Manning: https://arxiv.org/abs/1906.04341
- CAT-probing: A Metric-based Approach to Interpret How Pre-trained Models for Programming Language Attend Code Structure: https://arxiv.org/abs/2210.04633
In order to help this new field develop, we have included a few additional features in the BERT/GPT/GPT-2 models to
help people access the inner representations, mainly adapted from the great work of Paul Michel
(https://arxiv.org/abs/1905.10650):
- accessing all the hidden-states of BERT/GPT/GPT-2,
- accessing all the attention weights for each head of BERT/GPT/GPT-2,
- retrieving heads output values and gradients to be able to compute head importance score and prune head as explained
in https://arxiv.org/abs/1905.10650.
To help you understand and use these features, we have added a specific example script: [bertology.py](https://github.com/huggingface/transformers/tree/main/examples/research_projects/bertology/run_bertology.py) while extract information and prune a model pre-trained on
GLUE.
| huggingface/transformers/blob/main/docs/source/en/bertology.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Longformer
<div class="flex flex-wrap space-x-1">
<a href="https://huggingface.co/models?filter=longformer">
<img alt="Models" src="https://img.shields.io/badge/All_model_pages-longformer-blueviolet">
</a>
<a href="https://huggingface.co/spaces/docs-demos/longformer-base-4096-finetuned-squadv1">
<img alt="Spaces" src="https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-blue">
</a>
</div>
## Overview
The Longformer model was presented in [Longformer: The Long-Document Transformer](https://arxiv.org/pdf/2004.05150.pdf) by Iz Beltagy, Matthew E. Peters, Arman Cohan.
The abstract from the paper is the following:
*Transformer-based models are unable to process long sequences due to their self-attention operation, which scales
quadratically with the sequence length. To address this limitation, we introduce the Longformer with an attention
mechanism that scales linearly with sequence length, making it easy to process documents of thousands of tokens or
longer. Longformer's attention mechanism is a drop-in replacement for the standard self-attention and combines a local
windowed attention with a task motivated global attention. Following prior work on long-sequence transformers, we
evaluate Longformer on character-level language modeling and achieve state-of-the-art results on text8 and enwik8. In
contrast to most prior work, we also pretrain Longformer and finetune it on a variety of downstream tasks. Our
pretrained Longformer consistently outperforms RoBERTa on long document tasks and sets new state-of-the-art results on
WikiHop and TriviaQA.*
This model was contributed by [beltagy](https://huggingface.co/beltagy). The Authors' code can be found [here](https://github.com/allenai/longformer).
## Usage tips
- Since the Longformer is based on RoBERTa, it doesn't have `token_type_ids`. You don't need to indicate which
token belongs to which segment. Just separate your segments with the separation token `tokenizer.sep_token` (or
`</s>`).
- A transformer model replacing the attention matrices by sparse matrices to go faster. Often, the local context (e.g., what are the two tokens left and right?) is enough to take action for a given token. Some preselected input tokens are still given global attention, but the attention matrix has way less parameters, resulting in a speed-up. See the local attention section for more information.
## Longformer Self Attention
Longformer self attention employs self attention on both a "local" context and a "global" context. Most tokens only
attend "locally" to each other meaning that each token attends to its \\(\frac{1}{2} w\\) previous tokens and
\\(\frac{1}{2} w\\) succeeding tokens with \\(w\\) being the window length as defined in
`config.attention_window`. Note that `config.attention_window` can be of type `List` to define a
different \\(w\\) for each layer. A selected few tokens attend "globally" to all other tokens, as it is
conventionally done for all tokens in `BertSelfAttention`.
Note that "locally" and "globally" attending tokens are projected by different query, key and value matrices. Also note
that every "locally" attending token not only attends to tokens within its window \\(w\\), but also to all "globally"
attending tokens so that global attention is *symmetric*.
The user can define which tokens attend "locally" and which tokens attend "globally" by setting the tensor
`global_attention_mask` at run-time appropriately. All Longformer models employ the following logic for
`global_attention_mask`:
- 0: the token attends "locally",
- 1: the token attends "globally".
For more information please also refer to [`~LongformerModel.forward`] method.
Using Longformer self attention, the memory and time complexity of the query-key matmul operation, which usually
represents the memory and time bottleneck, can be reduced from \\(\mathcal{O}(n_s \times n_s)\\) to
\\(\mathcal{O}(n_s \times w)\\), with \\(n_s\\) being the sequence length and \\(w\\) being the average window
size. It is assumed that the number of "globally" attending tokens is insignificant as compared to the number of
"locally" attending tokens.
For more information, please refer to the official [paper](https://arxiv.org/pdf/2004.05150.pdf).
## Training
[`LongformerForMaskedLM`] is trained the exact same way [`RobertaForMaskedLM`] is
trained and should be used as follows:
```python
input_ids = tokenizer.encode("This is a sentence from [MASK] training data", return_tensors="pt")
mlm_labels = tokenizer.encode("This is a sentence from the training data", return_tensors="pt")
loss = model(input_ids, labels=input_ids, masked_lm_labels=mlm_labels)[0]
```
## Resources
- [Text classification task guide](../tasks/sequence_classification)
- [Token classification task guide](../tasks/token_classification)
- [Question answering task guide](../tasks/question_answering)
- [Masked language modeling task guide](../tasks/masked_language_modeling)
- [Multiple choice task guide](../tasks/multiple_choice)
## LongformerConfig
[[autodoc]] LongformerConfig
## LongformerTokenizer
[[autodoc]] LongformerTokenizer
## LongformerTokenizerFast
[[autodoc]] LongformerTokenizerFast
## Longformer specific outputs
[[autodoc]] models.longformer.modeling_longformer.LongformerBaseModelOutput
[[autodoc]] models.longformer.modeling_longformer.LongformerBaseModelOutputWithPooling
[[autodoc]] models.longformer.modeling_longformer.LongformerMaskedLMOutput
[[autodoc]] models.longformer.modeling_longformer.LongformerQuestionAnsweringModelOutput
[[autodoc]] models.longformer.modeling_longformer.LongformerSequenceClassifierOutput
[[autodoc]] models.longformer.modeling_longformer.LongformerMultipleChoiceModelOutput
[[autodoc]] models.longformer.modeling_longformer.LongformerTokenClassifierOutput
[[autodoc]] models.longformer.modeling_tf_longformer.TFLongformerBaseModelOutput
[[autodoc]] models.longformer.modeling_tf_longformer.TFLongformerBaseModelOutputWithPooling
[[autodoc]] models.longformer.modeling_tf_longformer.TFLongformerMaskedLMOutput
[[autodoc]] models.longformer.modeling_tf_longformer.TFLongformerQuestionAnsweringModelOutput
[[autodoc]] models.longformer.modeling_tf_longformer.TFLongformerSequenceClassifierOutput
[[autodoc]] models.longformer.modeling_tf_longformer.TFLongformerMultipleChoiceModelOutput
[[autodoc]] models.longformer.modeling_tf_longformer.TFLongformerTokenClassifierOutput
<frameworkcontent>
<pt>
## LongformerModel
[[autodoc]] LongformerModel
- forward
## LongformerForMaskedLM
[[autodoc]] LongformerForMaskedLM
- forward
## LongformerForSequenceClassification
[[autodoc]] LongformerForSequenceClassification
- forward
## LongformerForMultipleChoice
[[autodoc]] LongformerForMultipleChoice
- forward
## LongformerForTokenClassification
[[autodoc]] LongformerForTokenClassification
- forward
## LongformerForQuestionAnswering
[[autodoc]] LongformerForQuestionAnswering
- forward
</pt>
<tf>
## TFLongformerModel
[[autodoc]] TFLongformerModel
- call
## TFLongformerForMaskedLM
[[autodoc]] TFLongformerForMaskedLM
- call
## TFLongformerForQuestionAnswering
[[autodoc]] TFLongformerForQuestionAnswering
- call
## TFLongformerForSequenceClassification
[[autodoc]] TFLongformerForSequenceClassification
- call
## TFLongformerForTokenClassification
[[autodoc]] TFLongformerForTokenClassification
- call
## TFLongformerForMultipleChoice
[[autodoc]] TFLongformerForMultipleChoice
- call
</tf>
</frameworkcontent>
| huggingface/transformers/blob/main/docs/source/en/model_doc/longformer.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Utilities for Trainer
This page lists all the utility functions used by [`Trainer`].
Most of those are only useful if you are studying the code of the Trainer in the library.
## Utilities
[[autodoc]] EvalPrediction
[[autodoc]] IntervalStrategy
[[autodoc]] enable_full_determinism
[[autodoc]] set_seed
[[autodoc]] torch_distributed_zero_first
## Callbacks internals
[[autodoc]] trainer_callback.CallbackHandler
## Distributed Evaluation
[[autodoc]] trainer_pt_utils.DistributedTensorGatherer
## Trainer Argument Parser
[[autodoc]] HfArgumentParser
## Debug Utilities
[[autodoc]] debug_utils.DebugUnderflowOverflow
| huggingface/transformers/blob/main/docs/source/en/internal/trainer_utils.md |
!---
Copyright 2021 The Google Flax Team Authors and HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-->
# Question Answering examples
Based on the script [`run_qa.py`](https://github.com/huggingface/transformers/blob/main/examples/flax/question-answering/run_qa.py).
**Note:** This script only works with models that have a fast tokenizer (backed by the 🤗 Tokenizers library) as it
uses special features of those tokenizers. You can check if your favorite model has a fast tokenizer in
[this table](https://huggingface.co/transformers/index.html#supported-frameworks), if it doesn't you can still use the old version
of the script.
The following example fine-tunes BERT on SQuAD:
```bash
python run_qa.py \
--model_name_or_path bert-base-uncased \
--dataset_name squad \
--do_train \
--do_eval \
--max_seq_length 384 \
--doc_stride 128 \
--learning_rate 3e-5 \
--num_train_epochs 2 \
--per_device_train_batch_size 12 \
--output_dir ./bert-qa-squad \
--eval_steps 1000 \
--push_to_hub
```
Using the command above, the script will train for 2 epochs and run eval after each epoch.
Metrics and hyperparameters are stored in Tensorflow event files in `--output_dir`.
You can see the results by running `tensorboard` in that directory:
```bash
$ tensorboard --logdir .
```
or directly on the hub under *Training metrics*.
Training with the previously defined hyper-parameters yields the following results:
```bash
f1 = 88.62
exact_match = 81.34
```
sample Metrics - [tfhub.dev](https://tensorboard.dev/experiment/6gU75Hx8TGCnc6tr4ZgI9Q)
Here is an example training on 4 TITAN RTX GPUs and Bert Whole Word Masking uncased model to reach a F1 > 93 on SQuAD1.1:
```bash
export CUDA_VISIBLE_DEVICES=0,1,2,3
python run_qa.py \
--model_name_or_path bert-large-uncased-whole-word-masking \
--dataset_name squad \
--do_train \
--do_eval \
--per_device_train_batch_size 6 \
--learning_rate 3e-5 \
--num_train_epochs 2 \
--max_seq_length 384 \
--doc_stride 128 \
--output_dir ./wwm_uncased_finetuned_squad/ \
--eval_steps 1000 \
--push_to_hub
```
Training with the previously defined hyper-parameters yields the following results:
```bash
f1 = 93.31
exact_match = 87.04
```
### Usage notes
Note that when contexts are long they may be split into multiple training cases, not all of which may contain
the answer span.
As-is, the example script will train on SQuAD or any other question-answering dataset formatted the same way, and can handle user
inputs as well.
### Memory usage and data loading
One thing to note is that all data is loaded into memory in this script. Most question answering datasets are small
enough that this is not an issue, but if you have a very large dataset you will need to modify the script to handle
data streaming.
| huggingface/transformers/blob/main/examples/flax/question-answering/README.md |
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# TVLT
## Overview
The TVLT model was proposed in [TVLT: Textless Vision-Language Transformer](https://arxiv.org/abs/2209.14156)
by Zineng Tang, Jaemin Cho, Yixin Nie, Mohit Bansal (the first three authors contributed equally). The Textless Vision-Language Transformer (TVLT) is a model that uses raw visual and audio inputs for vision-and-language representation learning, without using text-specific modules such as tokenization or automatic speech recognition (ASR). It can perform various audiovisual and vision-language tasks like retrieval, question answering, etc.
The abstract from the paper is the following:
*In this work, we present the Textless Vision-Language Transformer (TVLT), where homogeneous transformer blocks take raw visual and audio inputs for vision-and-language representation learning with minimal modality-specific design, and do not use text-specific modules such as tokenization or automatic speech recognition (ASR). TVLT is trained by reconstructing masked patches of continuous video frames and audio spectrograms (masked autoencoding) and contrastive modeling to align video and audio. TVLT attains performance comparable to its text-based counterpart on various multimodal tasks, such as visual question answering, image retrieval, video retrieval, and multimodal sentiment analysis, with 28x faster inference speed and only 1/3 of the parameters. Our findings suggest the possibility of learning compact and efficient visual-linguistic representations from low-level visual and audio signals without assuming the prior existence of text.*
<p align="center">
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/tvlt_architecture.png"
alt="drawing" width="600"/>
</p>
<small> TVLT architecture. Taken from the <a href="[https://arxiv.org/abs/2102.03334](https://arxiv.org/abs/2209.14156)">original paper</a>. </small>
The original code can be found [here](https://github.com/zinengtang/TVLT). This model was contributed by [Zineng Tang](https://huggingface.co/ZinengTang).
## Usage tips
- TVLT is a model that takes both `pixel_values` and `audio_values` as input. One can use [`TvltProcessor`] to prepare data for the model.
This processor wraps an image processor (for the image/video modality) and an audio feature extractor (for the audio modality) into one.
- TVLT is trained with images/videos and audios of various sizes: the authors resize and crop the input images/videos to 224 and limit the length of audio spectrogram to 2048. To make batching of videos and audios possible, the authors use a `pixel_mask` that indicates which pixels are real/padding and `audio_mask` that indicates which audio values are real/padding.
- The design of TVLT is very similar to that of a standard Vision Transformer (ViT) and masked autoencoder (MAE) as in [ViTMAE](vitmae). The difference is that the model includes embedding layers for the audio modality.
- The PyTorch version of this model is only available in torch 1.10 and higher.
## TvltConfig
[[autodoc]] TvltConfig
## TvltProcessor
[[autodoc]] TvltProcessor
- __call__
## TvltImageProcessor
[[autodoc]] TvltImageProcessor
- preprocess
## TvltFeatureExtractor
[[autodoc]] TvltFeatureExtractor
- __call__
## TvltModel
[[autodoc]] TvltModel
- forward
## TvltForPreTraining
[[autodoc]] TvltForPreTraining
- forward
## TvltForAudioVisualClassification
[[autodoc]] TvltForAudioVisualClassification
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/tvlt.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# DistilBERT
<div class="flex flex-wrap space-x-1">
<a href="https://huggingface.co/models?filter=distilbert">
<img alt="Models" src="https://img.shields.io/badge/All_model_pages-distilbert-blueviolet">
</a>
<a href="https://huggingface.co/spaces/docs-demos/distilbert-base-uncased">
<img alt="Spaces" src="https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-blue">
</a>
<a href="https://huggingface.co/papers/1910.01108">
<img alt="Paper page" src="https://img.shields.io/badge/Paper%20page-1910.01108-green">
</a>
</div>
## Overview
The DistilBERT model was proposed in the blog post [Smaller, faster, cheaper, lighter: Introducing DistilBERT, a
distilled version of BERT](https://medium.com/huggingface/distilbert-8cf3380435b5), and the paper [DistilBERT, a
distilled version of BERT: smaller, faster, cheaper and lighter](https://arxiv.org/abs/1910.01108). DistilBERT is a
small, fast, cheap and light Transformer model trained by distilling BERT base. It has 40% less parameters than
*bert-base-uncased*, runs 60% faster while preserving over 95% of BERT's performances as measured on the GLUE language
understanding benchmark.
The abstract from the paper is the following:
*As Transfer Learning from large-scale pre-trained models becomes more prevalent in Natural Language Processing (NLP),
operating these large models in on-the-edge and/or under constrained computational training or inference budgets
remains challenging. In this work, we propose a method to pre-train a smaller general-purpose language representation
model, called DistilBERT, which can then be fine-tuned with good performances on a wide range of tasks like its larger
counterparts. While most prior work investigated the use of distillation for building task-specific models, we leverage
knowledge distillation during the pretraining phase and show that it is possible to reduce the size of a BERT model by
40%, while retaining 97% of its language understanding capabilities and being 60% faster. To leverage the inductive
biases learned by larger models during pretraining, we introduce a triple loss combining language modeling,
distillation and cosine-distance losses. Our smaller, faster and lighter model is cheaper to pre-train and we
demonstrate its capabilities for on-device computations in a proof-of-concept experiment and a comparative on-device
study.*
This model was contributed by [victorsanh](https://huggingface.co/victorsanh). This model jax version was
contributed by [kamalkraj](https://huggingface.co/kamalkraj). The original code can be found [here](https://github.com/huggingface/transformers/tree/main/examples/research_projects/distillation).
## Usage tips
- DistilBERT doesn't have `token_type_ids`, you don't need to indicate which token belongs to which segment. Just
separate your segments with the separation token `tokenizer.sep_token` (or `[SEP]`).
- DistilBERT doesn't have options to select the input positions (`position_ids` input). This could be added if
necessary though, just let us know if you need this option.
- Same as BERT but smaller. Trained by distillation of the pretrained BERT model, meaning it’s been trained to predict the same probabilities as the larger model. The actual objective is a combination of:
* finding the same probabilities as the teacher model
* predicting the masked tokens correctly (but no next-sentence objective)
* a cosine similarity between the hidden states of the student and the teacher model
## Resources
A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with DistilBERT. If you're interested in submitting a resource to be included here, please feel free to open a Pull Request and we'll review it! The resource should ideally demonstrate something new instead of duplicating an existing resource.
<PipelineTag pipeline="text-classification"/>
- A blog post on [Getting Started with Sentiment Analysis using Python](https://huggingface.co/blog/sentiment-analysis-python) with DistilBERT.
- A blog post on how to [train DistilBERT with Blurr for sequence classification](https://huggingface.co/blog/fastai).
- A blog post on how to use [Ray to tune DistilBERT hyperparameters](https://huggingface.co/blog/ray-tune).
- A blog post on how to [train DistilBERT with Hugging Face and Amazon SageMaker](https://huggingface.co/blog/the-partnership-amazon-sagemaker-and-hugging-face).
- A notebook on how to [finetune DistilBERT for multi-label classification](https://colab.research.google.com/github/DhavalTaunk08/Transformers_scripts/blob/master/Transformers_multilabel_distilbert.ipynb). 🌎
- A notebook on how to [finetune DistilBERT for multiclass classification with PyTorch](https://colab.research.google.com/github/abhimishra91/transformers-tutorials/blob/master/transformers_multiclass_classification.ipynb). 🌎
- A notebook on how to [finetune DistilBERT for text classification in TensorFlow](https://colab.research.google.com/github/peterbayerle/huggingface_notebook/blob/main/distilbert_tf.ipynb). 🌎
- [`DistilBertForSequenceClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/text-classification) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/text_classification.ipynb).
- [`TFDistilBertForSequenceClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/tensorflow/text-classification) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/text_classification-tf.ipynb).
- [`FlaxDistilBertForSequenceClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/flax/text-classification) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/text_classification_flax.ipynb).
- [Text classification task guide](../tasks/sequence_classification)
<PipelineTag pipeline="token-classification"/>
- [`DistilBertForTokenClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/token-classification) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/token_classification.ipynb).
- [`TFDistilBertForTokenClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/tensorflow/token-classification) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/token_classification-tf.ipynb).
- [`FlaxDistilBertForTokenClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/flax/token-classification).
- [Token classification](https://huggingface.co/course/chapter7/2?fw=pt) chapter of the 🤗 Hugging Face Course.
- [Token classification task guide](../tasks/token_classification)
<PipelineTag pipeline="fill-mask"/>
- [`DistilBertForMaskedLM`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/language-modeling#robertabertdistilbert-and-masked-language-modeling) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/language_modeling.ipynb).
- [`TFDistilBertForMaskedLM`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/tensorflow/language-modeling#run_mlmpy) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/language_modeling-tf.ipynb).
- [`FlaxDistilBertForMaskedLM`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/flax/language-modeling#masked-language-modeling) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/masked_language_modeling_flax.ipynb).
- [Masked language modeling](https://huggingface.co/course/chapter7/3?fw=pt) chapter of the 🤗 Hugging Face Course.
- [Masked language modeling task guide](../tasks/masked_language_modeling)
<PipelineTag pipeline="question-answering"/>
- [`DistilBertForQuestionAnswering`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/question-answering) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/question_answering.ipynb).
- [`TFDistilBertForQuestionAnswering`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/tensorflow/question-answering) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/question_answering-tf.ipynb).
- [`FlaxDistilBertForQuestionAnswering`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/flax/question-answering).
- [Question answering](https://huggingface.co/course/chapter7/7?fw=pt) chapter of the 🤗 Hugging Face Course.
- [Question answering task guide](../tasks/question_answering)
**Multiple choice**
- [`DistilBertForMultipleChoice`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/multiple-choice) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/multiple_choice.ipynb).
- [`TFDistilBertForMultipleChoice`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/tensorflow/multiple-choice) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/multiple_choice-tf.ipynb).
- [Multiple choice task guide](../tasks/multiple_choice)
⚗️ Optimization
- A blog post on how to [quantize DistilBERT with 🤗 Optimum and Intel](https://huggingface.co/blog/intel).
- A blog post on how [Optimizing Transformers for GPUs with 🤗 Optimum](https://www.philschmid.de/optimizing-transformers-with-optimum-gpu).
- A blog post on [Optimizing Transformers with Hugging Face Optimum](https://www.philschmid.de/optimizing-transformers-with-optimum).
⚡️ Inference
- A blog post on how to [Accelerate BERT inference with Hugging Face Transformers and AWS Inferentia](https://huggingface.co/blog/bert-inferentia-sagemaker) with DistilBERT.
- A blog post on [Serverless Inference with Hugging Face's Transformers, DistilBERT and Amazon SageMaker](https://www.philschmid.de/sagemaker-serverless-huggingface-distilbert).
🚀 Deploy
- A blog post on how to [deploy DistilBERT on Google Cloud](https://huggingface.co/blog/how-to-deploy-a-pipeline-to-google-clouds).
- A blog post on how to [deploy DistilBERT with Amazon SageMaker](https://huggingface.co/blog/deploy-hugging-face-models-easily-with-amazon-sagemaker).
- A blog post on how to [Deploy BERT with Hugging Face Transformers, Amazon SageMaker and Terraform module](https://www.philschmid.de/terraform-huggingface-amazon-sagemaker).
## Combining DistilBERT and Flash Attention 2
First, make sure to install the latest version of Flash Attention 2 to include the sliding window attention feature.
```bash
pip install -U flash-attn --no-build-isolation
```
Make also sure that you have a hardware that is compatible with Flash-Attention 2. Read more about it in the official documentation of flash-attn repository. Make also sure to load your model in half-precision (e.g. `torch.float16`)
To load and run a model using Flash Attention 2, refer to the snippet below:
```python
>>> import torch
>>> from transformers import AutoTokenizer, AutoModel
>>> device = "cuda" # the device to load the model onto
>>> tokenizer = AutoTokenizer.from_pretrained('distilbert-base-uncased')
>>> model = AutoModel.from_pretrained("distilbert-base-uncased", torch_dtype=torch.float16, attn_implementation="flash_attention_2")
>>> text = "Replace me by any text you'd like."
>>> encoded_input = tokenizer(text, return_tensors='pt').to(device)
>>> model.to(device)
>>> output = model(**encoded_input)
```
## DistilBertConfig
[[autodoc]] DistilBertConfig
## DistilBertTokenizer
[[autodoc]] DistilBertTokenizer
## DistilBertTokenizerFast
[[autodoc]] DistilBertTokenizerFast
<frameworkcontent>
<pt>
## DistilBertModel
[[autodoc]] DistilBertModel
- forward
## DistilBertForMaskedLM
[[autodoc]] DistilBertForMaskedLM
- forward
## DistilBertForSequenceClassification
[[autodoc]] DistilBertForSequenceClassification
- forward
## DistilBertForMultipleChoice
[[autodoc]] DistilBertForMultipleChoice
- forward
## DistilBertForTokenClassification
[[autodoc]] DistilBertForTokenClassification
- forward
## DistilBertForQuestionAnswering
[[autodoc]] DistilBertForQuestionAnswering
- forward
</pt>
<tf>
## TFDistilBertModel
[[autodoc]] TFDistilBertModel
- call
## TFDistilBertForMaskedLM
[[autodoc]] TFDistilBertForMaskedLM
- call
## TFDistilBertForSequenceClassification
[[autodoc]] TFDistilBertForSequenceClassification
- call
## TFDistilBertForMultipleChoice
[[autodoc]] TFDistilBertForMultipleChoice
- call
## TFDistilBertForTokenClassification
[[autodoc]] TFDistilBertForTokenClassification
- call
## TFDistilBertForQuestionAnswering
[[autodoc]] TFDistilBertForQuestionAnswering
- call
</tf>
<jax>
## FlaxDistilBertModel
[[autodoc]] FlaxDistilBertModel
- __call__
## FlaxDistilBertForMaskedLM
[[autodoc]] FlaxDistilBertForMaskedLM
- __call__
## FlaxDistilBertForSequenceClassification
[[autodoc]] FlaxDistilBertForSequenceClassification
- __call__
## FlaxDistilBertForMultipleChoice
[[autodoc]] FlaxDistilBertForMultipleChoice
- __call__
## FlaxDistilBertForTokenClassification
[[autodoc]] FlaxDistilBertForTokenClassification
- __call__
## FlaxDistilBertForQuestionAnswering
[[autodoc]] FlaxDistilBertForQuestionAnswering
- __call__
</jax>
</frameworkcontent>
| huggingface/transformers/blob/main/docs/source/en/model_doc/distilbert.md |
!---
Copyright 2021 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
-->
# Summarization example
This script shows an example of training a *summarization* model with the 🤗 Transformers library.
For straightforward use-cases you may be able to use these scripts without modification, although we have also
included comments in the code to indicate areas that you may need to adapt to your own projects.
### Multi-GPU and TPU usage
By default, these scripts use a `MirroredStrategy` and will use multiple GPUs effectively if they are available. TPUs
can also be used by passing the name of the TPU resource with the `--tpu` argument.
### Example command
```
python run_summarization.py \
--model_name_or_path facebook/bart-base \
--dataset_name cnn_dailymail \
--dataset_config "3.0.0" \
--output_dir /tmp/tst-summarization \
--per_device_train_batch_size 8 \
--per_device_eval_batch_size 16 \
--num_train_epochs 3 \
--do_train \
--do_eval
``` | huggingface/transformers/blob/main/examples/tensorflow/summarization/README.md |
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Optimizing LLMs for Speed and Memory
[[open-in-colab]]
Large Language Models (LLMs) such as GPT3/4, [Falcon](https://huggingface.co/tiiuae/falcon-40b), and [Llama](https://huggingface.co/meta-llama/Llama-2-70b-hf) are rapidly advancing in their ability to tackle human-centric tasks, establishing themselves as essential tools in modern knowledge-based industries.
Deploying these models in real-world tasks remains challenging, however:
- To exhibit near-human text understanding and generation capabilities, LLMs currently require to be composed of billions of parameters (see [Kaplan et al](https://arxiv.org/abs/2001.08361), [Wei et. al](https://arxiv.org/abs/2206.07682)). This consequently amplifies the memory demands for inference.
- In many real-world tasks, LLMs need to be given extensive contextual information. This necessitates the model's capability to manage very long input sequences during inference.
The crux of these challenges lies in augmenting the computational and memory capabilities of LLMs, especially when handling expansive input sequences.
In this guide, we will go over the effective techniques for efficient LLM deployment:
1. **Lower Precision:** Research has shown that operating at reduced numerical precision, namely [8-bit and 4-bit](./main_classes/quantization.md) can achieve computational advantages without a considerable decline in model performance.
2. **Flash Attention:** Flash Attention is a variation of the attention algorithm that not only provides a more memory-efficient approach but also realizes increased efficiency due to optimized GPU memory utilization.
3. **Architectural Innovations:** Considering that LLMs are always deployed in the same way during inference, namely autoregressive text generation with a long input context, specialized model architectures have been proposed that allow for more efficient inference. The most important advancement in model architectures hereby are [Alibi](https://arxiv.org/abs/2108.12409), [Rotary embeddings](https://arxiv.org/abs/2104.09864), [Multi-Query Attention (MQA)](https://arxiv.org/abs/1911.02150) and [Grouped-Query-Attention (GQA)]((https://arxiv.org/abs/2305.13245)).
Throughout this guide, we will offer an analysis of auto-regressive generation from a tensor's perspective. We delve into the pros and cons of adopting lower precision, provide a comprehensive exploration of the latest attention algorithms, and discuss improved LLM architectures. While doing so, we run practical examples showcasing each of the feature improvements.
## 1. Lower Precision
Memory requirements of LLMs can be best understood by seeing the LLM as a set of weight matrices and vectors and the text inputs as a sequence of vectors. In the following, the definition *weights* will be used to signify all model weight matrices and vectors.
At the time of writing this guide, LLMs consist of at least a couple billion parameters. Each parameter thereby is made of a decimal number, e.g. `4.5689` which is usually stored in either [float32](https://en.wikipedia.org/wiki/Single-precision_floating-point_format), [bfloat16](https://en.wikipedia.org/wiki/Bfloat16_floating-point_format), or [float16](https://en.wikipedia.org/wiki/Half-precision_floating-point_format) format. This allows us to easily compute the memory requirement to load the LLM into memory:
> *Loading the weights of a model having X billion parameters requires roughly 4 * X GB of VRAM in float32 precision*
Nowadays, models are however rarely trained in full float32 precision, but usually in bfloat16 precision or less frequently in float16 precision. Therefore the rule of thumb becomes:
> *Loading the weights of a model having X billion parameters requires roughly 2 * X GB of VRAM in bfloat16/float16 precision*
For shorter text inputs (less than 1024 tokens), the memory requirement for inference is very much dominated by the memory requirement to load the weights. Therefore, for now, let's assume that the memory requirement for inference is equal to the memory requirement to load the model into the GPU VRAM.
To give some examples of how much VRAM it roughly takes to load a model in bfloat16:
- **GPT3** requires 2 \* 175 GB = **350 GB** VRAM
- [**Bloom**](https://huggingface.co/bigscience/bloom) requires 2 \* 176 GB = **352 GB** VRAM
- [**Llama-2-70b**](https://huggingface.co/meta-llama/Llama-2-70b-hf) requires 2 \* 70 GB = **140 GB** VRAM
- [**Falcon-40b**](https://huggingface.co/tiiuae/falcon-40b) requires 2 \* 40 GB = **80 GB** VRAM
- [**MPT-30b**](https://huggingface.co/mosaicml/mpt-30b) requires 2 \* 30 GB = **60 GB** VRAM
- [**bigcode/starcoder**](https://huggingface.co/bigcode/starcoder) requires 2 \* 15.5 = **31 GB** VRAM
As of writing this document, the largest GPU chip on the market is the A100 & H100 offering 80GB of VRAM. Most of the models listed before require more than 80GB just to be loaded and therefore necessarily require [tensor parallelism](https://huggingface.co/docs/transformers/perf_train_gpu_many#tensor-parallelism) and/or [pipeline parallelism](https://huggingface.co/docs/transformers/perf_train_gpu_many#naive-model-parallelism-vertical-and-pipeline-parallelism).
🤗 Transformers does not support tensor parallelism out of the box as it requires the model architecture to be written in a specific way. If you're interested in writing models in a tensor-parallelism-friendly way, feel free to have a look at [the text-generation-inference library](https://github.com/huggingface/text-generation-inference/tree/main/server/text_generation_server/models/custom_modeling).
Naive pipeline parallelism is supported out of the box. For this, simply load the model with `device="auto"` which will automatically place the different layers on the available GPUs as explained [here](https://huggingface.co/docs/accelerate/v0.22.0/en/concept_guides/big_model_inference).
Note, however that while very effective, this naive pipeline parallelism does not tackle the issues of GPU idling. For this more advanced pipeline parallelism is required as explained [here](https://huggingface.co/docs/transformers/en/perf_train_gpu_many#naive-model-parallelism-vertical-and-pipeline-parallelism).
If you have access to an 8 x 80GB A100 node, you could load BLOOM as follows
```bash
!pip install transformers accelerate bitsandbytes optimum
```
```python
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("bigscience/bloom", device_map="auto", pad_token_id=0)
```
By using `device_map="auto"` the attention layers would be equally distributed over all available GPUs.
In this guide, we will use [bigcode/octocoder](https://huggingface.co/bigcode/octocoder) as it can be run on a single 40 GB A100 GPU device chip. Note that all memory and speed optimizations that we will apply going forward, are equally applicable to models that require model or tensor parallelism.
Since the model is loaded in bfloat16 precision, using our rule of thumb above, we would expect the memory requirement to run inference with `bigcode/octocoder` to be around 31 GB VRAM. Let's give it a try.
We first load the model and tokenizer and then pass both to Transformers' [pipeline](https://huggingface.co/docs/transformers/main_classes/pipelines) object.
```python
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
import torch
model = AutoModelForCausalLM.from_pretrained("bigcode/octocoder", torch_dtype=torch.bfloat16, device_map="auto", pad_token_id=0)
tokenizer = AutoTokenizer.from_pretrained("bigcode/octocoder")
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
```
```python
prompt = "Question: Please write a function in Python that transforms bytes to Giga bytes.\n\nAnswer:"
result = pipe(prompt, max_new_tokens=60)[0]["generated_text"][len(prompt):]
result
```
**Output**:
```
Here is a Python function that transforms bytes to Giga bytes:\n\n```python\ndef bytes_to_giga_bytes(bytes):\n return bytes / 1024 / 1024 / 1024\n```\n\nThis function takes a single
```
Nice, we can now directly use the result to convert bytes into Gigabytes.
```python
def bytes_to_giga_bytes(bytes):
return bytes / 1024 / 1024 / 1024
```
Let's call [`torch.cuda.max_memory_allocated`](https://pytorch.org/docs/stable/generated/torch.cuda.max_memory_allocated.html) to measure the peak GPU memory allocation.
```python
bytes_to_giga_bytes(torch.cuda.max_memory_allocated())
```
**Output**:
```bash
29.0260648727417
```
Close enough to our back-of-the-envelope computation! We can see the number is not exactly correct as going from bytes to kilobytes requires a multiplication of 1024 instead of 1000. Therefore the back-of-the-envelope formula can also be understood as an "at most X GB" computation.
Note that if we had tried to run the model in full float32 precision, a whopping 64 GB of VRAM would have been required.
> Almost all models are trained in bfloat16 nowadays, there is no reason to run the model in full float32 precision if [your GPU supports bfloat16](https://discuss.pytorch.org/t/bfloat16-native-support/117155/5). Float32 won't give better inference results than the precision that was used to train the model.
If you are unsure in which format the model weights are stored on the Hub, you can always look into the checkpoint's config under `"torch_dtype"`, *e.g.* [here](https://huggingface.co/meta-llama/Llama-2-7b-hf/blob/6fdf2e60f86ff2481f2241aaee459f85b5b0bbb9/config.json#L21). It is recommended to set the model to the same precision type as written in the config when loading with `from_pretrained(..., torch_dtype=...)` except when the original type is float32 in which case one can use both `float16` or `bfloat16` for inference.
Let's define a `flush(...)` function to free all allocated memory so that we can accurately measure the peak allocated GPU memory.
```python
del pipe
del model
import gc
import torch
def flush():
gc.collect()
torch.cuda.empty_cache()
torch.cuda.reset_peak_memory_stats()
```
Let's call it now for the next experiment.
```python
flush()
```
In the recent version of the accelerate library, you can also use an utility method called `release_memory()`
```python
from accelerate.utils import release_memory
# ...
release_memory(model)
```
Now what if your GPU does not have 32 GB of VRAM? It has been found that model weights can be quantized to 8-bit or 4-bits without a significant loss in performance (see [Dettmers et al.](https://arxiv.org/abs/2208.07339)).
Model can be quantized to even 3 or 2 bits with an acceptable loss in performance as shown in the recent [GPTQ paper](https://arxiv.org/abs/2210.17323) 🤯.
Without going into too many details, quantization schemes aim at reducing the precision of weights while trying to keep the model's inference results as accurate as possible (*a.k.a* as close as possible to bfloat16).
Note that quantization works especially well for text generation since all we care about is choosing the *set of most likely next tokens* and don't really care about the exact values of the next token *logit* distribution.
All that matters is that the next token *logit* distribution stays roughly the same so that an `argmax` or `topk` operation gives the same results.
There are various quantization techniques, which we won't discuss in detail here, but in general, all quantization techniques work as follows:
- 1. Quantize all weights to the target precision
- 2. Load the quantized weights, and pass the input sequence of vectors in bfloat16 precision
- 3. Dynamically dequantize weights to bfloat16 to perform the computation with their input vectors in bfloat16 precision
In a nutshell, this means that *inputs-weight matrix* multiplications, with \\( X \\) being the *inputs*, \\( W \\) being a weight matrix and \\( Y \\) being the output:
$$ Y = X * W $$
are changed to
$$ Y = X * \text{dequantize}(W) $$
for every matrix multiplication. Dequantization and re-quantization is performed sequentially for all weight matrices as the inputs run through the network graph.
Therefore, inference time is often **not** reduced when using quantized weights, but rather increases.
Enough theory, let's give it a try! To quantize the weights with Transformers, you need to make sure that
the [`bitsandbytes`](https://github.com/TimDettmers/bitsandbytes) library is installed.
```bash
!pip install bitsandbytes
```
We can then load models in 8-bit quantization by simply adding a `load_in_8bit=True` flag to `from_pretrained`.
```python
model = AutoModelForCausalLM.from_pretrained("bigcode/octocoder", load_in_8bit=True, pad_token_id=0)
```
Now, let's run our example again and measure the memory usage.
```python
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
result = pipe(prompt, max_new_tokens=60)[0]["generated_text"][len(prompt):]
result
```
**Output**:
```
Here is a Python function that transforms bytes to Giga bytes:\n\n```python\ndef bytes_to_giga_bytes(bytes):\n return bytes / 1024 / 1024 / 1024\n```\n\nThis function takes a single
```
Nice, we're getting the same result as before, so no loss in accuracy! Let's look at how much memory was used this time.
```python
bytes_to_giga_bytes(torch.cuda.max_memory_allocated())
```
**Output**:
```
15.219234466552734
```
Significantly less! We're down to just a bit over 15 GBs and could therefore run this model on consumer GPUs like the 4090.
We're seeing a very nice gain in memory efficiency and more or less no degradation to the model's output. However, we can also notice a slight slow-down during inference.
We delete the models and flush the memory again.
```python
del model
del pipe
```
```python
flush()
```
Let's see what peak GPU memory consumption 4-bit quantization gives. Quantizing the model to 4-bit can be done with the same API as before - this time by passing `load_in_4bit=True` instead of `load_in_8bit=True`.
```python
model = AutoModelForCausalLM.from_pretrained("bigcode/octocoder", load_in_4bit=True, low_cpu_mem_usage=True, pad_token_id=0)
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
result = pipe(prompt, max_new_tokens=60)[0]["generated_text"][len(prompt):]
result
```
**Output**:
```
Here is a Python function that transforms bytes to Giga bytes:\n\n```\ndef bytes_to_gigabytes(bytes):\n return bytes / 1024 / 1024 / 1024\n```\n\nThis function takes a single argument
```
We're almost seeing the same output text as before - just the `python` is missing just before the code snippet. Let's see how much memory was required.
```python
bytes_to_giga_bytes(torch.cuda.max_memory_allocated())
```
**Output**:
```
9.543574333190918
```
Just 9.5GB! That's really not a lot for a >15 billion parameter model.
While we see very little degradation in accuracy for our model here, 4-bit quantization can in practice often lead to different results compared to 8-bit quantization or full `bfloat16` inference. It is up to the user to try it out.
Also note that inference here was again a bit slower compared to 8-bit quantization which is due to the more aggressive quantization method used for 4-bit quantization leading to \\( \text{quantize} \\) and \\( \text{dequantize} \\) taking longer during inference.
```python
del model
del pipe
```
```python
flush()
```
Overall, we saw that running OctoCoder in 8-bit precision reduced the required GPU VRAM from 32G GPU VRAM to only 15GB and running the model in 4-bit precision further reduces the required GPU VRAM to just a bit over 9GB.
4-bit quantization allows the model to be run on GPUs such as RTX3090, V100, and T4 which are quite accessible for most people.
For more information on quantization and to see how one can quantize models to require even less GPU VRAM memory than 4-bit, we recommend looking into the [`AutoGPTQ`](https://huggingface.co/docs/transformers/main/en/main_classes/quantization#autogptq-integration%60) implementation.
> As a conclusion, it is important to remember that model quantization trades improved memory efficiency against accuracy and in some cases inference time.
If GPU memory is not a constraint for your use case, there is often no need to look into quantization. However many GPUs simply can't run LLMs without quantization methods and in this case, 4-bit and 8-bit quantization schemes are extremely useful tools.
For more in-detail usage information, we strongly recommend taking a look at the [Transformers Quantization Docs](https://huggingface.co/docs/transformers/main_classes/quantization#general-usage).
Next, let's look into how we can improve computational and memory efficiency by using better algorithms and an improved model architecture.
## 2. Flash Attention
Today's top-performing LLMs share more or less the same fundamental architecture that consists of feed-forward layers, activation layers, layer normalization layers, and most crucially, self-attention layers.
Self-attention layers are central to Large Language Models (LLMs) in that they enable the model to understand the contextual relationships between input tokens.
However, the peak GPU memory consumption for self-attention layers grows *quadratically* both in compute and memory complexity with number of input tokens (also called *sequence length*) that we denote in the following by \\( N \\) .
While this is not really noticeable for shorter input sequences (of up to 1000 input tokens), it becomes a serious problem for longer input sequences (at around 16000 input tokens).
Let's take a closer look. The formula to compute the output \\( \mathbf{O} \\) of a self-attention layer for an input \\( \mathbf{X} \\) of length \\( N \\) is:
$$ \textbf{O} = \text{Attn}(\mathbf{X}) = \mathbf{V} \times \text{Softmax}(\mathbf{QK}^T) \text{ with } \mathbf{Q} = \mathbf{W}_q \mathbf{X}, \mathbf{V} = \mathbf{W}_v \mathbf{X}, \mathbf{K} = \mathbf{W}_k \mathbf{X} $$
\\( \mathbf{X} = (\mathbf{x}_1, ... \mathbf{x}_{N}) \\) is thereby the input sequence to the attention layer. The projections \\( \mathbf{Q} \\) and \\( \mathbf{K} \\) will each consist of \\( N \\) vectors resulting in the \\( \mathbf{QK}^T \\) being of size \\( N^2 \\) .
LLMs usually have multiple attention heads, thus doing multiple self-attention computations in parallel.
Assuming, the LLM has 40 attention heads and runs in bfloat16 precision, we can calculate the memory requirement to store the \\( \mathbf{QK^T} \\) matrices to be \\( 40 * 2 * N^2 \\) bytes. For \\( N=1000 \\) only around 50 MB of VRAM are needed, however, for \\( N=16000 \\) we would need 19 GB of VRAM, and for \\( N=100,000 \\) we would need almost 1TB just to store the \\( \mathbf{QK}^T \\) matrices.
Long story short, the default self-attention algorithm quickly becomes prohibitively memory-expensive for large input contexts.
As LLMs improve in text comprehension and generation, they are applied to increasingly complex tasks. While models once handled the translation or summarization of a few sentences, they now manage entire pages, demanding the capability to process extensive input lengths.
How can we get rid of the exorbitant memory requirements for large input lengths? We need a new way to compute the self-attention mechanism that gets rid of the \\( QK^T \\) matrix. [Tri Dao et al.](https://arxiv.org/abs/2205.14135) developed exactly such a new algorithm and called it **Flash Attention**.
In a nutshell, Flash Attention breaks the \\(\mathbf{V} \times \text{Softmax}(\mathbf{QK}^T\\)) computation apart and instead computes smaller chunks of the output by iterating over multiple softmax computation steps:
$$ \textbf{O}_i \leftarrow s^a_{ij} * \textbf{O}_i + s^b_{ij} * \mathbf{V}_{j} \times \text{Softmax}(\mathbf{QK}^T_{i,j}) \text{ for multiple } i, j \text{ iterations} $$
with \\( s^a_{ij} \\) and \\( s^b_{ij} \\) being some softmax normalization statistics that need to be recomputed for every \\( i \\) and \\( j \\) .
Please note that the whole Flash Attention is a bit more complex and is greatly simplified here as going in too much depth is out of scope for this guide. The reader is invited to take a look at the well-written [Flash Attention paper](https://arxiv.org/abs/2205.14135) for more details.
The main takeaway here is:
> By keeping track of softmax normalization statistics and by using some smart mathematics, Flash Attention gives **numerical identical** outputs compared to the default self-attention layer at a memory cost that only increases linearly with \\( N \\) .
Looking at the formula, one would intuitively say that Flash Attention must be much slower compared to the default self-attention formula as more computation needs to be done. Indeed Flash Attention requires more FLOPs compared to normal attention as the softmax normalization statistics have to constantly be recomputed (see [paper](https://arxiv.org/abs/2205.14135) for more details if interested)
> However, Flash Attention is much faster in inference compared to default attention which comes from its ability to significantly reduce the demands on the slower, high-bandwidth memory of the GPU (VRAM), focusing instead on the faster on-chip memory (SRAM).
Essentially, Flash Attention makes sure that all intermediate write and read operations can be done using the fast *on-chip* SRAM memory instead of having to access the slower VRAM memory to compute the output vector \\( \mathbf{O} \\) .
In practice, there is currently absolutely no reason to **not** use Flash Attention if available. The algorithm gives mathematically the same outputs, and is both faster and more memory-efficient.
Let's look at a practical example.
Our OctoCoder model now gets a significantly longer input prompt which includes a so-called *system prompt*. System prompts are used to steer the LLM into a better assistant that is tailored to the users' task.
In the following, we use a system prompt that will make OctoCoder a better coding assistant.
```python
system_prompt = """Below are a series of dialogues between various people and an AI technical assistant.
The assistant tries to be helpful, polite, honest, sophisticated, emotionally aware, and humble but knowledgeable.
The assistant is happy to help with code questions and will do their best to understand exactly what is needed.
It also tries to avoid giving false or misleading information, and it caveats when it isn't entirely sure about the right answer.
That said, the assistant is practical really does its best, and doesn't let caution get too much in the way of being useful.
The Starcoder models are a series of 15.5B parameter models trained on 80+ programming languages from The Stack (v1.2) (excluding opt-out requests).
The model uses Multi Query Attention, was trained using the Fill-in-the-Middle objective, and with 8,192 tokens context window for a trillion tokens of heavily deduplicated data.
-----
Question: Write a function that takes two lists and returns a list that has alternating elements from each input list.
Answer: Sure. Here is a function that does that.
def alternating(list1, list2):
results = []
for i in range(len(list1)):
results.append(list1[i])
results.append(list2[i])
return results
Question: Can you write some test cases for this function?
Answer: Sure, here are some tests.
assert alternating([10, 20, 30], [1, 2, 3]) == [10, 1, 20, 2, 30, 3]
assert alternating([True, False], [4, 5]) == [True, 4, False, 5]
assert alternating([], []) == []
Question: Modify the function so that it returns all input elements when the lists have uneven length. The elements from the longer list should be at the end.
Answer: Here is the modified function.
def alternating(list1, list2):
results = []
for i in range(min(len(list1), len(list2))):
results.append(list1[i])
results.append(list2[i])
if len(list1) > len(list2):
results.extend(list1[i+1:])
else:
results.extend(list2[i+1:])
return results
-----
"""
```
For demonstration purposes, we duplicate the system prompt by ten so that the input length is long enough to observe Flash Attention's memory savings.
We append the original text prompt `"Question: Please write a function in Python that transforms bytes to Giga bytes.\n\nAnswer: Here"`
```python
long_prompt = 10 * system_prompt + prompt
```
We instantiate our model again in bfloat16 precision.
```python
model = AutoModelForCausalLM.from_pretrained("bigcode/octocoder", torch_dtype=torch.bfloat16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("bigcode/octocoder")
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)
```
Let's now run the model just like before *without Flash Attention* and measure the peak GPU memory requirement and inference time.
```python
import time
start_time = time.time()
result = pipe(long_prompt, max_new_tokens=60)[0]["generated_text"][len(long_prompt):]
print(f"Generated in {time.time() - start_time} seconds.")
result
```
**Output**:
```
Generated in 10.96854019165039 seconds.
Sure. Here is a function that does that.\n\ndef bytes_to_giga(bytes):\n return bytes / 1024 / 1024 / 1024\n\nAnswer: Sure. Here is a function that does that.\n\ndef
````
We're getting the same output as before, however this time, the model repeats the answer multiple times until it's 60 tokens cut-off. This is not surprising as we've repeated the system prompt ten times for demonstration purposes and thus cued the model to repeat itself.
**Note** that the system prompt should not be repeated ten times in real-world applications - one time is enough!
Let's measure the peak GPU memory requirement.
```python
bytes_to_giga_bytes(torch.cuda.max_memory_allocated())
```
**Output**:
```bash
37.668193340301514
```
As we can see the peak GPU memory requirement is now significantly higher than in the beginning, which is largely due to the longer input sequence. Also the generation takes a little over a minute now.
We call `flush()` to free GPU memory for our next experiment.
```python
flush()
```
For comparison, let's run the same function, but enable Flash Attention instead.
To do so, we convert the model to [BetterTransformer](https://huggingface.co/docs/optimum/bettertransformer/overview) and by doing so enabling PyTorch's [SDPA self-attention](https://pytorch.org/docs/master/generated/torch.nn.functional.scaled_dot_product_attention) which in turn is able to use Flash Attention.
```python
model.to_bettertransformer()
```
Now we run the exact same code snippet as before and under the hood Transformers will make use of Flash Attention.
```py
start_time = time.time()
with torch.backends.cuda.sdp_kernel(enable_flash=True, enable_math=False, enable_mem_efficient=False):
result = pipe(long_prompt, max_new_tokens=60)[0]["generated_text"][len(long_prompt):]
print(f"Generated in {time.time() - start_time} seconds.")
result
```
**Output**:
```
Generated in 3.0211617946624756 seconds.
Sure. Here is a function that does that.\n\ndef bytes_to_giga(bytes):\n return bytes / 1024 / 1024 / 1024\n\nAnswer: Sure. Here is a function that does that.\n\ndef
```
We're getting the exact same result as before, but can observe a very significant speed-up thanks to Flash Attention.
Let's measure the memory consumption one last time.
```python
bytes_to_giga_bytes(torch.cuda.max_memory_allocated())
```
**Output**:
```
32.617331981658936
```
And we're almost back to our original 29GB peak GPU memory from the beginning.
We can observe that we only use roughly 100MB more GPU memory when passing a very long input sequence with Flash Attention compared to passing a short input sequence as done in the beginning.
```py
flush()
```
For more information on how to use Flash Attention, please have a look at [this doc page](https://huggingface.co/docs/transformers/en/perf_infer_gpu_one#flashattention-2).
## 3. Architectural Innovations
So far we have looked into improving computational and memory efficiency by:
- Casting the weights to a lower precision format
- Replacing the self-attention algorithm with a more memory- and compute efficient version
Let's now look into how we can change the architecture of an LLM so that it is most effective and efficient for task that require long text inputs, *e.g.*:
- Retrieval augmented Questions Answering,
- Summarization,
- Chat
Note that *chat* not only requires the LLM to handle long text inputs, but it also necessitates that the LLM is able to efficiently handle the back-and-forth dialogue between user and assistant (such as ChatGPT).
Once trained, the fundamental LLM architecture is difficult to change, so it is important to make considerations about the LLM's tasks beforehand and accordingly optimize the model's architecture.
There are two important components of the model architecture that quickly become memory and/or performance bottlenecks for large input sequences.
- The positional embeddings
- The key-value cache
Let's go over each component in more detail
### 3.1 Improving positional embeddings of LLMs
Self-attention puts each token in relation to each other's tokens.
As an example, the \\( \text{Softmax}(\mathbf{QK}^T) \\) matrix of the text input sequence *"Hello", "I", "love", "you"* could look as follows:
Each word token is given a probability mass at which it attends all other word tokens and, therefore is put into relation with all other word tokens. E.g. the word *"love"* attends to the word *"Hello"* with 5%, to *"I"* with 30%, and to itself with 65%.
A LLM based on self-attention, but without position embeddings would have great difficulties in understanding the positions of the text inputs to each other.
This is because the probability score computed by \\( \mathbf{QK}^T \\) relates each word token to each other word token in \\( O(1) \\) computations regardless of their relative positional distance to each other.
Therefore, for the LLM without position embeddings each token appears to have the same distance to all other tokens, *e.g.* differentiating between *"Hello I love you"* and *"You love I hello"* would be very challenging.
For the LLM to understand sentence order, an additional *cue* is needed and is usually applied in the form of *positional encodings* (or also called *positional embeddings*).
Positional encodings, encode the position of each token into a numerical presentation that the LLM can leverage to better understand sentence order.
The authors of the [*Attention Is All You Need*](https://arxiv.org/abs/1706.03762) paper introduced sinusoidal positional embeddings \\( \mathbf{P} = \mathbf{p}_1, \ldots, \mathbf{p}_N \\) .
where each vector \\( \mathbf{p}_i \\) is computed as a sinusoidal function of its position \\( i \\) .
The positional encodings are then simply added to the input sequence vectors \\( \mathbf{\hat{X}} = \mathbf{\hat{x}}_1, \ldots, \mathbf{\hat{x}}_N \\) = \\( \mathbf{x}_1 + \mathbf{p}_1, \ldots, \mathbf{x}_N + \mathbf{p}_N \\) thereby cueing the model to better learn sentence order.
Instead of using fixed position embeddings, others (such as [Devlin et al.](https://arxiv.org/abs/1810.04805)) used learned positional encodings for which the positional embeddings
\\( \mathbf{P} \\) are learned during training.
Sinusoidal and learned position embeddings used to be the predominant methods to encode sentence order into LLMs, but a couple of problems related to these positional encodings were found:
1. Sinusoidal and learned position embeddings are both absolute positional embeddings, *i.e.* encoding a unique embedding for each position id: \\( 0, \ldots, N \\) . As shown by [Huang et al.](https://arxiv.org/abs/2009.13658) and [Su et al.](https://arxiv.org/abs/2104.09864), absolute positional embeddings lead to poor LLM performance for long text inputs. For long text inputs, it is advantageous if the model learns the relative positional distance input tokens have to each other instead of their absolute position.
2. When using learned position embeddings, the LLM has to be trained on a fixed input length \\( N \\), which makes it difficult to extrapolate to an input length longer than what it was trained on.
Recently, relative positional embeddings that can tackle the above mentioned problems have become more popular, most notably:
- [Rotary Position Embedding (RoPE)](https://arxiv.org/abs/2104.09864)
- [ALiBi](https://arxiv.org/abs/2108.12409)
Both *RoPE* and *ALiBi* argue that it's best to cue the LLM about sentence order directly in the self-attention algorithm as it's there that word tokens are put into relation with each other. More specifically, sentence order should be cued by modifying the \\( \mathbf{QK}^T \\) computation.
Without going into too many details, *RoPE* notes that positional information can be encoded into query-key pairs, *e.g.* \\( \mathbf{q}_i \\) and \\( \mathbf{x}_j \\) by rotating each vector by an angle \\( \theta * i \\) and \\( \theta * j \\) respectively with \\( i, j \\) describing each vectors sentence position:
$$ \mathbf{\hat{q}}_i^T \mathbf{\hat{x}}_j = \mathbf{{q}}_i^T \mathbf{R}_{\theta, i -j} \mathbf{{x}}_j. $$
\\( \mathbf{R}_{\theta, i - j} \\) thereby represents a rotational matrix. \\( \theta \\) is *not* learned during training, but instead set to a pre-defined value that depends on the maximum input sequence length during training.
> By doing so, the propability score between \\( \mathbf{q}_i \\) and \\( \mathbf{q}_j \\) is only affected if \\( i \ne j \\) and solely depends on the relative distance \\( i - j \\) regardless of each vector's specific positions \\( i \\) and \\( j \\) .
*RoPE* is used in multiple of today's most important LLMs, such as:
- [**Falcon**](https://huggingface.co/tiiuae/falcon-40b)
- [**Llama**](https://arxiv.org/abs/2302.13971)
- [**PaLM**](https://arxiv.org/abs/2204.02311)
As an alternative, *ALiBi* proposes a much simpler relative position encoding scheme. The relative distance that input tokens have to each other is added as a negative integer scaled by a pre-defined value `m` to each query-key entry of the \\( \mathbf{QK}^T \\) matrix right before the softmax computation.
As shown in the [ALiBi](https://arxiv.org/abs/2108.12409) paper, this simple relative positional encoding allows the model to retain a high performance even at very long text input sequences.
*ALiBi* is used in multiple of today's most important LLMs, such as:
- [**MPT**](https://huggingface.co/mosaicml/mpt-30b)
- [**BLOOM**](https://huggingface.co/bigscience/bloom)
Both *RoPE* and *ALiBi* position encodings can extrapolate to input lengths not seen during training whereas it has been shown that extrapolation works much better out-of-the-box for *ALiBi* as compared to *RoPE*.
For ALiBi, one simply increases the values of the lower triangular position matrix to match the length of the input sequence.
For *RoPE*, keeping the same \\( \theta \\) that was used during training leads to poor results when passing text inputs much longer than those seen during training, *c.f* [Press et al.](https://arxiv.org/abs/2108.12409). However, the community has found a couple of effective tricks that adapt \\( \theta \\), thereby allowing *RoPE* position embeddings to work well for extrapolated text input sequences (see [here](https://github.com/huggingface/transformers/pull/24653)).
> Both RoPE and ALiBi are relative positional embeddings that are *not* learned during training, but instead are based on the following intuitions:
- Positional cues about the text inputs should be given directly to the \\( QK^T \\) matrix of the self-attention layer
- The LLM should be incentivized to learn a constant *relative* distance positional encodings have to each other
- The further text input tokens are from each other, the lower the probability of their query-value probability. Both RoPE and ALiBi lower the query-key probability of tokens far away from each other. RoPE by decreasing their vector product by increasing the angle between the query-key vectors. ALiBi by adding large negative numbers to the vector product
In conclusion, LLMs that are intended to be deployed in tasks that require handling large text inputs are better trained with relative positional embeddings, such as RoPE and ALiBi. Also note that even if an LLM with RoPE and ALiBi has been trained only on a fixed length of say \\( N_1 = 2048 \\) it can still be used in practice with text inputs much larger than \\( N_1 \\), like \\( N_2 = 8192 > N_1 \\) by extrapolating the positional embeddings.
### 3.2 The key-value cache
Auto-regressive text generation with LLMs works by iteratively putting in an input sequence, sampling the next token, appending the next token to the input sequence, and continuing to do so until the LLM produces a token that signifies that the generation has finished.
Please have a look at [Transformer's Generate Text Tutorial](https://huggingface.co/docs/transformers/llm_tutorial#generate-text) to get a more visual explanation of how auto-regressive generation works.
Let's run a quick code snippet to show how auto-regressive works in practice. We will simply take the most likely next token via `torch.argmax`.
```python
input_ids = tokenizer(prompt, return_tensors="pt")["input_ids"].to("cuda")
for _ in range(5):
next_logits = model(input_ids)["logits"][:, -1:]
next_token_id = torch.argmax(next_logits,dim=-1)
input_ids = torch.cat([input_ids, next_token_id], dim=-1)
print("shape of input_ids", input_ids.shape)
generated_text = tokenizer.batch_decode(input_ids[:, -5:])
generated_text
```
**Output**:
```
shape of input_ids torch.Size([1, 21])
shape of input_ids torch.Size([1, 22])
shape of input_ids torch.Size([1, 23])
shape of input_ids torch.Size([1, 24])
shape of input_ids torch.Size([1, 25])
[' Here is a Python function']
```
As we can see every time we increase the text input tokens by the just sampled token.
With very few exceptions, LLMs are trained using the [causal language modeling objective](https://huggingface.co/docs/transformers/tasks/language_modeling#causal-language-modeling) and therefore mask the upper triangle matrix of the attention score - this is why in the two diagrams above the attention scores are left blank (*a.k.a* have 0 probability). For a quick recap on causal language modeling you can refer to the [*Illustrated Self Attention blog*](https://jalammar.github.io/illustrated-gpt2/#part-2-illustrated-self-attention).
As a consequence, tokens *never* depend on previous tokens, more specifically the \\( \mathbf{q}_i \\) vector is never put in relation with any key, values vectors \\( \mathbf{k}_j, \mathbf{v}_j \\) if \\( j > i \\) . Instead \\( \mathbf{q}_i \\) only attends to previous key-value vectors \\( \mathbf{k}_{m < i}, \mathbf{v}_{m < i} \text{ , for } m \in \{0, \ldots i - 1\} \\). In order to reduce unnecessary computation, one can therefore cache each layer's key-value vectors for all previous timesteps.
In the following, we will tell the LLM to make use of the key-value cache by retrieving and forwarding it for each forward pass.
In Transformers, we can retrieve the key-value cache by passing the `use_cache` flag to the `forward` call and can then pass it with the current token.
```python
past_key_values = None # past_key_values is the key-value cache
generated_tokens = []
next_token_id = tokenizer(prompt, return_tensors="pt")["input_ids"].to("cuda")
for _ in range(5):
next_logits, past_key_values = model(next_token_id, past_key_values=past_key_values, use_cache=True).to_tuple()
next_logits = next_logits[:, -1:]
next_token_id = torch.argmax(next_logits, dim=-1)
print("shape of input_ids", next_token_id.shape)
print("length of key-value cache", len(past_key_values[0][0])) # past_key_values are of shape [num_layers, 0 for k, 1 for v, batch_size, length, hidden_dim]
generated_tokens.append(next_token_id.item())
generated_text = tokenizer.batch_decode(generated_tokens)
generated_text
```
**Output**:
```
shape of input_ids torch.Size([1, 1])
length of key-value cache 20
shape of input_ids torch.Size([1, 1])
length of key-value cache 21
shape of input_ids torch.Size([1, 1])
length of key-value cache 22
shape of input_ids torch.Size([1, 1])
length of key-value cache 23
shape of input_ids torch.Size([1, 1])
length of key-value cache 24
[' Here', ' is', ' a', ' Python', ' function']
```
As one can see, when using the key-value cache the text input tokens are *not* increased in length, but remain a single input vector. The length of the key-value cache on the other hand is increased by one at every decoding step.
> Making use of the key-value cache means that the \\( \mathbf{QK}^T \\) is essentially reduced to \\( \mathbf{q}_c\mathbf{K}^T \\) with \\( \mathbf{q}_c \\) being the query projection of the currently passed input token which is *always* just a single vector.
Using the key-value cache has two advantages:
- Significant increase in computational efficiency as less computations are performed compared to computing the full \\( \mathbf{QK}^T \\) matrix. This leads to an increase in inference speed
- The maximum required memory is not increased quadratically with the number of generated tokens, but only increases linearly.
> One should *always* make use of the key-value cache as it leads to identical results and a significant speed-up for longer input sequences. Transformers has the key-value cache enabled by default when making use of the text pipeline or the [`generate` method](https://huggingface.co/docs/transformers/main_classes/text_generation).
<Tip warning={true}>
Note that, despite our advice to use key-value caches, your LLM output may be slightly different when you use them. This is a property of the matrix multiplication kernels themselves -- you can read more about it [here](https://github.com/huggingface/transformers/issues/25420#issuecomment-1775317535).
</Tip>
#### 3.2.1 Multi-round conversation
The key-value cache is especially useful for applications such as chat where multiple passes of auto-regressive decoding are required. Let's look at an example.
```
User: How many people live in France?
Assistant: Roughly 75 million people live in France
User: And how many are in Germany?
Assistant: Germany has ca. 81 million inhabitants
```
In this chat, the LLM runs auto-regressive decoding twice:
1. The first time, the key-value cache is empty and the input prompt is `"User: How many people live in France?"` and the model auto-regressively generates the text `"Roughly 75 million people live in France"` while increasing the key-value cache at every decoding step.
2. The second time the input prompt is `"User: How many people live in France? \n Assistant: Roughly 75 million people live in France \n User: And how many in Germany?"`. Thanks to the cache, all key-value vectors for the first two sentences are already computed. Therefore the input prompt only consists of `"User: And how many in Germany?"`. While processing the shortened input prompt, it's computed key-value vectors are concatenated to the key-value cache of the first decoding. The second Assistant's answer `"Germany has ca. 81 million inhabitants"` is then auto-regressively generated with the key-value cache consisting of encoded key-value vectors of `"User: How many people live in France? \n Assistant: Roughly 75 million people live in France \n User: And how many are in Germany?"`.
Two things should be noted here:
1. Keeping all the context is crucial for LLMs deployed in chat so that the LLM understands all the previous context of the conversation. E.g. for the example above the LLM needs to understand that the user refers to the population when asking `"And how many are in Germany"`.
2. The key-value cache is extremely useful for chat as it allows us to continuously grow the encoded chat history instead of having to re-encode the chat history again from scratch (as e.g. would be the case when using an encoder-decoder architecture).
In `transformers`, a `generate` call will return `past_key_values` when `return_dict_in_generate=True` is passed, in addition to the default `use_cache=True`. Note that it is not yet available through the `pipeline` interface.
```python
# Generation as usual
prompt = system_prompt + "Question: Please write a function in Python that transforms bytes to Giga bytes.\n\nAnswer: Here"
model_inputs = tokenizer(prompt, return_tensors='pt')
generation_output = model.generate(**model_inputs, max_new_tokens=60, return_dict_in_generate=True)
decoded_output = tokenizer.batch_decode(generation_output.sequences)[0]
# Piping the returned `past_key_values` to speed up the next conversation round
prompt = decoded_output + "\nQuestion: How can I modify the function above to return Mega bytes instead?\n\nAnswer: Here"
model_inputs = tokenizer(prompt, return_tensors='pt')
generation_output = model.generate(
**model_inputs,
past_key_values=generation_output.past_key_values,
max_new_tokens=60,
return_dict_in_generate=True
)
tokenizer.batch_decode(generation_output.sequences)[0][len(prompt):]
```
**Output**:
```
is a modified version of the function that returns Mega bytes instead.
def bytes_to_megabytes(bytes):
return bytes / 1024 / 1024
Answer: The function takes a number of bytes as input and returns the number of
```
Great, no additional time is spent recomputing the same key and values for the attention layer! There is however one catch. While the required peak memory for the \\( \mathbf{QK}^T \\) matrix is significantly reduced, holding the key-value cache in memory can become very memory expensive for long input sequences or multi-turn chat. Remember that the key-value cache needs to store the key-value vectors for all previous input vectors \\( \mathbf{x}_i \text{, for } i \in \{1, \ldots, c - 1\} \\) for all self-attention layers and for all attention heads.
Let's compute the number of float values that need to be stored in the key-value cache for the LLM `bigcode/octocoder` that we used before.
The number of float values amounts to two times the sequence length times the number of attention heads times the attention head dimension and times the number of layers.
Computing this for our LLM at a hypothetical input sequence length of 16000 gives:
```python
config = model.config
2 * 16_000 * config.n_layer * config.n_head * config.n_embd // config.n_head
```
**Output**:
```
7864320000
```
Roughly 8 billion float values! Storing 8 billion float values in `float16` precision requires around 15 GB of RAM which is circa half as much as the model weights themselves!
Researchers have proposed two methods that allow to significantly reduce the memory cost of storing the key-value cache, which are explored in the next subsections.
#### 3.2.2 Multi-Query-Attention (MQA)
[Multi-Query-Attention](https://arxiv.org/abs/1911.02150) was proposed in Noam Shazeer's *Fast Transformer Decoding: One Write-Head is All You Need* paper. As the title says, Noam found out that instead of using `n_head` key-value projections weights, one can use a single head-value projection weight pair that is shared across all attention heads without that the model's performance significantly degrades.
> By using a single head-value projection weight pair, the key value vectors \\( \mathbf{k}_i, \mathbf{v}_i \\) have to be identical across all attention heads which in turn means that we only need to store 1 key-value projection pair in the cache instead of `n_head` ones.
As most LLMs use between 20 and 100 attention heads, MQA significantly reduces the memory consumption of the key-value cache. For the LLM used in this notebook we could therefore reduce the required memory consumption from 15 GB to less than 400 MB at an input sequence length of 16000.
In addition to memory savings, MQA also leads to improved computational efficiency as explained in the following.
In auto-regressive decoding, large key-value vectors need to be reloaded, concatenated with the current key-value vector pair to be then fed into the \\( \mathbf{q}_c\mathbf{K}^T \\) computation at every step. For auto-regressive decoding, the required memory bandwidth for the constant reloading can become a serious time bottleneck. By reducing the size of the key-value vectors less memory needs to be accessed, thus reducing the memory bandwidth bottleneck. For more detail, please have a look at [Noam's paper](https://arxiv.org/abs/1911.02150).
The important part to understand here is that reducing the number of key-value attention heads to 1 only makes sense if a key-value cache is used. The peak memory consumption of the model for a single forward pass without key-value cache stays unchanged as every attention head still has a unique query vector so that each attention head still has a different \\( \mathbf{QK}^T \\) matrix.
MQA has seen wide adoption by the community and is now used by many of the most popular LLMs:
- [**Falcon**](https://huggingface.co/tiiuae/falcon-40b)
- [**PaLM**](https://arxiv.org/abs/2204.02311)
- [**MPT**](https://huggingface.co/mosaicml/mpt-30b)
- [**BLOOM**](https://huggingface.co/bigscience/bloom)
Also, the checkpoint used in this notebook - `bigcode/octocoder` - makes use of MQA.
#### 3.2.3 Grouped-Query-Attention (GQA)
[Grouped-Query-Attention](https://arxiv.org/abs/2305.13245), as proposed by Ainslie et al. from Google, found that using MQA can often lead to quality degradation compared to using vanilla multi-key-value head projections. The paper argues that more model performance can be kept by less drastically reducing the number of query head projection weights. Instead of using just a single key-value projection weight, `n < n_head` key-value projection weights should be used. By choosing `n` to a significantly smaller value than `n_head`, such as 2,4 or 8 almost all of the memory and speed gains from MQA can be kept while sacrificing less model capacity and thus arguably less performance.
Moreover, the authors of GQA found out that existing model checkpoints can be *uptrained* to have a GQA architecture with as little as 5% of the original pre-training compute. While 5% of the original pre-training compute can still be a massive amount, GQA *uptraining* allows existing checkpoints to be useful for longer input sequences.
GQA was only recently proposed which is why there is less adoption at the time of writing this notebook.
The most notable application of GQA is [Llama-v2](https://huggingface.co/meta-llama/Llama-2-70b-hf).
> As a conclusion, it is strongly recommended to make use of either GQA or MQA if the LLM is deployed with auto-regressive decoding and is required to handle large input sequences as is the case for example for chat.
## Conclusion
The research community is constantly coming up with new, nifty ways to speed up inference time for ever-larger LLMs. As an example, one such promising research direction is [speculative decoding](https://arxiv.org/abs/2211.17192) where "easy tokens" are generated by smaller, faster language models and only "hard tokens" are generated by the LLM itself. Going into more detail is out of the scope of this notebook, but can be read upon in this [nice blog post](https://huggingface.co/blog/assisted-generation).
The reason massive LLMs such as GPT3/4, Llama-2-70b, Claude, PaLM can run so quickly in chat-interfaces such as [Hugging Face Chat](https://huggingface.co/chat/) or ChatGPT is to a big part thanks to the above-mentioned improvements in precision, algorithms, and architecture.
Going forward, accelerators such as GPUs, TPUs, etc... will only get faster and allow for more memory, but one should nevertheless always make sure to use the best available algorithms and architectures to get the most bang for your buck 🤗
| huggingface/transformers/blob/main/docs/source/en/llm_tutorial_optimization.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# LED
## Overview
The LED model was proposed in [Longformer: The Long-Document Transformer](https://arxiv.org/abs/2004.05150) by Iz
Beltagy, Matthew E. Peters, Arman Cohan.
The abstract from the paper is the following:
*Transformer-based models are unable to process long sequences due to their self-attention operation, which scales
quadratically with the sequence length. To address this limitation, we introduce the Longformer with an attention
mechanism that scales linearly with sequence length, making it easy to process documents of thousands of tokens or
longer. Longformer's attention mechanism is a drop-in replacement for the standard self-attention and combines a local
windowed attention with a task motivated global attention. Following prior work on long-sequence transformers, we
evaluate Longformer on character-level language modeling and achieve state-of-the-art results on text8 and enwik8. In
contrast to most prior work, we also pretrain Longformer and finetune it on a variety of downstream tasks. Our
pretrained Longformer consistently outperforms RoBERTa on long document tasks and sets new state-of-the-art results on
WikiHop and TriviaQA. We finally introduce the Longformer-Encoder-Decoder (LED), a Longformer variant for supporting
long document generative sequence-to-sequence tasks, and demonstrate its effectiveness on the arXiv summarization
dataset.*
## Usage tips
- [`LEDForConditionalGeneration`] is an extension of
[`BartForConditionalGeneration`] exchanging the traditional *self-attention* layer with
*Longformer*'s *chunked self-attention* layer. [`LEDTokenizer`] is an alias of
[`BartTokenizer`].
- LED works very well on long-range *sequence-to-sequence* tasks where the `input_ids` largely exceed a length of
1024 tokens.
- LED pads the `input_ids` to be a multiple of `config.attention_window` if required. Therefore a small speed-up is
gained, when [`LEDTokenizer`] is used with the `pad_to_multiple_of` argument.
- LED makes use of *global attention* by means of the `global_attention_mask` (see
[`LongformerModel`]). For summarization, it is advised to put *global attention* only on the first
`<s>` token. For question answering, it is advised to put *global attention* on all tokens of the question.
- To fine-tune LED on all 16384, *gradient checkpointing* can be enabled in case training leads to out-of-memory (OOM)
errors. This can be done by executing `model.gradient_checkpointing_enable()`.
Moreover, the `use_cache=False`
flag can be used to disable the caching mechanism to save memory.
- LED is a model with absolute position embeddings so it's usually advised to pad the inputs on the right rather than
the left.
This model was contributed by [patrickvonplaten](https://huggingface.co/patrickvonplaten).
## Resources
- [A notebook showing how to evaluate LED](https://colab.research.google.com/drive/12INTTR6n64TzS4RrXZxMSXfrOd9Xzamo?usp=sharing).
- [A notebook showing how to fine-tune LED](https://colab.research.google.com/drive/12LjJazBl7Gam0XBPy_y0CTOJZeZ34c2v?usp=sharing).
- [Text classification task guide](../tasks/sequence_classification)
- [Question answering task guide](../tasks/question_answering)
- [Translation task guide](../tasks/translation)
- [Summarization task guide](../tasks/summarization)
## LEDConfig
[[autodoc]] LEDConfig
## LEDTokenizer
[[autodoc]] LEDTokenizer
- build_inputs_with_special_tokens
- get_special_tokens_mask
- create_token_type_ids_from_sequences
- save_vocabulary
## LEDTokenizerFast
[[autodoc]] LEDTokenizerFast
## LED specific outputs
[[autodoc]] models.led.modeling_led.LEDEncoderBaseModelOutput
[[autodoc]] models.led.modeling_led.LEDSeq2SeqModelOutput
[[autodoc]] models.led.modeling_led.LEDSeq2SeqLMOutput
[[autodoc]] models.led.modeling_led.LEDSeq2SeqSequenceClassifierOutput
[[autodoc]] models.led.modeling_led.LEDSeq2SeqQuestionAnsweringModelOutput
[[autodoc]] models.led.modeling_tf_led.TFLEDEncoderBaseModelOutput
[[autodoc]] models.led.modeling_tf_led.TFLEDSeq2SeqModelOutput
[[autodoc]] models.led.modeling_tf_led.TFLEDSeq2SeqLMOutput
<frameworkcontent>
<pt>
## LEDModel
[[autodoc]] LEDModel
- forward
## LEDForConditionalGeneration
[[autodoc]] LEDForConditionalGeneration
- forward
## LEDForSequenceClassification
[[autodoc]] LEDForSequenceClassification
- forward
## LEDForQuestionAnswering
[[autodoc]] LEDForQuestionAnswering
- forward
</pt>
<tf>
## TFLEDModel
[[autodoc]] TFLEDModel
- call
## TFLEDForConditionalGeneration
[[autodoc]] TFLEDForConditionalGeneration
- call
</tf>
</frameworkcontent>
| huggingface/transformers/blob/main/docs/source/en/model_doc/led.md |
!--Copyright 2021 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# DePlot
## Overview
DePlot was proposed in the paper [DePlot: One-shot visual language reasoning by plot-to-table translation](https://arxiv.org/abs/2212.10505) from Fangyu Liu, Julian Martin Eisenschlos, Francesco Piccinno, Syrine Krichene, Chenxi Pang, Kenton Lee, Mandar Joshi, Wenhu Chen, Nigel Collier, Yasemin Altun.
The abstract of the paper states the following:
*Visual language such as charts and plots is ubiquitous in the human world. Comprehending plots and charts requires strong reasoning skills. Prior state-of-the-art (SOTA) models require at least tens of thousands of training examples and their reasoning capabilities are still much limited, especially on complex human-written queries. This paper presents the first one-shot solution to visual language reasoning. We decompose the challenge of visual language reasoning into two steps: (1) plot-to-text translation, and (2) reasoning over the translated text. The key in this method is a modality conversion module, named as DePlot, which translates the image of a plot or chart to a linearized table. The output of DePlot can then be directly used to prompt a pretrained large language model (LLM), exploiting the few-shot reasoning capabilities of LLMs. To obtain DePlot, we standardize the plot-to-table task by establishing unified task formats and metrics, and train DePlot end-to-end on this task. DePlot can then be used off-the-shelf together with LLMs in a plug-and-play fashion. Compared with a SOTA model finetuned on more than >28k data points, DePlot+LLM with just one-shot prompting achieves a 24.0% improvement over finetuned SOTA on human-written queries from the task of chart QA.*
DePlot is a model that is trained using `Pix2Struct` architecture. You can find more information about `Pix2Struct` in the [Pix2Struct documentation](https://huggingface.co/docs/transformers/main/en/model_doc/pix2struct).
DePlot is a Visual Question Answering subset of `Pix2Struct` architecture. It renders the input question on the image and predicts the answer.
## Usage example
Currently one checkpoint is available for DePlot:
- `google/deplot`: DePlot fine-tuned on ChartQA dataset
```python
from transformers import AutoProcessor, Pix2StructForConditionalGeneration
import requests
from PIL import Image
model = Pix2StructForConditionalGeneration.from_pretrained("google/deplot")
processor = AutoProcessor.from_pretrained("google/deplot")
url = "https://raw.githubusercontent.com/vis-nlp/ChartQA/main/ChartQA%20Dataset/val/png/5090.png"
image = Image.open(requests.get(url, stream=True).raw)
inputs = processor(images=image, text="Generate underlying data table of the figure below:", return_tensors="pt")
predictions = model.generate(**inputs, max_new_tokens=512)
print(processor.decode(predictions[0], skip_special_tokens=True))
```
## Fine-tuning
To fine-tune DePlot, refer to the pix2struct [fine-tuning notebook](https://github.com/huggingface/notebooks/blob/main/examples/image_captioning_pix2struct.ipynb). For `Pix2Struct` models, we have found out that fine-tuning the model with Adafactor and cosine learning rate scheduler leads to faster convergence:
```python
from transformers.optimization import Adafactor, get_cosine_schedule_with_warmup
optimizer = Adafactor(self.parameters(), scale_parameter=False, relative_step=False, lr=0.01, weight_decay=1e-05)
scheduler = get_cosine_schedule_with_warmup(optimizer, num_warmup_steps=1000, num_training_steps=40000)
```
<Tip>
DePlot is a model trained using `Pix2Struct` architecture. For API reference, see [`Pix2Struct` documentation](pix2struct).
</Tip> | huggingface/transformers/blob/main/docs/source/en/model_doc/deplot.md |
*NOTE**: This example is outdated and is not longer actively maintained. Please
follow the new instructions of fine-tuning Wav2Vec2 [here](https://github.com/huggingface/transformers/blob/main/examples/pytorch/speech-recognition/README.md)
## Fine-tuning Wav2Vec2
The `run_asr.py` script allows one to fine-tune pretrained Wav2Vec2 models that can be found [here](https://huggingface.co/models?search=facebook/wav2vec2).
This finetuning script can also be run as a google colab [TODO: here]( ).
### Fine-Tuning with TIMIT
Let's take a look at the [script](./finetune_base_timit_asr.sh) used to fine-tune [wav2vec2-base](https://huggingface.co/facebook/wav2vec2-base)
with the [TIMIT dataset](https://huggingface.co/datasets/timit_asr):
```bash
#!/usr/bin/env bash
python run_asr.py \
--output_dir="./wav2vec2-base-timit-asr" \
--num_train_epochs="30" \
--per_device_train_batch_size="20" \
--per_device_eval_batch_size="20" \
--evaluation_strategy="steps" \
--save_steps="500" \
--eval_steps="100" \
--logging_steps="50" \
--learning_rate="5e-4" \
--warmup_steps="3000" \
--model_name_or_path="facebook/wav2vec2-base" \
--fp16 \
--dataset_name="timit_asr" \
--train_split_name="train" \
--validation_split_name="test" \
--orthography="timit" \
--preprocessing_num_workers="$(nproc)" \
--group_by_length \
--freeze_feature_extractor \
--verbose_logging \
```
The resulting model and inference examples can be found [here](https://huggingface.co/elgeish/wav2vec2-base-timit-asr).
Some of the arguments above may look unfamiliar, let's break down what's going on:
`--orthography="timit"` applies certain text preprocessing rules, for tokenization and normalization, to clean up the dataset.
In this case, we use the following instance of `Orthography`:
```python
Orthography(
do_lower_case=True,
# break compounds like "quarter-century-old" and replace pauses "--"
translation_table=str.maketrans({"-": " "}),
)
```
The instance above is used as follows:
* creates a tokenizer with `do_lower_case=True` (ignores casing for input and lowercases output when decoding)
* replaces `"-"` with `" "` to break compounds like `"quarter-century-old"` and to clean up suspended hyphens
* cleans up consecutive whitespaces (replaces them with a single space: `" "`)
* removes characters not in vocabulary (lacking respective sound units)
`--verbose_logging` logs text preprocessing updates and when evaluating, using the validation split every `eval_steps`,
logs references and predictions.
### Fine-Tuning with Arabic Speech Corpus
Other datasets, like the [Arabic Speech Corpus dataset](https://huggingface.co/datasets/arabic_speech_corpus),
require more work! Let's take a look at the [script](./finetune_large_xlsr_53_arabic_speech_corpus.sh)
used to fine-tune [wav2vec2-large-xlsr-53](https://huggingface.co/elgeish/wav2vec2-large-xlsr-53-arabic):
```bash
#!/usr/bin/env bash
python run_asr.py \
--output_dir="./wav2vec2-large-xlsr-53-arabic-speech-corpus" \
--num_train_epochs="50" \
--per_device_train_batch_size="1" \
--per_device_eval_batch_size="1" \
--gradient_accumulation_steps="8" \
--evaluation_strategy="steps" \
--save_steps="500" \
--eval_steps="100" \
--logging_steps="50" \
--learning_rate="5e-4" \
--warmup_steps="3000" \
--model_name_or_path="elgeish/wav2vec2-large-xlsr-53-arabic" \
--fp16 \
--dataset_name="arabic_speech_corpus" \
--train_split_name="train" \
--validation_split_name="test" \
--max_duration_in_seconds="15" \
--orthography="buckwalter" \
--preprocessing_num_workers="$(nproc)" \
--group_by_length \
--freeze_feature_extractor \
--target_feature_extractor_sampling_rate \
--verbose_logging \
```
First, let's understand how this dataset represents Arabic text; it uses a format called
[Buckwalter transliteration](https://en.wikipedia.org/wiki/Buckwalter_transliteration).
We use the [lang-trans](https://github.com/kariminf/lang-trans) package to convert back to Arabic when logging.
The Buckwalter format only includes ASCII characters, some of which are non-alpha (e.g., `">"` maps to `"أ"`).
`--orthography="buckwalter"` applies certain text preprocessing rules, for tokenization and normalization, to clean up the dataset. In this case, we use the following instance of `Orthography`:
```python
Orthography(
vocab_file=pathlib.Path(__file__).parent.joinpath("vocab/buckwalter.json"),
word_delimiter_token="/", # "|" is Arabic letter alef with madda above
words_to_remove={"sil"}, # fixing "sil" in arabic_speech_corpus dataset
untransliterator=arabic.buckwalter.untransliterate,
translation_table=str.maketrans(translation_table = {
"-": " ", # sometimes used to represent pauses
"^": "v", # fixing "tha" in arabic_speech_corpus dataset
}),
)
```
The instance above is used as follows:
* creates a tokenizer with Buckwalter vocabulary and `word_delimiter_token="/"`
* replaces `"-"` with `" "` to clean up hyphens and fixes the orthography for `"ث"`
* removes words used as indicators (in this case, `"sil"` is used for silence)
* cleans up consecutive whitespaces (replaces them with a single space: `" "`)
* removes characters not in vocabulary (lacking respective sound units)
`--verbose_logging` logs text preprocessing updates and when evaluating, using the validation split every `eval_steps`,
logs references and predictions. Using the Buckwalter format, text is also logged in Arabic abjad.
`--target_feature_extractor_sampling_rate` resamples audio to target feature extractor's sampling rate (16kHz).
`--max_duration_in_seconds="15"` filters out examples whose audio is longer than the specified limit,
which helps with capping GPU memory usage.
### DeepSpeed Integration
To learn how to deploy Deepspeed Integration please refer to [this guide](https://huggingface.co/transformers/main/main_classes/deepspeed.html#deepspeed-trainer-integration).
But to get started quickly all you need is to install:
```
pip install deepspeed
```
and then use the default configuration files in this directory:
* `ds_config_wav2vec2_zero2.json`
* `ds_config_wav2vec2_zero3.json`
Here are examples of how you can use DeepSpeed:
(edit the value for `--num_gpus` to match the number of GPUs you have)
ZeRO-2:
```
PYTHONPATH=../../../src deepspeed --num_gpus 2 \
run_asr.py \
--output_dir=output_dir --num_train_epochs=2 --per_device_train_batch_size=2 \
--per_device_eval_batch_size=2 --evaluation_strategy=steps --save_steps=500 --eval_steps=100 \
--logging_steps=5 --learning_rate=5e-4 --warmup_steps=3000 \
--model_name_or_path=patrickvonplaten/wav2vec2_tiny_random_robust \
--dataset_name=hf-internal-testing/librispeech_asr_dummy --dataset_config_name=clean \
--train_split_name=validation --validation_split_name=validation --orthography=timit \
--preprocessing_num_workers=1 --group_by_length --freeze_feature_extractor --verbose_logging \
--deepspeed ds_config_wav2vec2_zero2.json
```
For ZeRO-2 with more than 1 gpu you need to use (which is already in the example configuration file):
```
"zero_optimization": {
...
"find_unused_parameters": true,
...
}
```
ZeRO-3:
```
PYTHONPATH=../../../src deepspeed --num_gpus 2 \
run_asr.py \
--output_dir=output_dir --num_train_epochs=2 --per_device_train_batch_size=2 \
--per_device_eval_batch_size=2 --evaluation_strategy=steps --save_steps=500 --eval_steps=100 \
--logging_steps=5 --learning_rate=5e-4 --warmup_steps=3000 \
--model_name_or_path=patrickvonplaten/wav2vec2_tiny_random_robust \
--dataset_name=hf-internal-testing/librispeech_asr_dummy --dataset_config_name=clean \
--train_split_name=validation --validation_split_name=validation --orthography=timit \
--preprocessing_num_workers=1 --group_by_length --freeze_feature_extractor --verbose_logging \
--deepspeed ds_config_wav2vec2_zero3.json
```
### Pretraining Wav2Vec2
The `run_pretrain.py` script allows one to pretrain a Wav2Vec2 model from scratch using Wav2Vec2's contrastive loss objective (see official [paper](https://arxiv.org/abs/2006.11477) for more information).
It is recommended to pre-train Wav2Vec2 with Trainer + Deepspeed (please refer to [this guide](https://huggingface.co/transformers/main/main_classes/deepspeed.html#deepspeed-trainer-integration) for more information).
Here is an example of how you can use DeepSpeed ZeRO-2 to pretrain a small Wav2Vec2 model:
```
PYTHONPATH=../../../src deepspeed --num_gpus 4 run_pretrain.py \
--output_dir="./wav2vec2-base-libri-100h" \
--num_train_epochs="3" \
--per_device_train_batch_size="32" \
--per_device_eval_batch_size="32" \
--gradient_accumulation_steps="2" \
--save_total_limit="3" \
--save_steps="500" \
--logging_steps="10" \
--learning_rate="5e-4" \
--weight_decay="0.01" \
--warmup_steps="3000" \
--model_name_or_path="patrickvonplaten/wav2vec2-base-libri-100h" \
--dataset_name="librispeech_asr" \
--dataset_config_name="clean" \
--train_split_name="train.100" \
--preprocessing_num_workers="4" \
--max_duration_in_seconds="10.0" \
--group_by_length \
--verbose_logging \
--fp16 \
--deepspeed ds_config_wav2vec2_zero2.json \
```
### Forced Alignment
Character level forced alignment for audio and text pairs with wav2vec2 models finetuned on ASR task for a specific language.
Inspired by [this](https://pytorch.org/tutorials/intermediate/forced_alignment_with_torchaudio_tutorial.html) Pytorch tutorial.
#### Input Formats
Input format in script.txt Input format in wavs directroy
0000 sentence1 0000.wav
0001 sentence2 0001.wav
#### Output Format
Output directory will contain 0000.txt and 0001.txt. Each file will have format like below
char score start_ms end_ms
h 0.25 1440 1520
#### Run command
```
python alignment.py \
--model_name="arijitx/wav2vec2-xls-r-300m-bengali" \
--wav_dir="./wavs"
--text_file="script.txt" \
--input_wavs_sr=48000 \
--output_dir="./out_alignment" \
--cuda
```
| huggingface/transformers/blob/main/examples/research_projects/wav2vec2/README.md |
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# X-MOD
## Overview
The X-MOD model was proposed in [Lifting the Curse of Multilinguality by Pre-training Modular Transformers](http://dx.doi.org/10.18653/v1/2022.naacl-main.255) by Jonas Pfeiffer, Naman Goyal, Xi Lin, Xian Li, James Cross, Sebastian Riedel, and Mikel Artetxe.
X-MOD extends multilingual masked language models like [XLM-R](xlm-roberta) to include language-specific modular components (_language adapters_) during pre-training. For fine-tuning, the language adapters in each transformer layer are frozen.
The abstract from the paper is the following:
*Multilingual pre-trained models are known to suffer from the curse of multilinguality, which causes per-language performance to drop as they cover more languages. We address this issue by introducing language-specific modules, which allows us to grow the total capacity of the model, while keeping the total number of trainable parameters per language constant. In contrast with prior work that learns language-specific components post-hoc, we pre-train the modules of our Cross-lingual Modular (X-MOD) models from the start. Our experiments on natural language inference, named entity recognition and question answering show that our approach not only mitigates the negative interference between languages, but also enables positive transfer, resulting in improved monolingual and cross-lingual performance. Furthermore, our approach enables adding languages post-hoc with no measurable drop in performance, no longer limiting the model usage to the set of pre-trained languages.*
This model was contributed by [jvamvas](https://huggingface.co/jvamvas).
The original code can be found [here](https://github.com/facebookresearch/fairseq/tree/58cc6cca18f15e6d56e3f60c959fe4f878960a60/fairseq/models/xmod) and the original documentation is found [here](https://github.com/facebookresearch/fairseq/tree/58cc6cca18f15e6d56e3f60c959fe4f878960a60/examples/xmod).
## Usage tips
Tips:
- X-MOD is similar to [XLM-R](xlm-roberta), but a difference is that the input language needs to be specified so that the correct language adapter can be activated.
- The main models – base and large – have adapters for 81 languages.
## Adapter Usage
### Input language
There are two ways to specify the input language:
1. By setting a default language before using the model:
```python
from transformers import XmodModel
model = XmodModel.from_pretrained("facebook/xmod-base")
model.set_default_language("en_XX")
```
2. By explicitly passing the index of the language adapter for each sample:
```python
import torch
input_ids = torch.tensor(
[
[0, 581, 10269, 83, 99942, 136, 60742, 23, 70, 80583, 18276, 2],
[0, 1310, 49083, 443, 269, 71, 5486, 165, 60429, 660, 23, 2],
]
)
lang_ids = torch.LongTensor(
[
0, # en_XX
8, # de_DE
]
)
output = model(input_ids, lang_ids=lang_ids)
```
### Fine-tuning
The paper recommends that the embedding layer and the language adapters are frozen during fine-tuning. A method for doing this is provided:
```python
model.freeze_embeddings_and_language_adapters()
# Fine-tune the model ...
```
### Cross-lingual transfer
After fine-tuning, zero-shot cross-lingual transfer can be tested by activating the language adapter of the target language:
```python
model.set_default_language("de_DE")
# Evaluate the model on German examples ...
```
## Resources
- [Text classification task guide](../tasks/sequence_classification)
- [Token classification task guide](../tasks/token_classification)
- [Question answering task guide](../tasks/question_answering)
- [Causal language modeling task guide](../tasks/language_modeling)
- [Masked language modeling task guide](../tasks/masked_language_modeling)
- [Multiple choice task guide](../tasks/multiple_choice)
## XmodConfig
[[autodoc]] XmodConfig
## XmodModel
[[autodoc]] XmodModel
- forward
## XmodForCausalLM
[[autodoc]] XmodForCausalLM
- forward
## XmodForMaskedLM
[[autodoc]] XmodForMaskedLM
- forward
## XmodForSequenceClassification
[[autodoc]] XmodForSequenceClassification
- forward
## XmodForMultipleChoice
[[autodoc]] XmodForMultipleChoice
- forward
## XmodForTokenClassification
[[autodoc]] XmodForTokenClassification
- forward
## XmodForQuestionAnswering
[[autodoc]] XmodForQuestionAnswering
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/xmod.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Logging
🤗 Transformers has a centralized logging system, so that you can setup the verbosity of the library easily.
Currently the default verbosity of the library is `WARNING`.
To change the level of verbosity, just use one of the direct setters. For instance, here is how to change the verbosity
to the INFO level.
```python
import transformers
transformers.logging.set_verbosity_info()
```
You can also use the environment variable `TRANSFORMERS_VERBOSITY` to override the default verbosity. You can set it
to one of the following: `debug`, `info`, `warning`, `error`, `critical`. For example:
```bash
TRANSFORMERS_VERBOSITY=error ./myprogram.py
```
Additionally, some `warnings` can be disabled by setting the environment variable
`TRANSFORMERS_NO_ADVISORY_WARNINGS` to a true value, like *1*. This will disable any warning that is logged using
[`logger.warning_advice`]. For example:
```bash
TRANSFORMERS_NO_ADVISORY_WARNINGS=1 ./myprogram.py
```
Here is an example of how to use the same logger as the library in your own module or script:
```python
from transformers.utils import logging
logging.set_verbosity_info()
logger = logging.get_logger("transformers")
logger.info("INFO")
logger.warning("WARN")
```
All the methods of this logging module are documented below, the main ones are
[`logging.get_verbosity`] to get the current level of verbosity in the logger and
[`logging.set_verbosity`] to set the verbosity to the level of your choice. In order (from the least
verbose to the most verbose), those levels (with their corresponding int values in parenthesis) are:
- `transformers.logging.CRITICAL` or `transformers.logging.FATAL` (int value, 50): only report the most
critical errors.
- `transformers.logging.ERROR` (int value, 40): only report errors.
- `transformers.logging.WARNING` or `transformers.logging.WARN` (int value, 30): only reports error and
warnings. This the default level used by the library.
- `transformers.logging.INFO` (int value, 20): reports error, warnings and basic information.
- `transformers.logging.DEBUG` (int value, 10): report all information.
By default, `tqdm` progress bars will be displayed during model download. [`logging.disable_progress_bar`] and [`logging.enable_progress_bar`] can be used to suppress or unsuppress this behavior.
## `logging` vs `warnings`
Python has two logging systems that are often used in conjunction: `logging`, which is explained above, and `warnings`,
which allows further classification of warnings in specific buckets, e.g., `FutureWarning` for a feature or path
that has already been deprecated and `DeprecationWarning` to indicate an upcoming deprecation.
We use both in the `transformers` library. We leverage and adapt `logging`'s `captureWarning` method to allow
management of these warning messages by the verbosity setters above.
What does that mean for developers of the library? We should respect the following heuristic:
- `warnings` should be favored for developers of the library and libraries dependent on `transformers`
- `logging` should be used for end-users of the library using it in every-day projects
See reference of the `captureWarnings` method below.
[[autodoc]] logging.captureWarnings
## Base setters
[[autodoc]] logging.set_verbosity_error
[[autodoc]] logging.set_verbosity_warning
[[autodoc]] logging.set_verbosity_info
[[autodoc]] logging.set_verbosity_debug
## Other functions
[[autodoc]] logging.get_verbosity
[[autodoc]] logging.set_verbosity
[[autodoc]] logging.get_logger
[[autodoc]] logging.enable_default_handler
[[autodoc]] logging.disable_default_handler
[[autodoc]] logging.enable_explicit_format
[[autodoc]] logging.reset_format
[[autodoc]] logging.enable_progress_bar
[[autodoc]] logging.disable_progress_bar
| huggingface/transformers/blob/main/docs/source/en/main_classes/logging.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Callbacks
Callbacks are objects that can customize the behavior of the training loop in the PyTorch
[`Trainer`] (this feature is not yet implemented in TensorFlow) that can inspect the training loop
state (for progress reporting, logging on TensorBoard or other ML platforms...) and take decisions (like early
stopping).
Callbacks are "read only" pieces of code, apart from the [`TrainerControl`] object they return, they
cannot change anything in the training loop. For customizations that require changes in the training loop, you should
subclass [`Trainer`] and override the methods you need (see [trainer](trainer) for examples).
By default, `TrainingArguments.report_to` is set to `"all"`, so a [`Trainer`] will use the following callbacks.
- [`DefaultFlowCallback`] which handles the default behavior for logging, saving and evaluation.
- [`PrinterCallback`] or [`ProgressCallback`] to display progress and print the
logs (the first one is used if you deactivate tqdm through the [`TrainingArguments`], otherwise
it's the second one).
- [`~integrations.TensorBoardCallback`] if tensorboard is accessible (either through PyTorch >= 1.4
or tensorboardX).
- [`~integrations.WandbCallback`] if [wandb](https://www.wandb.com/) is installed.
- [`~integrations.CometCallback`] if [comet_ml](https://www.comet.ml/site/) is installed.
- [`~integrations.MLflowCallback`] if [mlflow](https://www.mlflow.org/) is installed.
- [`~integrations.NeptuneCallback`] if [neptune](https://neptune.ai/) is installed.
- [`~integrations.AzureMLCallback`] if [azureml-sdk](https://pypi.org/project/azureml-sdk/) is
installed.
- [`~integrations.CodeCarbonCallback`] if [codecarbon](https://pypi.org/project/codecarbon/) is
installed.
- [`~integrations.ClearMLCallback`] if [clearml](https://github.com/allegroai/clearml) is installed.
- [`~integrations.DagsHubCallback`] if [dagshub](https://dagshub.com/) is installed.
- [`~integrations.FlyteCallback`] if [flyte](https://flyte.org/) is installed.
- [`~integrations.DVCLiveCallback`] if [dvclive](https://dvc.org/doc/dvclive) is installed.
If a package is installed but you don't wish to use the accompanying integration, you can change `TrainingArguments.report_to` to a list of just those integrations you want to use (e.g. `["azure_ml", "wandb"]`).
The main class that implements callbacks is [`TrainerCallback`]. It gets the
[`TrainingArguments`] used to instantiate the [`Trainer`], can access that
Trainer's internal state via [`TrainerState`], and can take some actions on the training loop via
[`TrainerControl`].
## Available Callbacks
Here is the list of the available [`TrainerCallback`] in the library:
[[autodoc]] integrations.CometCallback
- setup
[[autodoc]] DefaultFlowCallback
[[autodoc]] PrinterCallback
[[autodoc]] ProgressCallback
[[autodoc]] EarlyStoppingCallback
[[autodoc]] integrations.TensorBoardCallback
[[autodoc]] integrations.WandbCallback
- setup
[[autodoc]] integrations.MLflowCallback
- setup
[[autodoc]] integrations.AzureMLCallback
[[autodoc]] integrations.CodeCarbonCallback
[[autodoc]] integrations.NeptuneCallback
[[autodoc]] integrations.ClearMLCallback
[[autodoc]] integrations.DagsHubCallback
[[autodoc]] integrations.FlyteCallback
[[autodoc]] integrations.DVCLiveCallback
- setup
## TrainerCallback
[[autodoc]] TrainerCallback
Here is an example of how to register a custom callback with the PyTorch [`Trainer`]:
```python
class MyCallback(TrainerCallback):
"A callback that prints a message at the beginning of training"
def on_train_begin(self, args, state, control, **kwargs):
print("Starting training")
trainer = Trainer(
model,
args,
train_dataset=train_dataset,
eval_dataset=eval_dataset,
callbacks=[MyCallback], # We can either pass the callback class this way or an instance of it (MyCallback())
)
```
Another way to register a callback is to call `trainer.add_callback()` as follows:
```python
trainer = Trainer(...)
trainer.add_callback(MyCallback)
# Alternatively, we can pass an instance of the callback class
trainer.add_callback(MyCallback())
```
## TrainerState
[[autodoc]] TrainerState
## TrainerControl
[[autodoc]] TrainerControl
| huggingface/transformers/blob/main/docs/source/en/main_classes/callback.md |
ow to add BigBird to 🤗 Transformers?
=====================================
Mentor: [Patrick](https://github.com/patrickvonplaten)
Begin: 12.02.2020
Estimated End: 19.03.2020
Contributor: [Vasudev](https://github.com/thevasudevgupta)
Adding a new model is often difficult and requires an in-depth knowledge
of the 🤗 Transformers library and ideally also of the model's original
repository. At Hugging Face, we are trying to empower the community more
and more to add models independently.
The following sections explain in detail how to add BigBird
to Transformers. You will work closely with Patrick to
integrate BigBird into Transformers. By doing so, you will both gain a
theoretical and deep practical understanding of BigBird.
But more importantly, you will have made a major
open-source contribution to Transformers. Along the way, you will:
- get insights into open-source best practices
- understand the design principles of one of the most popular NLP
libraries
- learn how to do efficiently test large NLP models
- learn how to integrate Python utilities like `black`, `ruff`,
`make fix-copies` into a library to always ensure clean and readable
code
To start, let's try to get a general overview of the Transformers
library.
General overview of 🤗 Transformers
----------------------------------
First, you should get a general overview of 🤗 Transformers. Transformers
is a very opinionated library, so there is a chance that
you don't agree with some of the library's philosophies or design
choices. From our experience, however, we found that the fundamental
design choices and philosophies of the library are crucial to
efficiently scale Transformers while keeping maintenance costs at a
reasonable level.
A good first starting point to better understand the library is to read
the [documentation of our philosophy](https://huggingface.co/transformers/philosophy.html).
As a result of our way of working, there are some choices that we try to apply to all models:
- Composition is generally favored over abstraction
- Duplicating code is not always bad if it strongly improves the
readability or accessibility of a model
- Model files are as self-contained as possible so that when you read
the code of a specific model, you ideally only have to look into the
respective `modeling_....py` file.
In our opinion, the library's code is not just a means to provide a
product, *e.g.*, the ability to use BERT for inference, but also as the
very product that we want to improve. Hence, when adding a model, the
user is not only the person that will use your model, but also everybody
that will read, try to understand, and possibly tweak your code.
With this in mind, let's go a bit deeper into the general library
design.
### Overview of models
To successfully add a model, it is important to understand the
interaction between your model and its config,
`PreTrainedModel`, and `PretrainedConfig`. For
exemplary purposes, we will call the PyTorch model to be added to 🤗 Transformers
`BrandNewBert`.
Let's take a look:
As you can see, we do make use of inheritance in 🤗 Transformers, but we
keep the level of abstraction to an absolute minimum. There are never
more than two levels of abstraction for any model in the library.
`BrandNewBertModel` inherits from
`BrandNewBertPreTrainedModel` which in
turn inherits from `PreTrainedModel` and that's it.
As a general rule, we want to make sure
that a new model only depends on `PreTrainedModel`. The
important functionalities that are automatically provided to every new
model are
`PreTrainedModel.from_pretrained` and `PreTrainedModel.save_pretrained`, which are
used for serialization and deserialization. All
of the other important functionalities, such as
`BrandNewBertModel.forward` should be
completely defined in the new `modeling_brand_new_bert.py` module. Next,
we want to make sure that a model with a specific head layer, such as
`BrandNewBertForMaskedLM` does not inherit
from `BrandNewBertModel`, but rather uses
`BrandNewBertModel` as a component that
can be called in its forward pass to keep the level of abstraction low.
Every new model requires a configuration class, called
`BrandNewBertConfig`. This configuration
is always stored as an attribute in
`PreTrainedModel`, and
thus can be accessed via the `config` attribute for all classes
inheriting from `BrandNewBertPreTrainedModel`
```python
# assuming that `brand_new_bert` belongs to the organization `brandy`
model = BrandNewBertModel.from_pretrained("brandy/brand_new_bert")
model.config # model has access to its config
```
Similar to the model, the configuration inherits basic serialization and
deserialization functionalities from
`PretrainedConfig`. Note
that the configuration and the model are always serialized into two
different formats - the model to a `pytorch_model.bin` file
and the configuration to a `config.json` file. Calling
`PreTrainedModel.save_pretrained` will automatically call
`PretrainedConfig.save_pretrained`, so that both model and configuration are saved.
### Overview of tokenizers
Not quite ready yet :-( This section will be added soon!
Step-by-step recipe to add a model to 🤗 Transformers
----------------------------------------------------
Everyone has different preferences of how to port a model so it can be
very helpful for you to take a look at summaries of how other
contributors ported models to Hugging Face. Here is a list of community
blog posts on how to port a model:
1. [Porting GPT2
Model](https://medium.com/huggingface/from-tensorflow-to-pytorch-265f40ef2a28)
by [Thomas](https://huggingface.co/thomwolf)
2. [Porting WMT19 MT Model](https://huggingface.co/blog/porting-fsmt)
by [Stas](https://huggingface.co/stas)
From experience, we can tell you that the most important things to keep
in mind when adding a model are:
- Don't reinvent the wheel! Most parts of the code you will add for
the new 🤗 Transformers model already exist somewhere in 🤗
Transformers. Take some time to find similar, already existing
models and tokenizers you can copy from.
[grep](https://www.gnu.org/software/grep/) and
[rg](https://github.com/BurntSushi/ripgrep) are your friends. Note
that it might very well happen that your model's tokenizer is based
on one model implementation, and your model's modeling code on
another one. *E.g.*, FSMT's modeling code is based on BART, while
FSMT's tokenizer code is based on XLM.
- It's more of an engineering challenge than a scientific challenge.
You should spend more time on creating an efficient debugging
environment than trying to understand all theoretical aspects of the
model in the paper.
- Ask for help when you're stuck! Models are the core component of 🤗
Transformers so we, at Hugging Face, are more than happy to help
you at every step to add your model. Don't hesitate to ask if you
notice you are not making progress.
In the following, we try to give you a general recipe that we found most
useful when porting a model to 🤗 Transformers.
The following list is a summary of everything that has to be done to add
a model and can be used by you as a To-Do List:
1. [ ] (Optional) Understood theoretical aspects
2. [ ] Prepared transformers dev environment
3. [ ] Set up debugging environment of the original repository
4. [ ] Created script that successfully runs forward pass using
original repository and checkpoint
5. [ ] Successfully opened a PR and added the model skeleton to Transformers
6. [ ] Successfully converted original checkpoint to Transformers
checkpoint
7. [ ] Successfully ran forward pass in Transformers that gives
identical output to original checkpoint
8. [ ] Finished model tests in Transformers
9. [ ] Successfully added Tokenizer in Transformers
10. [ ] Run end-to-end integration tests
11. [ ] Finished docs
12. [ ] Uploaded model weights to the hub
13. [ ] Submitted the pull request for review
14. [ ] (Optional) Added a demo notebook
To begin with, we usually recommend to start by getting a good
theoretical understanding of `BigBird`. However, if you prefer to
understand the theoretical aspects of the model *on-the-job*, then it is
totally fine to directly dive into the `BigBird`'s code-base. This
option might suit you better, if your engineering skills are better than
your theoretical skill, if you have trouble understanding
`BigBird`'s paper, or if you just enjoy programming much more than
reading scientific papers.
### 1. (Optional) Theoretical aspects of BigBird
You should take some time to read *BigBird's* paper, if such
descriptive work exists. There might be large sections of the paper that
are difficult to understand. If this is the case, this is fine - don't
worry! The goal is not to get a deep theoretical understanding of the
paper, but to extract the necessary information required to effectively
re-implement the model in 🤗 Transformers. That being said, you don't
have to spend too much time on the theoretical aspects, but rather focus
on the practical ones, namely:
- What type of model is *BigBird*? BERT-like encoder-only
model? GPT2-like decoder-only model? BART-like encoder-decoder
model? Look at the `model_summary` if
you're not familiar with the differences between those.
- What are the applications of *BigBird*? Text
classification? Text generation? Seq2Seq tasks, *e.g.,*
summarization?
- What is the novel feature of the model making it different from
BERT/GPT-2/BART?
- Which of the already existing [🤗 Transformers
models](https://huggingface.co/transformers/#contents) is most
similar to *BigBird*?
- What type of tokenizer is used? A sentencepiece tokenizer? Word
piece tokenizer? Is it the same tokenizer as used for BERT or BART?
After you feel like you have gotten a good overview of the architecture
of the model, you might want to write to Patrick with any
questions you might have. This might include questions regarding the
model's architecture, its attention layer, etc. We will be more than
happy to help you.
#### Additional resources
Before diving into the code, here are some additional resources that might be worth taking a look at:
- [Yannic Kilcher's paper summary](https://www.youtube.com/watch?v=WVPE62Gk3EM&ab_channel=YannicKilcher)
- [Yannic Kilcher's summary of Longformer](https://www.youtube.com/watch?v=_8KNb5iqblE&ab_channel=YannicKilcher) - Longformer and BigBird are **very** similar models. Since Longformer has already been ported to 🤗 Transformers, it is useful to understand the differences between the two models
- [Blog post](https://medium.com/dsc-msit/is-google-bigbird-gonna-be-the-new-leader-in-nlp-domain-8c95cecc30f8) - A relatively superficial blog post about BigBird. Might be a good starting point to understand BigBird
#### Make sure you've understood the fundamental aspects of BigBird
Alright, now you should be ready to take a closer look into the actual code of BigBird.
You should have understood the following aspects of BigBird by now:
- BigBird provides a new attention layer for long-range sequence modelling that can be used
as a drop-in replacement for already existing architectures. This means that every transformer-based model architecture can replace its [Self-attention layer](https://towardsdatascience.com/illustrated-self-attention-2d627e33b20a) with BigBird's self-attention layer.
- BigBird's self-attention layer is composed of three mechanisms: block sparse (local) self-attention, global self-attention, random self-attention
- BigBird's block sparse (local) self-attention is different from Longformer's local self-attention. How so? Why does that matter? => Can be deployed on TPU much easier this way
- BigBird can be implemented for both an encoder-only model **and**
for an encoder-decoder model, which means that we can reuse lots of [code from RoBERTa](https://github.com/huggingface/transformers/blob/main/src/transformers/models/roberta/modeling_roberta.py) and [from PEGASUS](https://github.com/huggingface/transformers/blob/main/src/transformers/models/pegasus/modeling_pegasus.py) at a later stage.
If any of the mentioned aspects above are **not** clear to you, now is a great time to talk to Patrick.
### 2. Next prepare your environment
1. Fork the [repository](https://github.com/huggingface/transformers)
by clicking on the 'Fork' button on the repository's page. This
creates a copy of the code under your GitHub user account.
2. Clone your `transformers` fork to your local disk, and add the base
repository as a remote:
```bash
git clone https://github.com/[your Github handle]/transformers.git
cd transformers
git remote add upstream https://github.com/huggingface/transformers.git
```
3. Set up a development environment, for instance by running the
following command:
```bash
python -m venv .env
source .env/bin/activate
pip install -e ".[dev]"
```
and return to the parent directory
```bash
cd ..
```
4. We recommend adding the PyTorch version of *BigBird* to
Transformers. To install PyTorch, please follow the instructions [here](https://pytorch.org/get-started/locally/).
**Note:** You don't need to have CUDA installed. Making the new model
work on CPU is sufficient.
5. To port *BigBird*, you will also need access to its
original repository:
```bash
git clone https://github.com/google-research/bigbird.git
cd big_bird
pip install -e .
```
Now you have set up a development environment to port *BigBird*
to 🤗 Transformers.
### Run a pretrained checkpoint using the original repository
**3. Set up debugging environment**
At first, you will work on the original *BigBird* repository.
Often, the original implementation is very "researchy". Meaning that
documentation might be lacking and the code can be difficult to
understand. But this should be exactly your motivation to reimplement
*BigBird*. At Hugging Face, one of our main goals is to *make
people stand on the shoulders of giants* which translates here very well
into taking a working model and rewriting it to make it as **accessible,
user-friendly, and beautiful** as possible. This is the number-one
motivation to re-implement models into 🤗 Transformers - trying to make
complex new NLP technology accessible to **everybody**.
You should start thereby by diving into the [original repository](https://github.com/google-research/bigbird).
Successfully running the official pretrained model in the original
repository is often **the most difficult** step. From our experience, it
is very important to spend some time getting familiar with the original
code-base. You need to figure out the following:
- Where to find the pretrained weights?
- How to load the pretrained weights into the corresponding model?
- How to run the tokenizer independently from the model?
- Trace one forward pass so that you know which classes and functions
are required for a simple forward pass. Usually, you only have to
reimplement those functions.
- Be able to locate the important components of the model: Where is
the model's class? Are there model sub-classes, *e.g.*,
EncoderModel, DecoderModel? Where is the self-attention layer? Are
there multiple different attention layers, *e.g.*, *self-attention*,
*cross-attention*...?
- How can you debug the model in the original environment of the repo?
Do you have to add `print` statements, can you work with
an interactive debugger like [ipdb](https://pypi.org/project/ipdb/), or should you use
an efficient IDE to debug the model, like PyCharm?
It is very important that before you start the porting process, that you
can **efficiently** debug code in the original repository! Also,
remember that you are working with an open-source library, so do not
hesitate to open an issue, or even a pull request in the original
repository. The maintainers of this repository are most likely very
happy about someone looking into their code!
At this point, it is really up to you which debugging environment and
strategy you prefer to use to debug the original model. We strongly
advise against setting up a costly GPU environment, but simply work on a
CPU both when starting to dive into the original repository and also
when starting to write the 🤗 Transformers implementation of the model.
Only at the very end, when the model has already been successfully
ported to 🤗 Transformers, one should verify that the model also works as
expected on GPU.
In general, there are two possible debugging environments for running
the original model
- [Jupyter notebooks](https://jupyter.org/) / [google colab](https://colab.research.google.com/notebooks/intro.ipynb)
- Local python scripts.
Jupyter notebooks have the advantage that they allow for cell-by-cell
execution which can be helpful to better split logical components from
one another and to have faster debugging cycles as intermediate results
can be stored. Also, notebooks are often easier to share with other
contributors, which might be very helpful if you want to ask the Hugging
Face team for help. If you are familiar with Jupyter notebooks, we
strongly recommend you to work with them.
The obvious disadvantage of Jupyter notebooks is that if you are not
used to working with them you will have to spend some time adjusting to
the new programming environment and that you might not be able to use
your known debugging tools anymore, like `ipdb`.
**4. Successfully run forward pass**
For each code-base, a good first step is always to load a **small**
pretrained checkpoint and to be able to reproduce a single forward pass
using a dummy integer vector of input IDs as an input. Such a script
could look something like this:
```python
from bigbird.core import modeling
model = modeling.BertModel(bert_config)
from bigbird.core import utils
params = utils.BigBirdConfig(vocab_size=32000, hidden_size=512,
num_hidden_layers=8, num_attention_heads=6, intermediate_size=1024)
ckpt_path = 'gs://bigbird-transformer/pretrain/bigbr_base/model.ckpt-0'
ckpt_reader = tf.compat.v1.train.NewCheckpointReader(ckpt_path)
model.set_weights([ckpt_reader.get_tensor(v.name[:-2]) for v in tqdm(model.trainable_weights, position=0)])
input_ids = tf.constant([[31, 51, 99], [15, 5, 0]])
_, pooled_output = model(input_ids=input_ids, token_type_ids=token_type_ids)
...
```
Next, regarding the debugging strategy, there are generally a few from
which to choose from:
- Decompose the original model into many small testable components and
run a forward pass on each of those for verification
- Decompose the original model only into the original *tokenizer* and
the original *model*, run a forward pass on those, and use
intermediate print statements or breakpoints for verification
Again, it is up to you which strategy to choose. Often, one or the other
is advantageous depending on the original code base.
If the original code-base allows you to decompose the model into smaller
sub-components, *e.g.*, if the original code-base can easily be run in
eager mode, it is usually worth the effort to do so. There are some
important advantages to taking the more difficult road in the beginning:
- at a later stage when comparing the original model to the Hugging
Face implementation, you can verify automatically for each component
individually that the corresponding component of the 🤗 Transformers
implementation matches instead of relying on visual comparison via
print statements
- it can give you some rope to decompose the big problem of porting a
model into smaller problems of just porting individual components
and thus structure your work better
- separating the model into logical meaningful components will help
you to get a better overview of the model's design and thus to
better understand the model
- at a later stage those component-by-component tests help you to
ensure that no regression occurs as you continue changing your code
[Lysandre's](https://gist.github.com/LysandreJik/db4c948f6b4483960de5cbac598ad4ed)
integration checks for ELECTRA gives a nice example of how this can be
done.
However, if the original code-base is very complex or only allows
intermediate components to be run in a compiled mode, it might be too
time-consuming or even impossible to separate the model into smaller
testable sub-components. A good example is [T5's
MeshTensorFlow](https://github.com/tensorflow/mesh/tree/master/mesh_tensorflow)
library which is very complex and does not offer a simple way to
decompose the model into its sub-components. For such libraries, one
often relies on verifying print statements.
No matter which strategy you choose, the recommended procedure is often
the same in that you should start to debug the starting layers first and
the ending layers last.
It is recommended that you retrieve the output, either by print
statements or sub-component functions, of the following layers in the
following order:
1. Retrieve the input IDs passed to the model
2. Retrieve the word embeddings
3. Retrieve the input of the first Transformer layer
4. Retrieve the output of the first Transformer layer
5. Retrieve the output of the following n - 1 Transformer layers
6. Retrieve the output of the whole BigBird Model
Input IDs should thereby consists of an array of integers, *e.g.*,
`input_ids = [0, 4, 4, 3, 2, 4, 1, 7, 19]`
The outputs of the following layers often consist of multi-dimensional
float arrays and can look like this:
```bash
[[
[-0.1465, -0.6501, 0.1993, ..., 0.1451, 0.3430, 0.6024],
[-0.4417, -0.5920, 0.3450, ..., -0.3062, 0.6182, 0.7132],
[-0.5009, -0.7122, 0.4548, ..., -0.3662, 0.6091, 0.7648],
...,
[-0.5613, -0.6332, 0.4324, ..., -0.3792, 0.7372, 0.9288],
[-0.5416, -0.6345, 0.4180, ..., -0.3564, 0.6992, 0.9191],
[-0.5334, -0.6403, 0.4271, ..., -0.3339, 0.6533, 0.8694]]],
```
We expect that every model added to 🤗 Transformers passes a couple of
integration tests, meaning that the original model and the reimplemented
version in 🤗 Transformers have to give the exact same output up to a
precision of 0.001! Since it is normal that the exact same model written
in different libraries can give a slightly different output depending on
the library framework, we accept an error tolerance of 1e-3 (0.001). It
is not enough if the model gives nearly the same output, they have to be
the almost identical. Therefore, you will certainly compare the
intermediate outputs of the 🤗 Transformers version multiple times
against the intermediate outputs of the original implementation of
*BigBird* in which case an **efficient** debugging environment
of the original repository is absolutely important. Here is some advice
to make your debugging environment as efficient as possible.
- Find the best way of debugging intermediate results. Is the original
repository written in PyTorch? Then you should probably take the
time to write a longer script that decomposes the original model
into smaller sub-components to retrieve intermediate values. Is the
original repository written in Tensorflow 1? Then you might have to
rely on TensorFlow print operations like
[tf.print](https://www.tensorflow.org/api_docs/python/tf/print) to
output intermediate values. Is the original repository written in
Jax? Then make sure that the model is **not jitted** when running
the forward pass, *e.g.*, check-out [this
link](https://github.com/google/jax/issues/196).
- Use the smallest pretrained checkpoint you can find. The smaller the
checkpoint, the faster your debug cycle becomes. It is not efficient
if your pretrained model is so big that your forward pass takes more
than 10 seconds. In case only very large checkpoints are available,
it might make more sense to create a dummy model in the new
environment with randomly initialized weights and save those weights
for comparison with the 🤗 Transformers version of your model
- Make sure you are using the easiest way of calling a forward pass in
the original repository. Ideally, you want to find the function in
the original repository that **only** calls a single forward pass,
*i.e.* that is often called `predict`, `evaluate`, `forward` or
`__call__`. You don't want to debug a function that calls `forward`
multiple times, *e.g.*, to generate text, like
`autoregressive_sample`, `generate`.
- Try to separate the tokenization from the model's
forward pass. If the original repository shows
examples where you have to input a string, then try to find out
where in the forward call the string input is changed to input ids
and start from this point. This might mean that you have to possibly
write a small script yourself or change the original code so that
you can directly input the ids instead of an input string.
- Make sure that the model in your debugging setup is **not** in
training mode, which often causes the model to yield random outputs
due to multiple dropout layers in the model. Make sure that the
forward pass in your debugging environment is **deterministic** so
that the dropout layers are not used. Or use
`transformers.utils.set_seed` if the old and new
implementations are in the same framework.
#### (Important) More details on how to create a debugging environment for BigBird
- BigBird has multiple pretrained checkpoints that should eventually all be ported to
🤗 Transformers. The pretrained checkpoints can be found [here](https://console.cloud.google.com/storage/browser/bigbird-transformer/pretrain;tab=objects?prefix=&forceOnObjectsSortingFiltering=false).
Those checkpoints include both pretrained weights for encoder-only (BERT/RoBERTa) under the folder `bigbr_base` and encoder-decoder (PEGASUS) under the folder `bigbp_large`.
You should start by porting the `bigbr_base` model. The encoder-decoder model
can be ported afterward.
for an encoder-decoder architecture as well as an encoder-only architecture.
- BigBird was written in tf.compat meaning that a mixture of a TensorFlow 1 and
TensorFlow 2 API was used.
- The most important part of the BigBird code-base is [bigbird.bigbird.core](https://github.com/google-research/bigbird/tree/master/bigbird/core) which includes all logic necessary
to implement BigBird.
- The first goal should be to successfully run a forward pass using the RoBERTa checkpoint `bigbr_base/model.ckpt-0.data-00000-of-00001` and `bigbr_base/model.ckpt-0.index`.
### Port BigBird to 🤗 Transformers
Next, you can finally start adding new code to 🤗 Transformers. Go into
the clone of your 🤗 Transformers' fork:
cd transformers
In the special case that you are adding a model whose architecture
exactly matches the model architecture of an existing model you only
have to add a conversion script as described in [this
section](#write-a-conversion-script). In this case, you can just re-use
the whole model architecture of the already existing model.
Otherwise, let's start generating a new model with the amazing
Cookiecutter!
**Use the Cookiecutter to automatically generate the model's code**
To begin with head over to the [🤗 Transformers
templates](https://github.com/huggingface/transformers/tree/main/templates/adding_a_new_model)
to make use of our `cookiecutter` implementation to automatically
generate all the relevant files for your model. Again, we recommend only
adding the PyTorch version of the model at first. Make sure you follow
the instructions of the `README.md` on the [🤗 Transformers
templates](https://github.com/huggingface/transformers/tree/main/templates/adding_a_new_model)
carefully.
Since you will first implement the Encoder-only/RoBERTa-like version of BigBird you should
select the `is_encoder_decoder_model = False` option in the cookiecutter. Also, it is recommended
that you implement the model only in PyTorch in the beginning and select "Standalone" as the
tokenizer type for now.
**Open a Pull Request on the main huggingface/transformers repo**
Before starting to adapt the automatically generated code, now is the
time to open a "Work in progress (WIP)" pull request, *e.g.*, "\[WIP\]
Add *BigBird*", in 🤗 Transformers so that you and the Hugging
Face team can work side-by-side on integrating the model into 🤗
Transformers.
You should do the following:
1. Create a branch with a descriptive name from your main branch
```
git checkout -b add_big_bird
```
2. Commit the automatically generated code:
```
git add .
git commit
```
3. Fetch and rebase to current main
```
git fetch upstream
git rebase upstream/main
```
4. Push the changes to your account using:
```
git push -u origin a-descriptive-name-for-my-changes
```
5. Once you are satisfied, go to the webpage of your fork on GitHub.
Click on "Pull request". Make sure to add the GitHub handle of Patrick
as one reviewer, so that the Hugging Face team gets notified for future changes.
6. Change the PR into a draft by clicking on "Convert to draft" on the
right of the GitHub pull request web page.
In the following, whenever you have done some progress, don't forget to
commit your work and push it to your account so that it shows in the
pull request. Additionally, you should make sure to update your work
with the current main from time to time by doing:
git fetch upstream
git merge upstream/main
In general, all questions you might have regarding the model or your
implementation should be asked in your PR and discussed/solved in the
PR. This way, Patrick will always be notified when you are
committing new code or if you have a question. It is often very helpful
to point Patrick to your added code so that the Hugging
Face team can efficiently understand your problem or question.
To do so, you can go to the "Files changed" tab where you see all of
your changes, go to a line regarding which you want to ask a question,
and click on the "+" symbol to add a comment. Whenever a question or
problem has been solved, you can click on the "Resolve" button of the
created comment.
In the same way, Patrick will open comments when reviewing
your code. We recommend asking most questions on GitHub on your PR. For
some very general questions that are not very useful for the public,
feel free to ping Patrick by Slack or email.
**5. Adapt the generated models code for BigBird**
At first, we will focus only on the model itself and not care about the
tokenizer. All the relevant code should be found in the generated files
`src/transformers/models/big_bird/modeling_big_bird.py` and
`src/transformers/models/big_bird/configuration_big_bird.py`.
Now you can finally start coding :). The generated code in
`src/transformers/models/big_bird/modeling_big_bird.py` will
either have the same architecture as BERT if it's an encoder-only model
or BART if it's an encoder-decoder model. At this point, you should
remind yourself what you've learned in the beginning about the
theoretical aspects of the model: *How is the model different from BERT
or BART?*\". Implement those changes which often means to change the
*self-attention* layer, the order of the normalization layer, etc...
Again, it is often useful to look at the similar architecture of already
existing models in Transformers to get a better feeling of how your
model should be implemented.
**Note** that at this point, you don't have to be very sure that your
code is fully correct or clean. Rather, it is advised to add a first
*unclean*, copy-pasted version of the original code to
`src/transformers/models/big_bird/modeling_big_bird.py`
until you feel like all the necessary code is added. From our
experience, it is much more efficient to quickly add a first version of
the required code and improve/correct the code iteratively with the
conversion script as described in the next section. The only thing that
has to work at this point is that you can instantiate the 🤗 Transformers
implementation of *BigBird*, *i.e.* the following command
should work:
```python
from transformers import BigBirdModel, BigBirdConfig
model = BigBirdModel(BigBirdConfig())
```
The above command will create a model according to the default
parameters as defined in `BigBirdConfig()` with random weights,
thus making sure that the `init()` methods of all components works.
Note that for BigBird you have to change the attention layer. BigBird's attention
layer is quite complex as you can see [here](https://github.com/google-research/bigbird/blob/103a3345f94bf6364749b51189ed93024ca5ef26/bigbird/core/attention.py#L560). Don't
feel discouraged by this! In a first step you should simply make sure that
the layer `BigBirdAttention` has the correct weights as can be found in the
pretrained checkpoints. This means that you have to make sure that in the
`__init__(self, ...)` function of `BigBirdAttention`, all submodules include all
necessary `nn.Module` layers. Only at a later stage do we need to fully rewrite
the complex attention function.
**6. Write a conversion script**
Next, you should write a conversion script that lets you convert the
checkpoint you used to debug *BigBird* in the original
repository to a checkpoint compatible with your just created 🤗
Transformers implementation of *BigBird*. It is not advised to
write the conversion script from scratch, but rather to look through
already existing conversion scripts in 🤗 Transformers for one that has
been used to convert a similar model that was written in the same
framework as *BigBird*. Usually, it is enough to copy an
already existing conversion script and slightly adapt it for your use
case. Don't hesitate to ask Patrick to point you to a
similar already existing conversion script for your model.
- A good starting point to convert the original TF BigBird implementation to the PT Hugging Face implementation is probably BERT's conversion script
[here](https://github.com/huggingface/transformers/blob/7acfa95afb8194f8f9c1f4d2c6028224dbed35a2/src/transformers/models/bert/modeling_bert.py#L91)
You can copy paste the conversion function into `modeling_big_bird.py` and then adapt it
to your needs.
In the following, we'll quickly explain how PyTorch models store layer
weights and define layer names. In PyTorch, the name of a layer is
defined by the name of the class attribute you give the layer. Let's
define a dummy model in PyTorch, called `SimpleModel` as follows:
```python
from torch import nn
class SimpleModel(nn.Module):
def __init__(self):
super().__init__()
self.dense = nn.Linear(10, 10)
self.intermediate = nn.Linear(10, 10)
self.layer_norm = nn.LayerNorm(10)
```
Now we can create an instance of this model definition which will fill
all weights: `dense`, `intermediate`, `layer_norm` with random weights.
We can print the model to see its architecture
```python
model = SimpleModel()
print(model)
```
This will print out the following:
```bash
SimpleModel(
(dense): Linear(in_features=10, out_features=10, bias=True)
(intermediate): Linear(in_features=10, out_features=10, bias=True)
(layer_norm): LayerNorm((10,), eps=1e-05, elementwise_affine=True)
)
```
We can see that the layer names are defined by the name of the class
attribute in PyTorch. You can print out the weight values of a specific
layer:
```python
print(model.dense.weight.data)
```
to see that the weights were randomly initialized
```bash
tensor([[-0.0818, 0.2207, -0.0749, -0.0030, 0.0045, -0.1569, -0.1598, 0.0212,
-0.2077, 0.2157],
[ 0.1044, 0.0201, 0.0990, 0.2482, 0.3116, 0.2509, 0.2866, -0.2190,
0.2166, -0.0212],
[-0.2000, 0.1107, -0.1999, -0.3119, 0.1559, 0.0993, 0.1776, -0.1950,
-0.1023, -0.0447],
[-0.0888, -0.1092, 0.2281, 0.0336, 0.1817, -0.0115, 0.2096, 0.1415,
-0.1876, -0.2467],
[ 0.2208, -0.2352, -0.1426, -0.2636, -0.2889, -0.2061, -0.2849, -0.0465,
0.2577, 0.0402],
[ 0.1502, 0.2465, 0.2566, 0.0693, 0.2352, -0.0530, 0.1859, -0.0604,
0.2132, 0.1680],
[ 0.1733, -0.2407, -0.1721, 0.1484, 0.0358, -0.0633, -0.0721, -0.0090,
0.2707, -0.2509],
[-0.1173, 0.1561, 0.2945, 0.0595, -0.1996, 0.2988, -0.0802, 0.0407,
0.1829, -0.1568],
[-0.1164, -0.2228, -0.0403, 0.0428, 0.1339, 0.0047, 0.1967, 0.2923,
0.0333, -0.0536],
[-0.1492, -0.1616, 0.1057, 0.1950, -0.2807, -0.2710, -0.1586, 0.0739,
0.2220, 0.2358]]).
```
In the conversion script, you should fill those randomly initialized
weights with the exact weights of the corresponding layer in the
checkpoint. *E.g.*,
```python
# retrieve matching layer weights, e.g. by
# recursive algorithm
layer_name = "dense"
pretrained_weight = array_of_dense_layer
model_pointer = getattr(model, "dense")
model_pointer.weight.data = torch.from_numpy(pretrained_weight)
```
While doing so, you must verify that each randomly initialized weight of
your PyTorch model and its corresponding pretrained checkpoint weight
exactly match in both **shape and name**. To do so, it is **necessary**
to add assert statements for the shape and print out the names of the
checkpoints weights. *E.g.*, you should add statements like:
```python
assert (
model_pointer.weight.shape == pretrained_weight.shape
), f"Pointer shape of random weight {model_pointer.shape} and array shape of checkpoint weight {pretrained_weight.shape} mismatched"
```
Besides, you should also print out the names of both weights to make
sure they match, *e.g.*,
```python
logger.info(f"Initialize PyTorch weight {layer_name} from {pretrained_weight.name}")
```
If either the shape or the name doesn't match, you probably assigned
the wrong checkpoint weight to a randomly initialized layer of the 🤗
Transformers implementation.
An incorrect shape is most likely due to an incorrect setting of the
config parameters in `BigBirdConfig()` that do not exactly match
those that were used for the checkpoint you want to convert. However, it
could also be that PyTorch's implementation of a layer requires the
weight to be transposed beforehand.
Finally, you should also check that **all** required weights are
initialized and print out all checkpoint weights that were not used for
initialization to make sure the model is correctly converted. It is
completely normal, that the conversion trials fail with either a wrong
shape statement or wrong name assignment. This is most likely because
either you used incorrect parameters in `BigBirdConfig()`, have a
wrong architecture in the 🤗 Transformers implementation, you have a bug
in the `init()` functions of one of the components of the 🤗 Transformers
implementation or you need to transpose one of the checkpoint weights.
This step should be iterated with the previous step until all weights of
the checkpoint are correctly loaded in the Transformers model. Having
correctly loaded the checkpoint into the 🤗 Transformers implementation,
you can then save the model under a folder of your choice
`/path/to/converted/checkpoint/folder` that should then contain both a
`pytorch_model.bin` file and a `config.json` file:
```python
model.save_pretrained("/path/to/converted/checkpoint/folder")
```
**7. Implement the forward pass**
Having managed to correctly load the pretrained weights into the 🤗
Transformers implementation, you should now make sure that the forward
pass is correctly implemented. In [Get familiar with the original
repository](#run-a-pretrained-checkpoint-using-the-original-repository),
you have already created a script that runs a forward pass of the model
using the original repository. Now you should write an analogous script
using the 🤗 Transformers implementation instead of the original one. It
should look as follows:
[Here the model name might have to be adapted, *e.g.*, maybe BigBirdForConditionalGeneration instead of BigBirdModel]
```python
model = BigBirdModel.from_pretrained("/path/to/converted/checkpoint/folder")
input_ids = [0, 4, 4, 3, 2, 4, 1, 7, 19]
output = model(input_ids).last_hidden_states
```
It is very likely that the 🤗 Transformers implementation and the
original model implementation don't give the exact same output the very
first time or that the forward pass throws an error. Don't be
disappointed - it's expected! First, you should make sure that the
forward pass doesn't throw any errors. It often happens that the wrong
dimensions are used leading to a `"Dimensionality mismatch"`
error or that the wrong data type object is used, *e.g.*, `torch.long`
instead of `torch.float32`. Don't hesitate to ask Patrick
for help, if you don't manage to solve certain errors.
The final part to make sure the 🤗 Transformers implementation works
correctly is to ensure that the outputs are equivalent to a precision of
`1e-3`. First, you should ensure that the output shapes are identical,
*i.e.* `outputs.shape` should yield the same value for the script of the
🤗 Transformers implementation and the original implementation. Next, you
should make sure that the output values are identical as well. This one
of the most difficult parts of adding a new model. Common mistakes why
the outputs are not identical are:
- Some layers were not added, *i.e.* an activation layer
was not added, or the residual connection was forgotten
- The word embedding matrix was not tied
- The wrong positional embeddings are used because the original
implementation uses on offset
- Dropout is applied during the forward pass. To fix this make sure
`model.training is False` and that no dropout layer is
falsely activated during the forward pass, *i.e.* pass
`self.training` to [PyTorch's functional
dropout](https://pytorch.org/docs/stable/nn.functional.html?highlight=dropout#torch.nn.functional.dropout)
The best way to fix the problem is usually to look at the forward pass
of the original implementation and the 🤗 Transformers implementation
side-by-side and check if there are any differences. Ideally, you should
debug/print out intermediate outputs of both implementations of the
forward pass to find the exact position in the network where the 🤗
Transformers implementation shows a different output than the original
implementation. First, make sure that the hard-coded `input_ids` in both
scripts are identical. Next, verify that the outputs of the first
transformation of the `input_ids` (usually the word embeddings) are
identical. And then work your way up to the very last layer of the
network. At some point, you will notice a difference between the two
implementations, which should point you to the bug in the 🤗 Transformers
implementation. From our experience, a simple and efficient way is to
add many print statements in both the original implementation and 🤗
Transformers implementation, at the same positions in the network
respectively, and to successively remove print statements showing the
same values for intermediate presentions.
When you're confident that both implementations yield the same output,
verifying the outputs with
`torch.allclose(original_output, output, atol=1e-3)`, you're done with
the most difficult part! Congratulations - the work left to be done
should be a cakewalk 😊.
**8. Adding all necessary model tests**
At this point, you have successfully added a new model. However, it is
very much possible that the model does not yet fully comply with the
required design. To make sure, the implementation is fully compatible
with 🤗 Transformers, all common tests should pass. The Cookiecutter
should have automatically added a test file for your model, probably
under the same `tests/test_modeling_big_bird.py`. Run this test
file to verify that all common tests pass:
```python
pytest tests/test_modeling_big_bird.py
```
Having fixed all common tests, it is now crucial to ensure that all the
nice work you have done is well tested, so that
- a) The community can easily understand your work by looking at
specific tests of *BigBird*
- b) Future changes to your model will not break any important
feature of the model.
At first, integration tests should be added. Those integration tests
essentially do the same as the debugging scripts you used earlier to
implement the model to 🤗 Transformers. A template of those model tests
is already added by the Cookiecutter, called
`BigBirdModelIntegrationTests` and only has to be filled out by
you. To ensure that those tests are passing, run
```python
RUN_SLOW=1 pytest -sv tests/test_modeling_big_bird.py::BigBirdModelIntegrationTests
```
**Note**: In case you are using Windows, you should replace `RUN_SLOW=1` with
`SET RUN_SLOW=1`
Second, all features that are special to *BigBird* should be
tested additionally in a separate test under
`BigBirdModelTester`/`BigBirdModelTest`. This part is often
forgotten but is extremely useful in two ways:
- It helps to transfer the knowledge you have acquired during the
model addition to the community by showing how the special features
of *BigBird* should work.
- Future contributors can quickly test changes to the model by running
those special tests.
BigBird has quite a complex attention layer, so it is very important
to add more tests verifying the all parts of BigBird's self-attention layer
works as expected. This means that there should be at least 3 additional tests:
- 1. Verify that the sparse attention works correctly
- 2. Verify that the global attention works correctly
- 3. Verify that the random attention works correctly
**9. Implement the tokenizer**
Next, we should add the tokenizer of *BigBird*. Usually, the
tokenizer is equivalent or very similar to an already existing tokenizer
of 🤗 Transformers.
In the case of BigBird you should be able to just rely on an already existing tokenizer.
If not mistaken, BigBird uses the same tokenizer that was used for `BertGenerationTokenizer`,
which is based on `sentencepiece`. So you should be able to just set the config parameter
`tokenizer_class` to `BertGenerationTokenizer` without having to implement any new tokenizer.
It is very important to find/extract the original tokenizer file and to
manage to load this file into the 🤗 Transformers' implementation of the
tokenizer.
For BigBird, the tokenizer (sentencepiece) files can be found [here](https://github.com/google-research/bigbird/blob/master/bigbird/vocab/gpt2.model), which you should be able to load
as easily as:
```python
from transformers import BertGenerationTokenizer
tokenizer = BertGenerationTokenizer("/path/to/gpt2.model/file")
```
To ensure that the tokenizer works correctly, it is recommended to first
create a script in the original repository that inputs a string and
returns the `input_ids`. It could look similar to this (in pseudo-code):
```bash
input_str = "This is a long example input string containing special characters .$?-, numbers 2872 234 12 and words."
model = BigBirdModel.load_pretrained_checkpoint("/path/to/checkpoint/")
input_ids = model.tokenize(input_str)
```
You might have to take a deeper look again into the original repository
to find the correct tokenizer function or you might even have to do
changes to your clone of the original repository to only output the
`input_ids`. Having written a functional tokenization script that uses
the original repository, an analogous script for 🤗 Transformers should
be created. It should look similar to this:
```python
from transformers import BertGenerationTokenizer
input_str = "This is a long example input string containing special characters .$?-, numbers 2872 234 12 and words."
tokenizer = BertGenerationTokenizer.from_pretrained("/path/big/bird/folder")
input_ids = tokenizer(input_str).input_ids
```
When both `input_ids` yield the same values, as a final step a tokenizer
test file should also be added.
Since BigBird is most likely fully based on `BertGenerationTokenizer`,
you should only add a couple of "slow" integration tests. However, in this
case you do **not** need to add any `BigBirdTokenizationTest`.
**10. Run End-to-end integration tests**
Having added the tokenizer, you should also add a couple of end-to-end
integration tests using both the model and the tokenizer to
`tests/test_modeling_big_bird.py` in 🤗 Transformers. Such a test
should show on a meaningful text-to-text sample that the 🤗 Transformers
implementation works as expected. A meaningful text-to-text sample can
include, *e.g.*, a source-to-target-translation pair, an
article-to-summary pair, a question-to-answer pair, etc... If none of
the ported checkpoints has been fine-tuned on a downstream task it is
enough to simply rely on the model tests. In a final step to ensure that
the model is fully functional, it is advised that you also run all tests
on GPU. It can happen that you forgot to add some `.to(self.device)`
statements to internal tensors of the model, which in such a test would
show in an error. In case you have no access to a GPU, the Hugging Face
team can take care of running those tests for you.
**11. Add Docstring**
Now, all the necessary functionality for *BigBird* is added -
you're almost done! The only thing left to add is a nice docstring and
a doc page. The Cookiecutter should have added a template file called
`docs/source/model_doc/big_bird.rst` that you should fill out.
Users of your model will usually first look at this page before using
your model. Hence, the documentation must be understandable and concise.
It is very useful for the community to add some *Tips* to show how the
model should be used. Don't hesitate to ping Patrick
regarding the docstrings.
Next, make sure that the docstring added to
`src/transformers/models/big_bird/modeling_big_bird.py` is
correct and included all necessary inputs and outputs. It is always to
good to remind oneself that documentation should be treated at least as
carefully as the code in 🤗 Transformers since the documentation is
usually the first contact point of the community with the model.
**Code refactor**
Great, now you have added all the necessary code for *BigBird*.
At this point, you should correct some potential incorrect code style by
running:
```bash
make style
```
and verify that your coding style passes the quality check:
```bash
make quality
```
There are a couple of other very strict design tests in 🤗 Transformers
that might still be failing, which shows up in the tests of your pull
request. This is often because of some missing information in the
docstring or some incorrect naming. Patrick will surely
help you if you're stuck here.
Lastly, it is always a good idea to refactor one's code after having
ensured that the code works correctly. With all tests passing, now it's
a good time to go over the added code again and do some refactoring.
You have now finished the coding part, congratulation! 🎉 You are
Awesome! 😎
**12. Upload the models to the model hub**
In this final part, you should convert and upload all checkpoints to the
model hub and add a model card for each uploaded model checkpoint. You
should work alongside Patrick here to decide on a fitting
name for each checkpoint and to get the required access rights to be
able to upload the model under the author's organization of
*BigBird*.
It is worth spending some time to create fitting model cards for each
checkpoint. The model cards should highlight the specific
characteristics of this particular checkpoint, *e.g.*, On which dataset
was the checkpoint pretrained/fine-tuned on? On what down-stream task
should the model be used? And also include some code on how to correctly
use the model.
**13. (Optional) Add notebook**
It is very helpful to add a notebook that showcases in-detail how
*BigBird* can be used for inference and/or fine-tuned on a
downstream task. This is not mandatory to merge your PR, but very useful
for the community.
**14. Submit your finished PR**
You're done programming now and can move to the last step, which is
getting your PR merged into main. Usually, Patrick
should have helped you already at this point, but it is worth taking
some time to give your finished PR a nice description and eventually add
comments to your code, if you want to point out certain design choices
to your reviewer.
### Share your work!!
Now, it's time to get some credit from the community for your work!
Having completed a model addition is a major contribution to
Transformers and the whole NLP community. Your code and the ported
pre-trained models will certainly be used by hundreds and possibly even
thousands of developers and researchers. You should be proud of your
work and share your achievement with the community.
**You have made another model that is super easy to access for everyone
in the community! 🤯**
| huggingface/transformers/blob/main/templates/adding_a_new_model/open_model_proposals/ADD_BIG_BIRD.md |
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# GPTBigCode
## Overview
The GPTBigCode model was proposed in [SantaCoder: don't reach for the stars!](https://arxiv.org/abs/2301.03988) by BigCode. The listed authors are: Loubna Ben Allal, Raymond Li, Denis Kocetkov, Chenghao Mou, Christopher Akiki, Carlos Munoz Ferrandis, Niklas Muennighoff, Mayank Mishra, Alex Gu, Manan Dey, Logesh Kumar Umapathi, Carolyn Jane Anderson, Yangtian Zi, Joel Lamy Poirier, Hailey Schoelkopf, Sergey Troshin, Dmitry Abulkhanov, Manuel Romero, Michael Lappert, Francesco De Toni, Bernardo García del Río, Qian Liu, Shamik Bose, Urvashi Bhattacharyya, Terry Yue Zhuo, Ian Yu, Paulo Villegas, Marco Zocca, Sourab Mangrulkar, David Lansky, Huu Nguyen, Danish Contractor, Luis Villa, Jia Li, Dzmitry Bahdanau, Yacine Jernite, Sean Hughes, Daniel Fried, Arjun Guha, Harm de Vries, Leandro von Werra.
The abstract from the paper is the following:
*The BigCode project is an open-scientific collaboration working on the responsible development of large language models for code. This tech report describes the progress of the collaboration until December 2022, outlining the current state of the Personally Identifiable Information (PII) redaction pipeline, the experiments conducted to de-risk the model architecture, and the experiments investigating better preprocessing methods for the training data. We train 1.1B parameter models on the Java, JavaScript, and Python subsets of The Stack and evaluate them on the MultiPL-E text-to-code benchmark. We find that more aggressive filtering of near-duplicates can further boost performance and, surprisingly, that selecting files from repositories with 5+ GitHub stars deteriorates performance significantly. Our best model outperforms previous open-source multilingual code generation models (InCoder-6.7B and CodeGen-Multi-2.7B) in both left-to-right generation and infilling on the Java, JavaScript, and Python portions of MultiPL-E, despite being a substantially smaller model. All models are released under an OpenRAIL license at [this https URL.](https://huggingface.co/bigcode)*
The model is an optimized [GPT2 model](https://huggingface.co/docs/transformers/model_doc/gpt2) with support for Multi-Query Attention.
## Implementation details
The main differences compared to GPT2.
- Added support for Multi-Query Attention.
- Use `gelu_pytorch_tanh` instead of classic `gelu`.
- Avoid unnecessary synchronizations (this has since been added to GPT2 in #20061, but wasn't in the reference codebase).
- Use Linear layers instead of Conv1D (good speedup but makes the checkpoints incompatible).
- Merge `_attn` and `_upcast_and_reordered_attn`. Always merge the matmul with scaling. Rename `reorder_and_upcast_attn`->`attention_softmax_in_fp32`
- Cache the attention mask value to avoid recreating it every time.
- Use jit to fuse the attention fp32 casting, masking, softmax, and scaling.
- Combine the attention and causal masks into a single one, pre-computed for the whole model instead of every layer.
- Merge the key and value caches into one (this changes the format of layer_past/ present, does it risk creating problems?)
- Use the memory layout (self.num_heads, 3, self.head_dim) instead of `(3, self.num_heads, self.head_dim)` for the QKV tensor with MHA. (prevents an overhead with the merged key and values, but makes the checkpoints incompatible with the original gpt2 model).
You can read more about the optimizations in the [original pull request](https://github.com/huggingface/transformers/pull/22575)
## Combining Starcoder and Flash Attention 2
First, make sure to install the latest version of Flash Attention 2 to include the sliding window attention feature.
```bash
pip install -U flash-attn --no-build-isolation
```
Make also sure that you have a hardware that is compatible with Flash-Attention 2. Read more about it in the official documentation of flash-attn repository. Make also sure to load your model in half-precision (e.g. `torch.float16``)
To load and run a model using Flash Attention 2, refer to the snippet below:
```python
>>> import torch
>>> from transformers import AutoModelForCausalLM, AutoTokenizer
>>> device = "cuda" # the device to load the model onto
>>> model = AutoModelForCausalLM.from_pretrained("bigcode/gpt_bigcode-santacoder", torch_dtype=torch.float16, attn_implementation="flash_attention_2")
>>> tokenizer = AutoTokenizer.from_pretrained("bigcode/gpt_bigcode-santacoder")
>>> prompt = "def hello_world():"
>>> model_inputs = tokenizer([prompt], return_tensors="pt").to(device)
>>> model.to(device)
>>> generated_ids = model.generate(**model_inputs, max_new_tokens=30, do_sample=False)
>>> tokenizer.batch_decode(generated_ids)[0]
'def hello_world():\n print("hello world")\n\nif __name__ == "__main__":\n print("hello world")\n<|endoftext|>'
```
### Expected speedups
Below is a expected speedup diagram that compares pure inference time between the native implementation in transformers using `bigcode/starcoder` checkpoint and the Flash Attention 2 version of the model using two different sequence lengths.
<div style="text-align: center">
<img src="https://huggingface.co/datasets/ybelkada/documentation-images/resolve/main/starcoder-speedup.png">
</div>
## GPTBigCodeConfig
[[autodoc]] GPTBigCodeConfig
## GPTBigCodeModel
[[autodoc]] GPTBigCodeModel
- forward
## GPTBigCodeForCausalLM
[[autodoc]] GPTBigCodeForCausalLM
- forward
## GPTBigCodeForSequenceClassification
[[autodoc]] GPTBigCodeForSequenceClassification
- forward
## GPTBigCodeForTokenClassification
[[autodoc]] GPTBigCodeForTokenClassification
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/gpt_bigcode.md |
Examples
In this folder we showcase some examples to use code models for downstream tasks.
## Complexity prediction
In this task we want to predict the complexity of Java programs in [CodeComplex](https://huggingface.co/datasets/codeparrot/codecomplex) dataset. Using Hugging Face `trainer`, we finetuned [multilingual CodeParrot](https://huggingface.co/codeparrot/codeparrot-small-multi) and [UniXcoder](https://huggingface.co/microsoft/unixcoder-base-nine) on it, and we used the latter to build this Java complexity prediction [space](https://huggingface.co/spaces/codeparrot/code-complexity-predictor) on Hugging Face hub.
To fine-tune a model on this dataset you can use the following commands:
```python
python train_complexity_predictor.py \
--model_ckpt microsoft/unixcoder-base-nine \
--num_epochs 60 \
--num_warmup_steps 10 \
--batch_size 8 \
--learning_rate 5e-4
```
## Code generation: text to python
In this task we want to train a model to generate code from english text. We finetuned Codeparrot-small on [github-jupyter-text-to-code](https://huggingface.co/datasets/codeparrot/github-jupyter-text-to-code), a dataset where the samples are a succession of docstrings and their Python code, originally extracted from Jupyter notebooks parsed in this [dataset](https://huggingface.co/datasets/codeparrot/github-jupyter-parsed).
To fine-tune a model on this dataset we use the same [script](https://github.com/huggingface/transformers/blob/main/examples/research_projects/codeparrot/scripts/codeparrot_training.py) as the pretraining of codeparrot:
```python
accelerate launch scripts/codeparrot_training.py \
--model_ckpt codeparrot/codeparrot-small \
--dataset_name_train codeparrot/github-jupyter-text-to-code \
--dataset_name_valid codeparrot/github-jupyter-text-to-code \
--train_batch_size 12 \
--valid_batch_size 12 \
--learning_rate 5e-4 \
--num_warmup_steps 100 \
--gradient_accumulation 1 \
--gradient_checkpointing False \
--max_train_steps 3000 \
--save_checkpoint_steps 200 \
--save_dir jupyter-text-to-python
```
## Code explanation: python to text
In this task we want to train a model to explain python code. We finetuned Codeparrot-small on [github-jupyter-code-to-text](https://huggingface.co/datasets/codeparrot/github-jupyter-code-to-text), a dataset where the samples are a succession of Python code and its explanation as a docstring, we just inverted the order of text and code pairs in github-jupyter-code-to-text dataset and added the delimiters "Explanation:" and "End of explanation" inside the doctrings.
To fine-tune a model on this dataset we use the same [script](https://github.com/huggingface/transformers/blob/main/examples/research_projects/codeparrot/scripts/codeparrot_training.py) as the pretraining of codeparrot:
```python
accelerate launch scripts/codeparrot_training.py \
--model_ckpt codeparrot/codeparrot-small \
--dataset_name_train codeparrot/github-jupyter-code-to-text \
--dataset_name_valid codeparrot/github-jupyter-code-to-text \
--train_batch_size 12 \
--valid_batch_size 12 \
--learning_rate 5e-4 \
--num_warmup_steps 100 \
--gradient_accumulation 1 \
--gradient_checkpointing False \
--max_train_steps 3000 \
--save_checkpoint_steps 200 \
--save_dir jupyter-python-to-text
``` | huggingface/transformers/blob/main/examples/research_projects/codeparrot/examples/README.md |
!--Copyright 2021 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Splinter
## Overview
The Splinter model was proposed in [Few-Shot Question Answering by Pretraining Span Selection](https://arxiv.org/abs/2101.00438) by Ori Ram, Yuval Kirstain, Jonathan Berant, Amir Globerson, Omer Levy. Splinter
is an encoder-only transformer (similar to BERT) pretrained using the recurring span selection task on a large corpus
comprising Wikipedia and the Toronto Book Corpus.
The abstract from the paper is the following:
In several question answering benchmarks, pretrained models have reached human parity through fine-tuning on an order
of 100,000 annotated questions and answers. We explore the more realistic few-shot setting, where only a few hundred
training examples are available, and observe that standard models perform poorly, highlighting the discrepancy between
current pretraining objectives and question answering. We propose a new pretraining scheme tailored for question
answering: recurring span selection. Given a passage with multiple sets of recurring spans, we mask in each set all
recurring spans but one, and ask the model to select the correct span in the passage for each masked span. Masked spans
are replaced with a special token, viewed as a question representation, that is later used during fine-tuning to select
the answer span. The resulting model obtains surprisingly good results on multiple benchmarks (e.g., 72.7 F1 on SQuAD
with only 128 training examples), while maintaining competitive performance in the high-resource setting.
This model was contributed by [yuvalkirstain](https://huggingface.co/yuvalkirstain) and [oriram](https://huggingface.co/oriram). The original code can be found [here](https://github.com/oriram/splinter).
## Usage tips
- Splinter was trained to predict answers spans conditioned on a special [QUESTION] token. These tokens contextualize
to question representations which are used to predict the answers. This layer is called QASS, and is the default
behaviour in the [`SplinterForQuestionAnswering`] class. Therefore:
- Use [`SplinterTokenizer`] (rather than [`BertTokenizer`]), as it already
contains this special token. Also, its default behavior is to use this token when two sequences are given (for
example, in the *run_qa.py* script).
- If you plan on using Splinter outside *run_qa.py*, please keep in mind the question token - it might be important for
the success of your model, especially in a few-shot setting.
- Please note there are two different checkpoints for each size of Splinter. Both are basically the same, except that
one also has the pretrained weights of the QASS layer (*tau/splinter-base-qass* and *tau/splinter-large-qass*) and one
doesn't (*tau/splinter-base* and *tau/splinter-large*). This is done to support randomly initializing this layer at
fine-tuning, as it is shown to yield better results for some cases in the paper.
## Resources
- [Question answering task guide](../tasks/question-answering)
## SplinterConfig
[[autodoc]] SplinterConfig
## SplinterTokenizer
[[autodoc]] SplinterTokenizer
- build_inputs_with_special_tokens
- get_special_tokens_mask
- create_token_type_ids_from_sequences
- save_vocabulary
## SplinterTokenizerFast
[[autodoc]] SplinterTokenizerFast
## SplinterModel
[[autodoc]] SplinterModel
- forward
## SplinterForQuestionAnswering
[[autodoc]] SplinterForQuestionAnswering
- forward
## SplinterForPreTraining
[[autodoc]] SplinterForPreTraining
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/splinter.md |
!--Copyright 2022 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Jukebox
## Overview
The Jukebox model was proposed in [Jukebox: A generative model for music](https://arxiv.org/pdf/2005.00341.pdf)
by Prafulla Dhariwal, Heewoo Jun, Christine Payne, Jong Wook Kim, Alec Radford,
Ilya Sutskever. It introduces a generative music model which can produce minute long samples that can be conditioned on
an artist, genres and lyrics.
The abstract from the paper is the following:
*We introduce Jukebox, a model that generates music with singing in the raw audio domain. We tackle the long context of raw audio using a multiscale VQ-VAE to compress it to discrete codes, and modeling those using autoregressive Transformers. We show that the combined model at scale can generate high-fidelity and diverse songs with coherence up to multiple minutes. We can condition on artist and genre to steer the musical and vocal style, and on unaligned lyrics to make the singing more controllable. We are releasing thousands of non cherry-picked samples, along with model weights and code.*
As shown on the following figure, Jukebox is made of 3 `priors` which are decoder only models. They follow the architecture described in [Generating Long Sequences with Sparse Transformers](https://arxiv.org/abs/1904.10509), modified to support longer context length.
First, a autoencoder is used to encode the text lyrics. Next, the first (also called `top_prior`) prior attends to the last hidden states extracted from the lyrics encoder. The priors are linked to the previous priors respectively via an `AudioConditionner` module. The`AudioConditioner` upsamples the outputs of the previous prior to raw tokens at a certain audio frame per second resolution.
The metadata such as *artist, genre and timing* are passed to each prior, in the form of a start token and positional embedding for the timing data. The hidden states are mapped to the closest codebook vector from the VQVAE in order to convert them to raw audio.
This model was contributed by [Arthur Zucker](https://huggingface.co/ArthurZ).
The original code can be found [here](https://github.com/openai/jukebox).
## Usage tips
- This model only supports inference. This is for a few reasons, mostly because it requires a crazy amount of memory to train. Feel free to open a PR and add what's missing to have a full integration with the hugging face traineer!
- This model is very slow, and takes 8h to generate a minute long audio using the 5b top prior on a V100 GPU. In order automaticallay handle the device on which the model should execute, use `accelerate`.
- Contrary to the paper, the order of the priors goes from `0` to `1` as it felt more intuitive : we sample starting from `0`.
- Primed sampling (conditioning the sampling on raw audio) requires more memory than ancestral sampling and should be used with `fp16` set to `True`.
This model was contributed by [Arthur Zucker](https://huggingface.co/ArthurZ).
The original code can be found [here](https://github.com/openai/jukebox).
## JukeboxConfig
[[autodoc]] JukeboxConfig
## JukeboxPriorConfig
[[autodoc]] JukeboxPriorConfig
## JukeboxVQVAEConfig
[[autodoc]] JukeboxVQVAEConfig
## JukeboxTokenizer
[[autodoc]] JukeboxTokenizer
- save_vocabulary
## JukeboxModel
[[autodoc]] JukeboxModel
- ancestral_sample
- primed_sample
- continue_sample
- upsample
- _sample
## JukeboxPrior
[[autodoc]] JukeboxPrior
- sample
- forward
## JukeboxVQVAE
[[autodoc]] JukeboxVQVAE
- forward
- encode
- decode
| huggingface/transformers/blob/main/docs/source/en/model_doc/jukebox.md |
!--Copyright 2022 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Neighborhood Attention Transformer
## Overview
NAT was proposed in [Neighborhood Attention Transformer](https://arxiv.org/abs/2204.07143)
by Ali Hassani, Steven Walton, Jiachen Li, Shen Li, and Humphrey Shi.
It is a hierarchical vision transformer based on Neighborhood Attention, a sliding-window self attention pattern.
The abstract from the paper is the following:
*We present Neighborhood Attention (NA), the first efficient and scalable sliding-window attention mechanism for vision.
NA is a pixel-wise operation, localizing self attention (SA) to the nearest neighboring pixels, and therefore enjoys a
linear time and space complexity compared to the quadratic complexity of SA. The sliding-window pattern allows NA's
receptive field to grow without needing extra pixel shifts, and preserves translational equivariance, unlike
Swin Transformer's Window Self Attention (WSA). We develop NATTEN (Neighborhood Attention Extension), a Python package
with efficient C++ and CUDA kernels, which allows NA to run up to 40% faster than Swin's WSA while using up to 25% less
memory. We further present Neighborhood Attention Transformer (NAT), a new hierarchical transformer design based on NA
that boosts image classification and downstream vision performance. Experimental results on NAT are competitive;
NAT-Tiny reaches 83.2% top-1 accuracy on ImageNet, 51.4% mAP on MS-COCO and 48.4% mIoU on ADE20K, which is 1.9%
ImageNet accuracy, 1.0% COCO mAP, and 2.6% ADE20K mIoU improvement over a Swin model with similar size. *
<img
src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/neighborhood-attention-pattern.jpg"
alt="drawing" width="600"/>
<small> Neighborhood Attention compared to other attention patterns.
Taken from the <a href="https://arxiv.org/abs/2204.07143">original paper</a>.</small>
This model was contributed by [Ali Hassani](https://huggingface.co/alihassanijr).
The original code can be found [here](https://github.com/SHI-Labs/Neighborhood-Attention-Transformer).
## Usage tips
- One can use the [`AutoImageProcessor`] API to prepare images for the model.
- NAT can be used as a *backbone*. When `output_hidden_states = True`,
it will output both `hidden_states` and `reshaped_hidden_states`.
The `reshaped_hidden_states` have a shape of `(batch, num_channels, height, width)` rather than
`(batch_size, height, width, num_channels)`.
Notes:
- NAT depends on [NATTEN](https://github.com/SHI-Labs/NATTEN/)'s implementation of Neighborhood Attention.
You can install it with pre-built wheels for Linux by referring to [shi-labs.com/natten](https://shi-labs.com/natten),
or build on your system by running `pip install natten`.
Note that the latter will likely take time to compile. NATTEN does not support Windows devices yet.
- Patch size of 4 is only supported at the moment.
## Resources
A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with NAT.
<PipelineTag pipeline="image-classification"/>
- [`NatForImageClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/image-classification) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/image_classification.ipynb).
- See also: [Image classification task guide](../tasks/image_classification)
If you're interested in submitting a resource to be included here, please feel free to open a Pull Request and we'll review it! The resource should ideally demonstrate something new instead of duplicating an existing resource.
## NatConfig
[[autodoc]] NatConfig
## NatModel
[[autodoc]] NatModel
- forward
## NatForImageClassification
[[autodoc]] NatForImageClassification
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/nat.md |
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# LLaVa
## Overview
LLaVa is an open-source chatbot trained by fine-tuning LlamA/Vicuna on GPT-generated multimodal instruction-following data. It is an auto-regressive language model, based on the transformer architecture. In other words, it is an multi-modal version of LLMs fine-tuned for chat / instructions.
The LLaVa model was proposed in [Visual Instruction Tuning](https://arxiv.org/abs/2304.08485) and improved in [Improved Baselines with Visual Instruction Tuning](https://arxiv.org/pdf/2310.03744) by Haotian Liu, Chunyuan Li, Yuheng Li and Yong Jae Lee.
The abstract from the paper is the following:
*Large multimodal models (LMM) have recently shown encouraging progress with visual instruction tuning. In this note, we show that the fully-connected vision-language cross-modal connector in LLaVA is surprisingly powerful and data-efficient. With simple modifications to LLaVA, namely, using CLIP-ViT-L-336px with an MLP projection and adding academic-task-oriented VQA data with simple response formatting prompts, we establish stronger baselines that achieve state-of-the-art across 11 benchmarks. Our final 13B checkpoint uses merely 1.2M publicly available data, and finishes full training in ∼1 day on a single 8-A100 node. We hope this can make state-of-the-art LMM research more accessible. Code and model will be publicly available*
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/llava_architecture.jpg"
alt="drawing" width="600"/>
<small> LLaVa architecture. Taken from the <a href="https://arxiv.org/abs/2304.08485">original paper.</a> </small>
This model was contributed by [ArthurZ](https://huggingface.co/ArthurZ) and [ybelkada](https://huggingface.co/ybelkada).
The original code can be found [here](https://github.com/haotian-liu/LLaVA/tree/main/llava).
## Usage tips
- We advise users to use `padding_side="left"` when computing batched generation as it leads to more accurate results. Simply make sure to call `processor.tokenizer.padding_side = "left"` before generating.
- Note the model has not been explicitly trained to process multiple images in the same prompt, although this is technically possible, you may experience inaccurate results.
- For better results, we recommend users to prompt the model with the correct prompt format:
```bash
"USER: <image>\n<prompt>ASSISTANT:"
```
For multiple turns conversation:
```bash
"USER: <image>\n<prompt1>ASSISTANT: <answer1>USER: <prompt2>ASSISTANT: <answer2>USER: <prompt3>ASSISTANT:"
```
### Using Flash Attention 2
Flash Attention 2 is an even faster, optimized version of the previous optimization, please refer to the [Flash Attention 2 section of performance docs](https://huggingface.co/docs/transformers/perf_infer_gpu_one).
## Resources
A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with BEiT.
<PipelineTag pipeline="image-to-text"/>
- A [Google Colab demo](https://colab.research.google.com/drive/1qsl6cd2c8gGtEW1xV5io7S8NHh-Cp1TV?usp=sharing) on how to run Llava on a free-tier Google colab instance leveraging 4-bit inference.
- A [similar notebook](https://github.com/NielsRogge/Transformers-Tutorials/blob/master/LLaVa/Inference_with_LLaVa_for_multimodal_generation.ipynb) showcasing batched inference. 🌎
## LlavaConfig
[[autodoc]] LlavaConfig
## LlavaProcessor
[[autodoc]] LlavaProcessor
## LlavaForConditionalGeneration
[[autodoc]] LlavaForConditionalGeneration
- forward
| huggingface/transformers/blob/main/docs/source/en/model_doc/llava.md |
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
-->
# InstructBLIP
## Overview
The InstructBLIP model was proposed in [InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning](https://arxiv.org/abs/2305.06500) by Wenliang Dai, Junnan Li, Dongxu Li, Anthony Meng Huat Tiong, Junqi Zhao, Weisheng Wang, Boyang Li, Pascale Fung, Steven Hoi.
InstructBLIP leverages the [BLIP-2](blip2) architecture for visual instruction tuning.
The abstract from the paper is the following:
*General-purpose language models that can solve various language-domain tasks have emerged driven by the pre-training and instruction-tuning pipeline. However, building general-purpose vision-language models is challenging due to the increased task discrepancy introduced by the additional visual input. Although vision-language pre-training has been widely studied, vision-language instruction tuning remains relatively less explored. In this paper, we conduct a systematic and comprehensive study on vision-language instruction tuning based on the pre-trained BLIP-2 models. We gather a wide variety of 26 publicly available datasets, transform them into instruction tuning format and categorize them into two clusters for held-in instruction tuning and held-out zero-shot evaluation. Additionally, we introduce instruction-aware visual feature extraction, a crucial method that enables the model to extract informative features tailored to the given instruction. The resulting InstructBLIP models achieve state-of-the-art zero-shot performance across all 13 held-out datasets, substantially outperforming BLIP-2 and the larger Flamingo. Our models also lead to state-of-the-art performance when finetuned on individual downstream tasks (e.g., 90.7% accuracy on ScienceQA IMG). Furthermore, we qualitatively demonstrate the advantages of InstructBLIP over concurrent multimodal models.*
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/instructblip_architecture.jpg"
alt="drawing" width="600"/>
<small> InstructBLIP architecture. Taken from the <a href="https://arxiv.org/abs/2305.06500">original paper.</a> </small>
This model was contributed by [nielsr](https://huggingface.co/nielsr).
The original code can be found [here](https://github.com/salesforce/LAVIS/tree/main/projects/instructblip).
## Usage tips
InstructBLIP uses the same architecture as [BLIP-2](blip2) with a tiny but important difference: it also feeds the text prompt (instruction) to the Q-Former.
## InstructBlipConfig
[[autodoc]] InstructBlipConfig
- from_vision_qformer_text_configs
## InstructBlipVisionConfig
[[autodoc]] InstructBlipVisionConfig
## InstructBlipQFormerConfig
[[autodoc]] InstructBlipQFormerConfig
## InstructBlipProcessor
[[autodoc]] InstructBlipProcessor
## InstructBlipVisionModel
[[autodoc]] InstructBlipVisionModel
- forward
## InstructBlipQFormerModel
[[autodoc]] InstructBlipQFormerModel
- forward
## InstructBlipForConditionalGeneration
[[autodoc]] InstructBlipForConditionalGeneration
- forward
- generate | huggingface/transformers/blob/main/docs/source/en/model_doc/instructblip.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Auto Classes
In many cases, the architecture you want to use can be guessed from the name or the path of the pretrained model you
are supplying to the `from_pretrained()` method. AutoClasses are here to do this job for you so that you
automatically retrieve the relevant model given the name/path to the pretrained weights/config/vocabulary.
Instantiating one of [`AutoConfig`], [`AutoModel`], and
[`AutoTokenizer`] will directly create a class of the relevant architecture. For instance
```python
model = AutoModel.from_pretrained("bert-base-cased")
```
will create a model that is an instance of [`BertModel`].
There is one class of `AutoModel` for each task, and for each backend (PyTorch, TensorFlow, or Flax).
## Extending the Auto Classes
Each of the auto classes has a method to be extended with your custom classes. For instance, if you have defined a
custom class of model `NewModel`, make sure you have a `NewModelConfig` then you can add those to the auto
classes like this:
```python
from transformers import AutoConfig, AutoModel
AutoConfig.register("new-model", NewModelConfig)
AutoModel.register(NewModelConfig, NewModel)
```
You will then be able to use the auto classes like you would usually do!
<Tip warning={true}>
If your `NewModelConfig` is a subclass of [`~transformers.PretrainedConfig`], make sure its
`model_type` attribute is set to the same key you use when registering the config (here `"new-model"`).
Likewise, if your `NewModel` is a subclass of [`PreTrainedModel`], make sure its
`config_class` attribute is set to the same class you use when registering the model (here
`NewModelConfig`).
</Tip>
## AutoConfig
[[autodoc]] AutoConfig
## AutoTokenizer
[[autodoc]] AutoTokenizer
## AutoFeatureExtractor
[[autodoc]] AutoFeatureExtractor
## AutoImageProcessor
[[autodoc]] AutoImageProcessor
## AutoProcessor
[[autodoc]] AutoProcessor
## Generic model classes
The following auto classes are available for instantiating a base model class without a specific head.
### AutoModel
[[autodoc]] AutoModel
### TFAutoModel
[[autodoc]] TFAutoModel
### FlaxAutoModel
[[autodoc]] FlaxAutoModel
## Generic pretraining classes
The following auto classes are available for instantiating a model with a pretraining head.
### AutoModelForPreTraining
[[autodoc]] AutoModelForPreTraining
### TFAutoModelForPreTraining
[[autodoc]] TFAutoModelForPreTraining
### FlaxAutoModelForPreTraining
[[autodoc]] FlaxAutoModelForPreTraining
## Natural Language Processing
The following auto classes are available for the following natural language processing tasks.
### AutoModelForCausalLM
[[autodoc]] AutoModelForCausalLM
### TFAutoModelForCausalLM
[[autodoc]] TFAutoModelForCausalLM
### FlaxAutoModelForCausalLM
[[autodoc]] FlaxAutoModelForCausalLM
### AutoModelForMaskedLM
[[autodoc]] AutoModelForMaskedLM
### TFAutoModelForMaskedLM
[[autodoc]] TFAutoModelForMaskedLM
### FlaxAutoModelForMaskedLM
[[autodoc]] FlaxAutoModelForMaskedLM
### AutoModelForMaskGeneration
[[autodoc]] AutoModelForMaskGeneration
### TFAutoModelForMaskGeneration
[[autodoc]] TFAutoModelForMaskGeneration
### AutoModelForSeq2SeqLM
[[autodoc]] AutoModelForSeq2SeqLM
### TFAutoModelForSeq2SeqLM
[[autodoc]] TFAutoModelForSeq2SeqLM
### FlaxAutoModelForSeq2SeqLM
[[autodoc]] FlaxAutoModelForSeq2SeqLM
### AutoModelForSequenceClassification
[[autodoc]] AutoModelForSequenceClassification
### TFAutoModelForSequenceClassification
[[autodoc]] TFAutoModelForSequenceClassification
### FlaxAutoModelForSequenceClassification
[[autodoc]] FlaxAutoModelForSequenceClassification
### AutoModelForMultipleChoice
[[autodoc]] AutoModelForMultipleChoice
### TFAutoModelForMultipleChoice
[[autodoc]] TFAutoModelForMultipleChoice
### FlaxAutoModelForMultipleChoice
[[autodoc]] FlaxAutoModelForMultipleChoice
### AutoModelForNextSentencePrediction
[[autodoc]] AutoModelForNextSentencePrediction
### TFAutoModelForNextSentencePrediction
[[autodoc]] TFAutoModelForNextSentencePrediction
### FlaxAutoModelForNextSentencePrediction
[[autodoc]] FlaxAutoModelForNextSentencePrediction
### AutoModelForTokenClassification
[[autodoc]] AutoModelForTokenClassification
### TFAutoModelForTokenClassification
[[autodoc]] TFAutoModelForTokenClassification
### FlaxAutoModelForTokenClassification
[[autodoc]] FlaxAutoModelForTokenClassification
### AutoModelForQuestionAnswering
[[autodoc]] AutoModelForQuestionAnswering
### TFAutoModelForQuestionAnswering
[[autodoc]] TFAutoModelForQuestionAnswering
### FlaxAutoModelForQuestionAnswering
[[autodoc]] FlaxAutoModelForQuestionAnswering
### AutoModelForTextEncoding
[[autodoc]] AutoModelForTextEncoding
### TFAutoModelForTextEncoding
[[autodoc]] TFAutoModelForTextEncoding
## Computer vision
The following auto classes are available for the following computer vision tasks.
### AutoModelForDepthEstimation
[[autodoc]] AutoModelForDepthEstimation
### AutoModelForImageClassification
[[autodoc]] AutoModelForImageClassification
### TFAutoModelForImageClassification
[[autodoc]] TFAutoModelForImageClassification
### FlaxAutoModelForImageClassification
[[autodoc]] FlaxAutoModelForImageClassification
### AutoModelForVideoClassification
[[autodoc]] AutoModelForVideoClassification
### AutoModelForMaskedImageModeling
[[autodoc]] AutoModelForMaskedImageModeling
### TFAutoModelForMaskedImageModeling
[[autodoc]] TFAutoModelForMaskedImageModeling
### AutoModelForObjectDetection
[[autodoc]] AutoModelForObjectDetection
### AutoModelForImageSegmentation
[[autodoc]] AutoModelForImageSegmentation
### AutoModelForImageToImage
[[autodoc]] AutoModelForImageToImage
### AutoModelForSemanticSegmentation
[[autodoc]] AutoModelForSemanticSegmentation
### TFAutoModelForSemanticSegmentation
[[autodoc]] TFAutoModelForSemanticSegmentation
### AutoModelForInstanceSegmentation
[[autodoc]] AutoModelForInstanceSegmentation
### AutoModelForUniversalSegmentation
[[autodoc]] AutoModelForUniversalSegmentation
### AutoModelForZeroShotImageClassification
[[autodoc]] AutoModelForZeroShotImageClassification
### TFAutoModelForZeroShotImageClassification
[[autodoc]] TFAutoModelForZeroShotImageClassification
### AutoModelForZeroShotObjectDetection
[[autodoc]] AutoModelForZeroShotObjectDetection
## Audio
The following auto classes are available for the following audio tasks.
### AutoModelForAudioClassification
[[autodoc]] AutoModelForAudioClassification
### AutoModelForAudioFrameClassification
[[autodoc]] TFAutoModelForAudioClassification
### TFAutoModelForAudioFrameClassification
[[autodoc]] AutoModelForAudioFrameClassification
### AutoModelForCTC
[[autodoc]] AutoModelForCTC
### AutoModelForSpeechSeq2Seq
[[autodoc]] AutoModelForSpeechSeq2Seq
### TFAutoModelForSpeechSeq2Seq
[[autodoc]] TFAutoModelForSpeechSeq2Seq
### FlaxAutoModelForSpeechSeq2Seq
[[autodoc]] FlaxAutoModelForSpeechSeq2Seq
### AutoModelForAudioXVector
[[autodoc]] AutoModelForAudioXVector
### AutoModelForTextToSpectrogram
[[autodoc]] AutoModelForTextToSpectrogram
### AutoModelForTextToWaveform
[[autodoc]] AutoModelForTextToWaveform
## Multimodal
The following auto classes are available for the following multimodal tasks.
### AutoModelForTableQuestionAnswering
[[autodoc]] AutoModelForTableQuestionAnswering
### TFAutoModelForTableQuestionAnswering
[[autodoc]] TFAutoModelForTableQuestionAnswering
### AutoModelForDocumentQuestionAnswering
[[autodoc]] AutoModelForDocumentQuestionAnswering
### TFAutoModelForDocumentQuestionAnswering
[[autodoc]] TFAutoModelForDocumentQuestionAnswering
### AutoModelForVisualQuestionAnswering
[[autodoc]] AutoModelForVisualQuestionAnswering
### AutoModelForVision2Seq
[[autodoc]] AutoModelForVision2Seq
### TFAutoModelForVision2Seq
[[autodoc]] TFAutoModelForVision2Seq
### FlaxAutoModelForVision2Seq
[[autodoc]] FlaxAutoModelForVision2Seq
| huggingface/transformers/blob/main/docs/source/en/model_doc/auto.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# CamemBERT
## Overview
The CamemBERT model was proposed in [CamemBERT: a Tasty French Language Model](https://arxiv.org/abs/1911.03894) by
Louis Martin, Benjamin Muller, Pedro Javier Ortiz Suárez, Yoann Dupont, Laurent Romary, Éric Villemonte de la
Clergerie, Djamé Seddah, and Benoît Sagot. It is based on Facebook's RoBERTa model released in 2019. It is a model
trained on 138GB of French text.
The abstract from the paper is the following:
*Pretrained language models are now ubiquitous in Natural Language Processing. Despite their success, most available
models have either been trained on English data or on the concatenation of data in multiple languages. This makes
practical use of such models --in all languages except English-- very limited. Aiming to address this issue for French,
we release CamemBERT, a French version of the Bi-directional Encoders for Transformers (BERT). We measure the
performance of CamemBERT compared to multilingual models in multiple downstream tasks, namely part-of-speech tagging,
dependency parsing, named-entity recognition, and natural language inference. CamemBERT improves the state of the art
for most of the tasks considered. We release the pretrained model for CamemBERT hoping to foster research and
downstream applications for French NLP.*
This model was contributed by [camembert](https://huggingface.co/camembert). The original code can be found [here](https://camembert-model.fr/).
<Tip>
This implementation is the same as RoBERTa. Refer to the [documentation of RoBERTa](roberta) for usage examples as well
as the information relative to the inputs and outputs.
</Tip>
## Resources
- [Text classification task guide](../tasks/sequence_classification)
- [Token classification task guide](../tasks/token_classification)
- [Question answering task guide](../tasks/question_answering)
- [Causal language modeling task guide](../tasks/language_modeling)
- [Masked language modeling task guide](../tasks/masked_language_modeling)
- [Multiple choice task guide](../tasks/multiple_choice)
## CamembertConfig
[[autodoc]] CamembertConfig
## CamembertTokenizer
[[autodoc]] CamembertTokenizer
- build_inputs_with_special_tokens
- get_special_tokens_mask
- create_token_type_ids_from_sequences
- save_vocabulary
## CamembertTokenizerFast
[[autodoc]] CamembertTokenizerFast
<frameworkcontent>
<pt>
## CamembertModel
[[autodoc]] CamembertModel
## CamembertForCausalLM
[[autodoc]] CamembertForCausalLM
## CamembertForMaskedLM
[[autodoc]] CamembertForMaskedLM
## CamembertForSequenceClassification
[[autodoc]] CamembertForSequenceClassification
## CamembertForMultipleChoice
[[autodoc]] CamembertForMultipleChoice
## CamembertForTokenClassification
[[autodoc]] CamembertForTokenClassification
## CamembertForQuestionAnswering
[[autodoc]] CamembertForQuestionAnswering
</pt>
<tf>
## TFCamembertModel
[[autodoc]] TFCamembertModel
## TFCamembertForCasualLM
[[autodoc]] TFCamembertForCausalLM
## TFCamembertForMaskedLM
[[autodoc]] TFCamembertForMaskedLM
## TFCamembertForSequenceClassification
[[autodoc]] TFCamembertForSequenceClassification
## TFCamembertForMultipleChoice
[[autodoc]] TFCamembertForMultipleChoice
## TFCamembertForTokenClassification
[[autodoc]] TFCamembertForTokenClassification
## TFCamembertForQuestionAnswering
[[autodoc]] TFCamembertForQuestionAnswering
</tf>
</frameworkcontent>
| huggingface/transformers/blob/main/docs/source/en/model_doc/camembert.md |
# Token classification
Based on the scripts [`run_ner.py`](https://github.com/huggingface/transformers/blob/main/examples/legacy/token-classification/run_ner.py).
The following examples are covered in this section:
* NER on the GermEval 2014 (German NER) dataset
* Emerging and Rare Entities task: WNUT’17 (English NER) dataset
Details and results for the fine-tuning provided by @stefan-it.
### GermEval 2014 (German NER) dataset
#### Data (Download and pre-processing steps)
Data can be obtained from the [GermEval 2014](https://sites.google.com/site/germeval2014ner/data) shared task page.
Here are the commands for downloading and pre-processing train, dev and test datasets. The original data format has four (tab-separated) columns, in a pre-processing step only the two relevant columns (token and outer span NER annotation) are extracted:
```bash
curl -L 'https://drive.google.com/uc?export=download&id=1Jjhbal535VVz2ap4v4r_rN1UEHTdLK5P' \
| grep -v "^#" | cut -f 2,3 | tr '\t' ' ' > train.txt.tmp
curl -L 'https://drive.google.com/uc?export=download&id=1ZfRcQThdtAR5PPRjIDtrVP7BtXSCUBbm' \
| grep -v "^#" | cut -f 2,3 | tr '\t' ' ' > dev.txt.tmp
curl -L 'https://drive.google.com/uc?export=download&id=1u9mb7kNJHWQCWyweMDRMuTFoOHOfeBTH' \
| grep -v "^#" | cut -f 2,3 | tr '\t' ' ' > test.txt.tmp
```
The GermEval 2014 dataset contains some strange "control character" tokens like `'\x96', '\u200e', '\x95', '\xad' or '\x80'`.
One problem with these tokens is, that `BertTokenizer` returns an empty token for them, resulting in misaligned `InputExample`s.
The `preprocess.py` script located in the `scripts` folder a) filters these tokens and b) splits longer sentences into smaller ones (once the max. subtoken length is reached).
Let's define some variables that we need for further pre-processing steps and training the model:
```bash
export MAX_LENGTH=128
export BERT_MODEL=bert-base-multilingual-cased
```
Run the pre-processing script on training, dev and test datasets:
```bash
python3 scripts/preprocess.py train.txt.tmp $BERT_MODEL $MAX_LENGTH > train.txt
python3 scripts/preprocess.py dev.txt.tmp $BERT_MODEL $MAX_LENGTH > dev.txt
python3 scripts/preprocess.py test.txt.tmp $BERT_MODEL $MAX_LENGTH > test.txt
```
The GermEval 2014 dataset has much more labels than CoNLL-2002/2003 datasets, so an own set of labels must be used:
```bash
cat train.txt dev.txt test.txt | cut -d " " -f 2 | grep -v "^$"| sort | uniq > labels.txt
```
#### Prepare the run
Additional environment variables must be set:
```bash
export OUTPUT_DIR=germeval-model
export BATCH_SIZE=32
export NUM_EPOCHS=3
export SAVE_STEPS=750
export SEED=1
```
#### Run the Pytorch version
To start training, just run:
```bash
python3 run_ner.py --data_dir ./ \
--labels ./labels.txt \
--model_name_or_path $BERT_MODEL \
--output_dir $OUTPUT_DIR \
--max_seq_length $MAX_LENGTH \
--num_train_epochs $NUM_EPOCHS \
--per_device_train_batch_size $BATCH_SIZE \
--save_steps $SAVE_STEPS \
--seed $SEED \
--do_train \
--do_eval \
--do_predict
```
If your GPU supports half-precision training, just add the `--fp16` flag. After training, the model will be both evaluated on development and test datasets.
#### JSON-based configuration file
Instead of passing all parameters via commandline arguments, the `run_ner.py` script also supports reading parameters from a json-based configuration file:
```json
{
"data_dir": ".",
"labels": "./labels.txt",
"model_name_or_path": "bert-base-multilingual-cased",
"output_dir": "germeval-model",
"max_seq_length": 128,
"num_train_epochs": 3,
"per_device_train_batch_size": 32,
"save_steps": 750,
"seed": 1,
"do_train": true,
"do_eval": true,
"do_predict": true
}
```
It must be saved with a `.json` extension and can be used by running `python3 run_ner.py config.json`.
#### Evaluation
Evaluation on development dataset outputs the following for our example:
```bash
10/04/2019 00:42:06 - INFO - __main__ - ***** Eval results *****
10/04/2019 00:42:06 - INFO - __main__ - f1 = 0.8623348017621146
10/04/2019 00:42:06 - INFO - __main__ - loss = 0.07183869666975543
10/04/2019 00:42:06 - INFO - __main__ - precision = 0.8467916366258111
10/04/2019 00:42:06 - INFO - __main__ - recall = 0.8784592370979806
```
On the test dataset the following results could be achieved:
```bash
10/04/2019 00:42:42 - INFO - __main__ - ***** Eval results *****
10/04/2019 00:42:42 - INFO - __main__ - f1 = 0.8614389652384803
10/04/2019 00:42:42 - INFO - __main__ - loss = 0.07064602487454782
10/04/2019 00:42:42 - INFO - __main__ - precision = 0.8604651162790697
10/04/2019 00:42:42 - INFO - __main__ - recall = 0.8624150210424085
```
#### Run the Tensorflow 2 version
To start training, just run:
```bash
python3 run_tf_ner.py --data_dir ./ \
--labels ./labels.txt \
--model_name_or_path $BERT_MODEL \
--output_dir $OUTPUT_DIR \
--max_seq_length $MAX_LENGTH \
--num_train_epochs $NUM_EPOCHS \
--per_device_train_batch_size $BATCH_SIZE \
--save_steps $SAVE_STEPS \
--seed $SEED \
--do_train \
--do_eval \
--do_predict
```
Such as the Pytorch version, if your GPU supports half-precision training, just add the `--fp16` flag. After training, the model will be both evaluated on development and test datasets.
#### Evaluation
Evaluation on development dataset outputs the following for our example:
```bash
precision recall f1-score support
LOCderiv 0.7619 0.6154 0.6809 52
PERpart 0.8724 0.8997 0.8858 4057
OTHpart 0.9360 0.9466 0.9413 711
ORGpart 0.7015 0.6989 0.7002 269
LOCpart 0.7668 0.8488 0.8057 496
LOC 0.8745 0.9191 0.8963 235
ORGderiv 0.7723 0.8571 0.8125 91
OTHderiv 0.4800 0.6667 0.5581 18
OTH 0.5789 0.6875 0.6286 16
PERderiv 0.5385 0.3889 0.4516 18
PER 0.5000 0.5000 0.5000 2
ORG 0.0000 0.0000 0.0000 3
micro avg 0.8574 0.8862 0.8715 5968
macro avg 0.8575 0.8862 0.8713 5968
```
On the test dataset the following results could be achieved:
```bash
precision recall f1-score support
PERpart 0.8847 0.8944 0.8896 9397
OTHpart 0.9376 0.9353 0.9365 1639
ORGpart 0.7307 0.7044 0.7173 697
LOC 0.9133 0.9394 0.9262 561
LOCpart 0.8058 0.8157 0.8107 1150
ORG 0.0000 0.0000 0.0000 8
OTHderiv 0.5882 0.4762 0.5263 42
PERderiv 0.6571 0.5227 0.5823 44
OTH 0.4906 0.6667 0.5652 39
ORGderiv 0.7016 0.7791 0.7383 172
LOCderiv 0.8256 0.6514 0.7282 109
PER 0.0000 0.0000 0.0000 11
micro avg 0.8722 0.8774 0.8748 13869
macro avg 0.8712 0.8774 0.8740 13869
```
### Emerging and Rare Entities task: WNUT’17 (English NER) dataset
Description of the WNUT’17 task from the [shared task website](http://noisy-text.github.io/2017/index.html):
> The WNUT’17 shared task focuses on identifying unusual, previously-unseen entities in the context of emerging discussions.
> Named entities form the basis of many modern approaches to other tasks (like event clustering and summarization), but recall on
> them is a real problem in noisy text - even among annotators. This drop tends to be due to novel entities and surface forms.
Six labels are available in the dataset. An overview can be found on this [page](http://noisy-text.github.io/2017/files/).
#### Data (Download and pre-processing steps)
The dataset can be downloaded from the [official GitHub](https://github.com/leondz/emerging_entities_17) repository.
The following commands show how to prepare the dataset for fine-tuning:
```bash
mkdir -p data_wnut_17
curl -L 'https://github.com/leondz/emerging_entities_17/raw/master/wnut17train.conll' | tr '\t' ' ' > data_wnut_17/train.txt.tmp
curl -L 'https://github.com/leondz/emerging_entities_17/raw/master/emerging.dev.conll' | tr '\t' ' ' > data_wnut_17/dev.txt.tmp
curl -L 'https://raw.githubusercontent.com/leondz/emerging_entities_17/master/emerging.test.annotated' | tr '\t' ' ' > data_wnut_17/test.txt.tmp
```
Let's define some variables that we need for further pre-processing steps:
```bash
export MAX_LENGTH=128
export BERT_MODEL=bert-large-cased
```
Here we use the English BERT large model for fine-tuning.
The `preprocess.py` scripts splits longer sentences into smaller ones (once the max. subtoken length is reached):
```bash
python3 scripts/preprocess.py data_wnut_17/train.txt.tmp $BERT_MODEL $MAX_LENGTH > data_wnut_17/train.txt
python3 scripts/preprocess.py data_wnut_17/dev.txt.tmp $BERT_MODEL $MAX_LENGTH > data_wnut_17/dev.txt
python3 scripts/preprocess.py data_wnut_17/test.txt.tmp $BERT_MODEL $MAX_LENGTH > data_wnut_17/test.txt
```
In the last pre-processing step, the `labels.txt` file needs to be generated. This file contains all available labels:
```bash
cat data_wnut_17/train.txt data_wnut_17/dev.txt data_wnut_17/test.txt | cut -d " " -f 2 | grep -v "^$"| sort | uniq > data_wnut_17/labels.txt
```
#### Run the Pytorch version
Fine-tuning with the PyTorch version can be started using the `run_ner.py` script. In this example we use a JSON-based configuration file.
This configuration file looks like:
```json
{
"data_dir": "./data_wnut_17",
"labels": "./data_wnut_17/labels.txt",
"model_name_or_path": "bert-large-cased",
"output_dir": "wnut-17-model-1",
"max_seq_length": 128,
"num_train_epochs": 3,
"per_device_train_batch_size": 32,
"save_steps": 425,
"seed": 1,
"do_train": true,
"do_eval": true,
"do_predict": true,
"fp16": false
}
```
If your GPU supports half-precision training, please set `fp16` to `true`.
Save this JSON-based configuration under `wnut_17.json`. The fine-tuning can be started with `python3 run_ner_old.py wnut_17.json`.
#### Evaluation
Evaluation on development dataset outputs the following:
```bash
05/29/2020 23:33:44 - INFO - __main__ - ***** Eval results *****
05/29/2020 23:33:44 - INFO - __main__ - eval_loss = 0.26505235286212275
05/29/2020 23:33:44 - INFO - __main__ - eval_precision = 0.7008264462809918
05/29/2020 23:33:44 - INFO - __main__ - eval_recall = 0.507177033492823
05/29/2020 23:33:44 - INFO - __main__ - eval_f1 = 0.5884802220680084
05/29/2020 23:33:44 - INFO - __main__ - epoch = 3.0
```
On the test dataset the following results could be achieved:
```bash
05/29/2020 23:33:44 - INFO - transformers.trainer - ***** Running Prediction *****
05/29/2020 23:34:02 - INFO - __main__ - eval_loss = 0.30948806500973547
05/29/2020 23:34:02 - INFO - __main__ - eval_precision = 0.5840108401084011
05/29/2020 23:34:02 - INFO - __main__ - eval_recall = 0.3994439295644115
05/29/2020 23:34:02 - INFO - __main__ - eval_f1 = 0.47440836543753434
```
WNUT’17 is a very difficult task. Current state-of-the-art results on this dataset can be found [here](https://nlpprogress.com/english/named_entity_recognition.html).
| huggingface/transformers/blob/main/examples/legacy/token-classification/README.md |
!--Copyright 2022 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Audio classification
[[open-in-colab]]
<Youtube id="KWwzcmG98Ds"/>
Audio classification - just like with text - assigns a class label output from the input data. The only difference is instead of text inputs, you have raw audio waveforms. Some practical applications of audio classification include identifying speaker intent, language classification, and even animal species by their sounds.
This guide will show you how to:
1. Finetune [Wav2Vec2](https://huggingface.co/facebook/wav2vec2-base) on the [MInDS-14](https://huggingface.co/datasets/PolyAI/minds14) dataset to classify speaker intent.
2. Use your finetuned model for inference.
<Tip>
The task illustrated in this tutorial is supported by the following model architectures:
<!--This tip is automatically generated by `make fix-copies`, do not fill manually!-->
[Audio Spectrogram Transformer](../model_doc/audio-spectrogram-transformer), [Data2VecAudio](../model_doc/data2vec-audio), [Hubert](../model_doc/hubert), [SEW](../model_doc/sew), [SEW-D](../model_doc/sew-d), [UniSpeech](../model_doc/unispeech), [UniSpeechSat](../model_doc/unispeech-sat), [Wav2Vec2](../model_doc/wav2vec2), [Wav2Vec2-Conformer](../model_doc/wav2vec2-conformer), [WavLM](../model_doc/wavlm), [Whisper](../model_doc/whisper)
<!--End of the generated tip-->
</Tip>
Before you begin, make sure you have all the necessary libraries installed:
```bash
pip install transformers datasets evaluate
```
We encourage you to login to your Hugging Face account so you can upload and share your model with the community. When prompted, enter your token to login:
```py
>>> from huggingface_hub import notebook_login
>>> notebook_login()
```
## Load MInDS-14 dataset
Start by loading the MInDS-14 dataset from the 🤗 Datasets library:
```py
>>> from datasets import load_dataset, Audio
>>> minds = load_dataset("PolyAI/minds14", name="en-US", split="train")
```
Split the dataset's `train` split into a smaller train and test set with the [`~datasets.Dataset.train_test_split`] method. This'll give you a chance to experiment and make sure everything works before spending more time on the full dataset.
```py
>>> minds = minds.train_test_split(test_size=0.2)
```
Then take a look at the dataset:
```py
>>> minds
DatasetDict({
train: Dataset({
features: ['path', 'audio', 'transcription', 'english_transcription', 'intent_class', 'lang_id'],
num_rows: 450
})
test: Dataset({
features: ['path', 'audio', 'transcription', 'english_transcription', 'intent_class', 'lang_id'],
num_rows: 113
})
})
```
While the dataset contains a lot of useful information, like `lang_id` and `english_transcription`, you'll focus on the `audio` and `intent_class` in this guide. Remove the other columns with the [`~datasets.Dataset.remove_columns`] method:
```py
>>> minds = minds.remove_columns(["path", "transcription", "english_transcription", "lang_id"])
```
Take a look at an example now:
```py
>>> minds["train"][0]
{'audio': {'array': array([ 0. , 0. , 0. , ..., -0.00048828,
-0.00024414, -0.00024414], dtype=float32),
'path': '/root/.cache/huggingface/datasets/downloads/extracted/f14948e0e84be638dd7943ac36518a4cf3324e8b7aa331c5ab11541518e9368c/en-US~APP_ERROR/602b9a5fbb1e6d0fbce91f52.wav',
'sampling_rate': 8000},
'intent_class': 2}
```
There are two fields:
- `audio`: a 1-dimensional `array` of the speech signal that must be called to load and resample the audio file.
- `intent_class`: represents the class id of the speaker's intent.
To make it easier for the model to get the label name from the label id, create a dictionary that maps the label name to an integer and vice versa:
```py
>>> labels = minds["train"].features["intent_class"].names
>>> label2id, id2label = dict(), dict()
>>> for i, label in enumerate(labels):
... label2id[label] = str(i)
... id2label[str(i)] = label
```
Now you can convert the label id to a label name:
```py
>>> id2label[str(2)]
'app_error'
```
## Preprocess
The next step is to load a Wav2Vec2 feature extractor to process the audio signal:
```py
>>> from transformers import AutoFeatureExtractor
>>> feature_extractor = AutoFeatureExtractor.from_pretrained("facebook/wav2vec2-base")
```
The MInDS-14 dataset has a sampling rate of 8000khz (you can find this information in it's [dataset card](https://huggingface.co/datasets/PolyAI/minds14)), which means you'll need to resample the dataset to 16000kHz to use the pretrained Wav2Vec2 model:
```py
>>> minds = minds.cast_column("audio", Audio(sampling_rate=16_000))
>>> minds["train"][0]
{'audio': {'array': array([ 2.2098757e-05, 4.6582241e-05, -2.2803260e-05, ...,
-2.8419291e-04, -2.3305941e-04, -1.1425107e-04], dtype=float32),
'path': '/root/.cache/huggingface/datasets/downloads/extracted/f14948e0e84be638dd7943ac36518a4cf3324e8b7aa331c5ab11541518e9368c/en-US~APP_ERROR/602b9a5fbb1e6d0fbce91f52.wav',
'sampling_rate': 16000},
'intent_class': 2}
```
Now create a preprocessing function that:
1. Calls the `audio` column to load, and if necessary, resample the audio file.
2. Checks if the sampling rate of the audio file matches the sampling rate of the audio data a model was pretrained with. You can find this information in the Wav2Vec2 [model card](https://huggingface.co/facebook/wav2vec2-base).
3. Set a maximum input length to batch longer inputs without truncating them.
```py
>>> def preprocess_function(examples):
... audio_arrays = [x["array"] for x in examples["audio"]]
... inputs = feature_extractor(
... audio_arrays, sampling_rate=feature_extractor.sampling_rate, max_length=16000, truncation=True
... )
... return inputs
```
To apply the preprocessing function over the entire dataset, use 🤗 Datasets [`~datasets.Dataset.map`] function. You can speed up `map` by setting `batched=True` to process multiple elements of the dataset at once. Remove the columns you don't need, and rename `intent_class` to `label` because that's the name the model expects:
```py
>>> encoded_minds = minds.map(preprocess_function, remove_columns="audio", batched=True)
>>> encoded_minds = encoded_minds.rename_column("intent_class", "label")
```
## Evaluate
Including a metric during training is often helpful for evaluating your model's performance. You can quickly load a evaluation method with the 🤗 [Evaluate](https://huggingface.co/docs/evaluate/index) library. For this task, load the [accuracy](https://huggingface.co/spaces/evaluate-metric/accuracy) metric (see the 🤗 Evaluate [quick tour](https://huggingface.co/docs/evaluate/a_quick_tour) to learn more about how to load and compute a metric):
```py
>>> import evaluate
>>> accuracy = evaluate.load("accuracy")
```
Then create a function that passes your predictions and labels to [`~evaluate.EvaluationModule.compute`] to calculate the accuracy:
```py
>>> import numpy as np
>>> def compute_metrics(eval_pred):
... predictions = np.argmax(eval_pred.predictions, axis=1)
... return accuracy.compute(predictions=predictions, references=eval_pred.label_ids)
```
Your `compute_metrics` function is ready to go now, and you'll return to it when you setup your training.
## Train
<frameworkcontent>
<pt>
<Tip>
If you aren't familiar with finetuning a model with the [`Trainer`], take a look at the basic tutorial [here](../training#train-with-pytorch-trainer)!
</Tip>
You're ready to start training your model now! Load Wav2Vec2 with [`AutoModelForAudioClassification`] along with the number of expected labels, and the label mappings:
```py
>>> from transformers import AutoModelForAudioClassification, TrainingArguments, Trainer
>>> num_labels = len(id2label)
>>> model = AutoModelForAudioClassification.from_pretrained(
... "facebook/wav2vec2-base", num_labels=num_labels, label2id=label2id, id2label=id2label
... )
```
At this point, only three steps remain:
1. Define your training hyperparameters in [`TrainingArguments`]. The only required parameter is `output_dir` which specifies where to save your model. You'll push this model to the Hub by setting `push_to_hub=True` (you need to be signed in to Hugging Face to upload your model). At the end of each epoch, the [`Trainer`] will evaluate the accuracy and save the training checkpoint.
2. Pass the training arguments to [`Trainer`] along with the model, dataset, tokenizer, data collator, and `compute_metrics` function.
3. Call [`~Trainer.train`] to finetune your model.
```py
>>> training_args = TrainingArguments(
... output_dir="my_awesome_mind_model",
... evaluation_strategy="epoch",
... save_strategy="epoch",
... learning_rate=3e-5,
... per_device_train_batch_size=32,
... gradient_accumulation_steps=4,
... per_device_eval_batch_size=32,
... num_train_epochs=10,
... warmup_ratio=0.1,
... logging_steps=10,
... load_best_model_at_end=True,
... metric_for_best_model="accuracy",
... push_to_hub=True,
... )
>>> trainer = Trainer(
... model=model,
... args=training_args,
... train_dataset=encoded_minds["train"],
... eval_dataset=encoded_minds["test"],
... tokenizer=feature_extractor,
... compute_metrics=compute_metrics,
... )
>>> trainer.train()
```
Once training is completed, share your model to the Hub with the [`~transformers.Trainer.push_to_hub`] method so everyone can use your model:
```py
>>> trainer.push_to_hub()
```
</pt>
</frameworkcontent>
<Tip>
For a more in-depth example of how to finetune a model for audio classification, take a look at the corresponding [PyTorch notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/audio_classification.ipynb).
</Tip>
## Inference
Great, now that you've finetuned a model, you can use it for inference!
Load an audio file you'd like to run inference on. Remember to resample the sampling rate of the audio file to match the sampling rate of the model if you need to!
```py
>>> from datasets import load_dataset, Audio
>>> dataset = load_dataset("PolyAI/minds14", name="en-US", split="train")
>>> dataset = dataset.cast_column("audio", Audio(sampling_rate=16000))
>>> sampling_rate = dataset.features["audio"].sampling_rate
>>> audio_file = dataset[0]["audio"]["path"]
```
The simplest way to try out your finetuned model for inference is to use it in a [`pipeline`]. Instantiate a `pipeline` for audio classification with your model, and pass your audio file to it:
```py
>>> from transformers import pipeline
>>> classifier = pipeline("audio-classification", model="stevhliu/my_awesome_minds_model")
>>> classifier(audio_file)
[
{'score': 0.09766869246959686, 'label': 'cash_deposit'},
{'score': 0.07998877018690109, 'label': 'app_error'},
{'score': 0.0781070664525032, 'label': 'joint_account'},
{'score': 0.07667109370231628, 'label': 'pay_bill'},
{'score': 0.0755252093076706, 'label': 'balance'}
]
```
You can also manually replicate the results of the `pipeline` if you'd like:
<frameworkcontent>
<pt>
Load a feature extractor to preprocess the audio file and return the `input` as PyTorch tensors:
```py
>>> from transformers import AutoFeatureExtractor
>>> feature_extractor = AutoFeatureExtractor.from_pretrained("stevhliu/my_awesome_minds_model")
>>> inputs = feature_extractor(dataset[0]["audio"]["array"], sampling_rate=sampling_rate, return_tensors="pt")
```
Pass your inputs to the model and return the logits:
```py
>>> from transformers import AutoModelForAudioClassification
>>> model = AutoModelForAudioClassification.from_pretrained("stevhliu/my_awesome_minds_model")
>>> with torch.no_grad():
... logits = model(**inputs).logits
```
Get the class with the highest probability, and use the model's `id2label` mapping to convert it to a label:
```py
>>> import torch
>>> predicted_class_ids = torch.argmax(logits).item()
>>> predicted_label = model.config.id2label[predicted_class_ids]
>>> predicted_label
'cash_deposit'
```
</pt>
</frameworkcontent> | huggingface/transformers/blob/main/docs/source/en/tasks/audio_classification.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# BERTweet
## Overview
The BERTweet model was proposed in [BERTweet: A pre-trained language model for English Tweets](https://www.aclweb.org/anthology/2020.emnlp-demos.2.pdf) by Dat Quoc Nguyen, Thanh Vu, Anh Tuan Nguyen.
The abstract from the paper is the following:
*We present BERTweet, the first public large-scale pre-trained language model for English Tweets. Our BERTweet, having
the same architecture as BERT-base (Devlin et al., 2019), is trained using the RoBERTa pre-training procedure (Liu et
al., 2019). Experiments show that BERTweet outperforms strong baselines RoBERTa-base and XLM-R-base (Conneau et al.,
2020), producing better performance results than the previous state-of-the-art models on three Tweet NLP tasks:
Part-of-speech tagging, Named-entity recognition and text classification.*
This model was contributed by [dqnguyen](https://huggingface.co/dqnguyen). The original code can be found [here](https://github.com/VinAIResearch/BERTweet).
## Usage example
```python
>>> import torch
>>> from transformers import AutoModel, AutoTokenizer
>>> bertweet = AutoModel.from_pretrained("vinai/bertweet-base")
>>> # For transformers v4.x+:
>>> tokenizer = AutoTokenizer.from_pretrained("vinai/bertweet-base", use_fast=False)
>>> # For transformers v3.x:
>>> # tokenizer = AutoTokenizer.from_pretrained("vinai/bertweet-base")
>>> # INPUT TWEET IS ALREADY NORMALIZED!
>>> line = "SC has first two presumptive cases of coronavirus , DHEC confirms HTTPURL via @USER :cry:"
>>> input_ids = torch.tensor([tokenizer.encode(line)])
>>> with torch.no_grad():
... features = bertweet(input_ids) # Models outputs are now tuples
>>> # With TensorFlow 2.0+:
>>> # from transformers import TFAutoModel
>>> # bertweet = TFAutoModel.from_pretrained("vinai/bertweet-base")
```
<Tip>
This implementation is the same as BERT, except for tokenization method. Refer to [BERT documentation](bert) for
API reference information.
</Tip>
## BertweetTokenizer
[[autodoc]] BertweetTokenizer
| huggingface/transformers/blob/main/docs/source/en/model_doc/bertweet.md |
!--Copyright 2023 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Text generation strategies
Text generation is essential to many NLP tasks, such as open-ended text generation, summarization, translation, and
more. It also plays a role in a variety of mixed-modality applications that have text as an output like speech-to-text
and vision-to-text. Some of the models that can generate text include
GPT2, XLNet, OpenAI GPT, CTRL, TransformerXL, XLM, Bart, T5, GIT, Whisper.
Check out a few examples that use [`~transformers.generation_utils.GenerationMixin.generate`] method to produce
text outputs for different tasks:
* [Text summarization](./tasks/summarization#inference)
* [Image captioning](./model_doc/git#transformers.GitForCausalLM.forward.example)
* [Audio transcription](./model_doc/whisper#transformers.WhisperForConditionalGeneration.forward.example)
Note that the inputs to the generate method depend on the model's modality. They are returned by the model's preprocessor
class, such as AutoTokenizer or AutoProcessor. If a model's preprocessor creates more than one kind of input, pass all
the inputs to generate(). You can learn more about the individual model's preprocessor in the corresponding model's documentation.
The process of selecting output tokens to generate text is known as decoding, and you can customize the decoding strategy
that the `generate()` method will use. Modifying a decoding strategy does not change the values of any trainable parameters.
However, it can have a noticeable impact on the quality of the generated output. It can help reduce repetition in the text
and make it more coherent.
This guide describes:
* default generation configuration
* common decoding strategies and their main parameters
* saving and sharing custom generation configurations with your fine-tuned model on 🤗 Hub
## Default text generation configuration
A decoding strategy for a model is defined in its generation configuration. When using pre-trained models for inference
within a [`pipeline`], the models call the `PreTrainedModel.generate()` method that applies a default generation
configuration under the hood. The default configuration is also used when no custom configuration has been saved with
the model.
When you load a model explicitly, you can inspect the generation configuration that comes with it through
`model.generation_config`:
```python
>>> from transformers import AutoModelForCausalLM
>>> model = AutoModelForCausalLM.from_pretrained("distilgpt2")
>>> model.generation_config
GenerationConfig {
"bos_token_id": 50256,
"eos_token_id": 50256,
}
```
Printing out the `model.generation_config` reveals only the values that are different from the default generation
configuration, and does not list any of the default values.
The default generation configuration limits the size of the output combined with the input prompt to a maximum of 20
tokens to avoid running into resource limitations. The default decoding strategy is greedy search, which is the simplest decoding strategy that picks a token with the highest probability as the next token. For many tasks
and small output sizes this works well. However, when used to generate longer outputs, greedy search can start
producing highly repetitive results.
## Customize text generation
You can override any `generation_config` by passing the parameters and their values directly to the [`generate`] method:
```python
>>> my_model.generate(**inputs, num_beams=4, do_sample=True) # doctest: +SKIP
```
Even if the default decoding strategy mostly works for your task, you can still tweak a few things. Some of the
commonly adjusted parameters include:
- `max_new_tokens`: the maximum number of tokens to generate. In other words, the size of the output sequence, not
including the tokens in the prompt. As an alternative to using the output's length as a stopping criteria, you can choose
to stop generation whenever the full generation exceeds some amount of time. To learn more, check [`StoppingCriteria`].
- `num_beams`: by specifying a number of beams higher than 1, you are effectively switching from greedy search to
beam search. This strategy evaluates several hypotheses at each time step and eventually chooses the hypothesis that
has the overall highest probability for the entire sequence. This has the advantage of identifying high-probability
sequences that start with a lower probability initial tokens and would've been ignored by the greedy search.
- `do_sample`: if set to `True`, this parameter enables decoding strategies such as multinomial sampling, beam-search
multinomial sampling, Top-K sampling and Top-p sampling. All these strategies select the next token from the probability
distribution over the entire vocabulary with various strategy-specific adjustments.
- `num_return_sequences`: the number of sequence candidates to return for each input. This option is only available for
the decoding strategies that support multiple sequence candidates, e.g. variations of beam search and sampling. Decoding
strategies like greedy search and contrastive search return a single output sequence.
## Save a custom decoding strategy with your model
If you would like to share your fine-tuned model with a specific generation configuration, you can:
* Create a [`GenerationConfig`] class instance
* Specify the decoding strategy parameters
* Save your generation configuration with [`GenerationConfig.save_pretrained`], making sure to leave its `config_file_name` argument empty
* Set `push_to_hub` to `True` to upload your config to the model's repo
```python
>>> from transformers import AutoModelForCausalLM, GenerationConfig
>>> model = AutoModelForCausalLM.from_pretrained("my_account/my_model") # doctest: +SKIP
>>> generation_config = GenerationConfig(
... max_new_tokens=50, do_sample=True, top_k=50, eos_token_id=model.config.eos_token_id
... )
>>> generation_config.save_pretrained("my_account/my_model", push_to_hub=True) # doctest: +SKIP
```
You can also store several generation configurations in a single directory, making use of the `config_file_name`
argument in [`GenerationConfig.save_pretrained`]. You can later instantiate them with [`GenerationConfig.from_pretrained`]. This is useful if you want to
store several generation configurations for a single model (e.g. one for creative text generation with sampling, and
one for summarization with beam search). You must have the right Hub permissions to add configuration files to a model.
```python
>>> from transformers import AutoModelForSeq2SeqLM, AutoTokenizer, GenerationConfig
>>> tokenizer = AutoTokenizer.from_pretrained("t5-small")
>>> model = AutoModelForSeq2SeqLM.from_pretrained("t5-small")
>>> translation_generation_config = GenerationConfig(
... num_beams=4,
... early_stopping=True,
... decoder_start_token_id=0,
... eos_token_id=model.config.eos_token_id,
... pad_token=model.config.pad_token_id,
... )
>>> # Tip: add `push_to_hub=True` to push to the Hub
>>> translation_generation_config.save_pretrained("/tmp", "translation_generation_config.json")
>>> # You could then use the named generation config file to parameterize generation
>>> generation_config = GenerationConfig.from_pretrained("/tmp", "translation_generation_config.json")
>>> inputs = tokenizer("translate English to French: Configuration files are easy to use!", return_tensors="pt")
>>> outputs = model.generate(**inputs, generation_config=generation_config)
>>> print(tokenizer.batch_decode(outputs, skip_special_tokens=True))
['Les fichiers de configuration sont faciles à utiliser!']
```
## Streaming
The `generate()` supports streaming, through its `streamer` input. The `streamer` input is compatible with any instance
from a class that has the following methods: `put()` and `end()`. Internally, `put()` is used to push new tokens and
`end()` is used to flag the end of text generation.
<Tip warning={true}>
The API for the streamer classes is still under development and may change in the future.
</Tip>
In practice, you can craft your own streaming class for all sorts of purposes! We also have basic streaming classes
ready for you to use. For example, you can use the [`TextStreamer`] class to stream the output of `generate()` into
your screen, one word at a time:
```python
>>> from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer
>>> tok = AutoTokenizer.from_pretrained("gpt2")
>>> model = AutoModelForCausalLM.from_pretrained("gpt2")
>>> inputs = tok(["An increasing sequence: one,"], return_tensors="pt")
>>> streamer = TextStreamer(tok)
>>> # Despite returning the usual output, the streamer will also print the generated text to stdout.
>>> _ = model.generate(**inputs, streamer=streamer, max_new_tokens=20)
An increasing sequence: one, two, three, four, five, six, seven, eight, nine, ten, eleven,
```
## Decoding strategies
Certain combinations of the `generate()` parameters, and ultimately `generation_config`, can be used to enable specific
decoding strategies. If you are new to this concept, we recommend reading [this blog post that illustrates how common decoding strategies work](https://huggingface.co/blog/how-to-generate).
Here, we'll show some of the parameters that control the decoding strategies and illustrate how you can use them.
### Greedy Search
[`generate`] uses greedy search decoding by default so you don't have to pass any parameters to enable it. This means the parameters `num_beams` is set to 1 and `do_sample=False`.
```python
>>> from transformers import AutoModelForCausalLM, AutoTokenizer
>>> prompt = "I look forward to"
>>> checkpoint = "distilgpt2"
>>> tokenizer = AutoTokenizer.from_pretrained(checkpoint)
>>> inputs = tokenizer(prompt, return_tensors="pt")
>>> model = AutoModelForCausalLM.from_pretrained(checkpoint)
>>> outputs = model.generate(**inputs)
>>> tokenizer.batch_decode(outputs, skip_special_tokens=True)
['I look forward to seeing you all again!\n\n\n\n\n\n\n\n\n\n\n']
```
### Contrastive search
The contrastive search decoding strategy was proposed in the 2022 paper [A Contrastive Framework for Neural Text Generation](https://arxiv.org/abs/2202.06417).
It demonstrates superior results for generating non-repetitive yet coherent long outputs. To learn how contrastive search
works, check out [this blog post](https://huggingface.co/blog/introducing-csearch).
The two main parameters that enable and control the behavior of contrastive search are `penalty_alpha` and `top_k`:
```python
>>> from transformers import AutoTokenizer, AutoModelForCausalLM
>>> checkpoint = "gpt2-large"
>>> tokenizer = AutoTokenizer.from_pretrained(checkpoint)
>>> model = AutoModelForCausalLM.from_pretrained(checkpoint)
>>> prompt = "Hugging Face Company is"
>>> inputs = tokenizer(prompt, return_tensors="pt")
>>> outputs = model.generate(**inputs, penalty_alpha=0.6, top_k=4, max_new_tokens=100)
>>> tokenizer.batch_decode(outputs, skip_special_tokens=True)
['Hugging Face Company is a family owned and operated business. We pride ourselves on being the best
in the business and our customer service is second to none.\n\nIf you have any questions about our
products or services, feel free to contact us at any time. We look forward to hearing from you!']
```
### Multinomial sampling
As opposed to greedy search that always chooses a token with the highest probability as the
next token, multinomial sampling (also called ancestral sampling) randomly selects the next token based on the probability distribution over the entire
vocabulary given by the model. Every token with a non-zero probability has a chance of being selected, thus reducing the
risk of repetition.
To enable multinomial sampling set `do_sample=True` and `num_beams=1`.
```python
>>> from transformers import AutoTokenizer, AutoModelForCausalLM, set_seed
>>> set_seed(0) # For reproducibility
>>> checkpoint = "gpt2-large"
>>> tokenizer = AutoTokenizer.from_pretrained(checkpoint)
>>> model = AutoModelForCausalLM.from_pretrained(checkpoint)
>>> prompt = "Today was an amazing day because"
>>> inputs = tokenizer(prompt, return_tensors="pt")
>>> outputs = model.generate(**inputs, do_sample=True, num_beams=1, max_new_tokens=100)
>>> tokenizer.batch_decode(outputs, skip_special_tokens=True)
['Today was an amazing day because when you go to the World Cup and you don\'t, or when you don\'t get invited,
that\'s a terrible feeling."']
```
### Beam-search decoding
Unlike greedy search, beam-search decoding keeps several hypotheses at each time step and eventually chooses
the hypothesis that has the overall highest probability for the entire sequence. This has the advantage of identifying high-probability
sequences that start with lower probability initial tokens and would've been ignored by the greedy search.
To enable this decoding strategy, specify the `num_beams` (aka number of hypotheses to keep track of) that is greater than 1.
```python
>>> from transformers import AutoModelForCausalLM, AutoTokenizer
>>> prompt = "It is astonishing how one can"
>>> checkpoint = "gpt2-medium"
>>> tokenizer = AutoTokenizer.from_pretrained(checkpoint)
>>> inputs = tokenizer(prompt, return_tensors="pt")
>>> model = AutoModelForCausalLM.from_pretrained(checkpoint)
>>> outputs = model.generate(**inputs, num_beams=5, max_new_tokens=50)
>>> tokenizer.batch_decode(outputs, skip_special_tokens=True)
['It is astonishing how one can have such a profound impact on the lives of so many people in such a short period of
time."\n\nHe added: "I am very proud of the work I have been able to do in the last few years.\n\n"I have']
```
### Beam-search multinomial sampling
As the name implies, this decoding strategy combines beam search with multinomial sampling. You need to specify
the `num_beams` greater than 1, and set `do_sample=True` to use this decoding strategy.
```python
>>> from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, set_seed
>>> set_seed(0) # For reproducibility
>>> prompt = "translate English to German: The house is wonderful."
>>> checkpoint = "t5-small"
>>> tokenizer = AutoTokenizer.from_pretrained(checkpoint)
>>> inputs = tokenizer(prompt, return_tensors="pt")
>>> model = AutoModelForSeq2SeqLM.from_pretrained(checkpoint)
>>> outputs = model.generate(**inputs, num_beams=5, do_sample=True)
>>> tokenizer.decode(outputs[0], skip_special_tokens=True)
'Das Haus ist wunderbar.'
```
### Diverse beam search decoding
The diverse beam search decoding strategy is an extension of the beam search strategy that allows for generating a more diverse
set of beam sequences to choose from. To learn how it works, refer to [Diverse Beam Search: Decoding Diverse Solutions from Neural Sequence Models](https://arxiv.org/pdf/1610.02424.pdf).
This approach has three main parameters: `num_beams`, `num_beam_groups`, and `diversity_penalty`.
The diversity penalty ensures the outputs are distinct across groups, and beam search is used within each group.
```python
>>> from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
>>> checkpoint = "google/pegasus-xsum"
>>> prompt = (
... "The Permaculture Design Principles are a set of universal design principles "
... "that can be applied to any location, climate and culture, and they allow us to design "
... "the most efficient and sustainable human habitation and food production systems. "
... "Permaculture is a design system that encompasses a wide variety of disciplines, such "
... "as ecology, landscape design, environmental science and energy conservation, and the "
... "Permaculture design principles are drawn from these various disciplines. Each individual "
... "design principle itself embodies a complete conceptual framework based on sound "
... "scientific principles. When we bring all these separate principles together, we can "
... "create a design system that both looks at whole systems, the parts that these systems "
... "consist of, and how those parts interact with each other to create a complex, dynamic, "
... "living system. Each design principle serves as a tool that allows us to integrate all "
... "the separate parts of a design, referred to as elements, into a functional, synergistic, "
... "whole system, where the elements harmoniously interact and work together in the most "
... "efficient way possible."
... )
>>> tokenizer = AutoTokenizer.from_pretrained(checkpoint)
>>> inputs = tokenizer(prompt, return_tensors="pt")
>>> model = AutoModelForSeq2SeqLM.from_pretrained(checkpoint)
>>> outputs = model.generate(**inputs, num_beams=5, num_beam_groups=5, max_new_tokens=30, diversity_penalty=1.0)
>>> tokenizer.decode(outputs[0], skip_special_tokens=True)
'The Design Principles are a set of universal design principles that can be applied to any location, climate and
culture, and they allow us to design the'
```
This guide illustrates the main parameters that enable various decoding strategies. More advanced parameters exist for the
[`generate`] method, which gives you even further control over the [`generate`] method's behavior.
For the complete list of the available parameters, refer to the [API documentation](./main_classes/text_generation.md).
### Speculative Decoding
Speculative decoding (also known as assisted decoding) is a modification of the decoding strategies above, that uses an
assistant model (ideally a much smaller one) with the same tokenizer, to generate a few candidate tokens. The main
model then validates the candidate tokens in a single forward pass, which speeds up the decoding process. If
`do_sample=True`, then the token validation with resampling introduced in the
[speculative decoding paper](https://arxiv.org/pdf/2211.17192.pdf) is used.
Currently, only greedy search and sampling are supported with assisted decoding, and assisted decoding doesn't support batched inputs.
To learn more about assisted decoding, check [this blog post](https://huggingface.co/blog/assisted-generation).
To enable assisted decoding, set the `assistant_model` argument with a model.
```python
>>> from transformers import AutoModelForCausalLM, AutoTokenizer
>>> prompt = "Alice and Bob"
>>> checkpoint = "EleutherAI/pythia-1.4b-deduped"
>>> assistant_checkpoint = "EleutherAI/pythia-160m-deduped"
>>> tokenizer = AutoTokenizer.from_pretrained(checkpoint)
>>> inputs = tokenizer(prompt, return_tensors="pt")
>>> model = AutoModelForCausalLM.from_pretrained(checkpoint)
>>> assistant_model = AutoModelForCausalLM.from_pretrained(assistant_checkpoint)
>>> outputs = model.generate(**inputs, assistant_model=assistant_model)
>>> tokenizer.batch_decode(outputs, skip_special_tokens=True)
['Alice and Bob are sitting in a bar. Alice is drinking a beer and Bob is drinking a']
```
When using assisted decoding with sampling methods, you can use the `temperature` argument to control the randomness,
just like in multinomial sampling. However, in assisted decoding, reducing the temperature may help improve the latency.
```python
>>> from transformers import AutoModelForCausalLM, AutoTokenizer, set_seed
>>> set_seed(42) # For reproducibility
>>> prompt = "Alice and Bob"
>>> checkpoint = "EleutherAI/pythia-1.4b-deduped"
>>> assistant_checkpoint = "EleutherAI/pythia-160m-deduped"
>>> tokenizer = AutoTokenizer.from_pretrained(checkpoint)
>>> inputs = tokenizer(prompt, return_tensors="pt")
>>> model = AutoModelForCausalLM.from_pretrained(checkpoint)
>>> assistant_model = AutoModelForCausalLM.from_pretrained(assistant_checkpoint)
>>> outputs = model.generate(**inputs, assistant_model=assistant_model, do_sample=True, temperature=0.5)
>>> tokenizer.batch_decode(outputs, skip_special_tokens=True)
['Alice and Bob are going to the same party. It is a small party, in a small']
```
| huggingface/transformers/blob/main/docs/source/en/generation_strategies.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Pipelines
The pipelines are a great and easy way to use models for inference. These pipelines are objects that abstract most of
the complex code from the library, offering a simple API dedicated to several tasks, including Named Entity
Recognition, Masked Language Modeling, Sentiment Analysis, Feature Extraction and Question Answering. See the
[task summary](../task_summary) for examples of use.
There are two categories of pipeline abstractions to be aware about:
- The [`pipeline`] which is the most powerful object encapsulating all other pipelines.
- Task-specific pipelines are available for [audio](#audio), [computer vision](#computer-vision), [natural language processing](#natural-language-processing), and [multimodal](#multimodal) tasks.
## The pipeline abstraction
The *pipeline* abstraction is a wrapper around all the other available pipelines. It is instantiated as any other
pipeline but can provide additional quality of life.
Simple call on one item:
```python
>>> pipe = pipeline("text-classification")
>>> pipe("This restaurant is awesome")
[{'label': 'POSITIVE', 'score': 0.9998743534088135}]
```
If you want to use a specific model from the [hub](https://huggingface.co) you can ignore the task if the model on
the hub already defines it:
```python
>>> pipe = pipeline(model="roberta-large-mnli")
>>> pipe("This restaurant is awesome")
[{'label': 'NEUTRAL', 'score': 0.7313136458396912}]
```
To call a pipeline on many items, you can call it with a *list*.
```python
>>> pipe = pipeline("text-classification")
>>> pipe(["This restaurant is awesome", "This restaurant is awful"])
[{'label': 'POSITIVE', 'score': 0.9998743534088135},
{'label': 'NEGATIVE', 'score': 0.9996669292449951}]
```
To iterate over full datasets it is recommended to use a `dataset` directly. This means you don't need to allocate
the whole dataset at once, nor do you need to do batching yourself. This should work just as fast as custom loops on
GPU. If it doesn't don't hesitate to create an issue.
```python
import datasets
from transformers import pipeline
from transformers.pipelines.pt_utils import KeyDataset
from tqdm.auto import tqdm
pipe = pipeline("automatic-speech-recognition", model="facebook/wav2vec2-base-960h", device=0)
dataset = datasets.load_dataset("superb", name="asr", split="test")
# KeyDataset (only *pt*) will simply return the item in the dict returned by the dataset item
# as we're not interested in the *target* part of the dataset. For sentence pair use KeyPairDataset
for out in tqdm(pipe(KeyDataset(dataset, "file"))):
print(out)
# {"text": "NUMBER TEN FRESH NELLY IS WAITING ON YOU GOOD NIGHT HUSBAND"}
# {"text": ....}
# ....
```
For ease of use, a generator is also possible:
```python
from transformers import pipeline
pipe = pipeline("text-classification")
def data():
while True:
# This could come from a dataset, a database, a queue or HTTP request
# in a server
# Caveat: because this is iterative, you cannot use `num_workers > 1` variable
# to use multiple threads to preprocess data. You can still have 1 thread that
# does the preprocessing while the main runs the big inference
yield "This is a test"
for out in pipe(data()):
print(out)
# {"text": "NUMBER TEN FRESH NELLY IS WAITING ON YOU GOOD NIGHT HUSBAND"}
# {"text": ....}
# ....
```
[[autodoc]] pipeline
## Pipeline batching
All pipelines can use batching. This will work
whenever the pipeline uses its streaming ability (so when passing lists or `Dataset` or `generator`).
```python
from transformers import pipeline
from transformers.pipelines.pt_utils import KeyDataset
import datasets
dataset = datasets.load_dataset("imdb", name="plain_text", split="unsupervised")
pipe = pipeline("text-classification", device=0)
for out in pipe(KeyDataset(dataset, "text"), batch_size=8, truncation="only_first"):
print(out)
# [{'label': 'POSITIVE', 'score': 0.9998743534088135}]
# Exactly the same output as before, but the content are passed
# as batches to the model
```
<Tip warning={true}>
However, this is not automatically a win for performance. It can be either a 10x speedup or 5x slowdown depending
on hardware, data and the actual model being used.
Example where it's mostly a speedup:
</Tip>
```python
from transformers import pipeline
from torch.utils.data import Dataset
from tqdm.auto import tqdm
pipe = pipeline("text-classification", device=0)
class MyDataset(Dataset):
def __len__(self):
return 5000
def __getitem__(self, i):
return "This is a test"
dataset = MyDataset()
for batch_size in [1, 8, 64, 256]:
print("-" * 30)
print(f"Streaming batch_size={batch_size}")
for out in tqdm(pipe(dataset, batch_size=batch_size), total=len(dataset)):
pass
```
```
# On GTX 970
------------------------------
Streaming no batching
100%|██████████████████████████████████████████████████████████████████████| 5000/5000 [00:26<00:00, 187.52it/s]
------------------------------
Streaming batch_size=8
100%|█████████████████████████████████████████████████████████████████████| 5000/5000 [00:04<00:00, 1205.95it/s]
------------------------------
Streaming batch_size=64
100%|█████████████████████████████████████████████████████████████████████| 5000/5000 [00:02<00:00, 2478.24it/s]
------------------------------
Streaming batch_size=256
100%|█████████████████████████████████████████████████████████████████████| 5000/5000 [00:01<00:00, 2554.43it/s]
(diminishing returns, saturated the GPU)
```
Example where it's most a slowdown:
```python
class MyDataset(Dataset):
def __len__(self):
return 5000
def __getitem__(self, i):
if i % 64 == 0:
n = 100
else:
n = 1
return "This is a test" * n
```
This is a occasional very long sentence compared to the other. In that case, the **whole** batch will need to be 400
tokens long, so the whole batch will be [64, 400] instead of [64, 4], leading to the high slowdown. Even worse, on
bigger batches, the program simply crashes.
```
------------------------------
Streaming no batching
100%|█████████████████████████████████████████████████████████████████████| 1000/1000 [00:05<00:00, 183.69it/s]
------------------------------
Streaming batch_size=8
100%|█████████████████████████████████████████████████████████████████████| 1000/1000 [00:03<00:00, 265.74it/s]
------------------------------
Streaming batch_size=64
100%|██████████████████████████████████████████████████████████████████████| 1000/1000 [00:26<00:00, 37.80it/s]
------------------------------
Streaming batch_size=256
0%| | 0/1000 [00:00<?, ?it/s]
Traceback (most recent call last):
File "/home/nicolas/src/transformers/test.py", line 42, in <module>
for out in tqdm(pipe(dataset, batch_size=256), total=len(dataset)):
....
q = q / math.sqrt(dim_per_head) # (bs, n_heads, q_length, dim_per_head)
RuntimeError: CUDA out of memory. Tried to allocate 376.00 MiB (GPU 0; 3.95 GiB total capacity; 1.72 GiB already allocated; 354.88 MiB free; 2.46 GiB reserved in total by PyTorch)
```
There are no good (general) solutions for this problem, and your mileage may vary depending on your use cases. Rule of
thumb:
For users, a rule of thumb is:
- **Measure performance on your load, with your hardware. Measure, measure, and keep measuring. Real numbers are the
only way to go.**
- If you are latency constrained (live product doing inference), don't batch.
- If you are using CPU, don't batch.
- If you are using throughput (you want to run your model on a bunch of static data), on GPU, then:
- If you have no clue about the size of the sequence_length ("natural" data), by default don't batch, measure and
try tentatively to add it, add OOM checks to recover when it will fail (and it will at some point if you don't
control the sequence_length.)
- If your sequence_length is super regular, then batching is more likely to be VERY interesting, measure and push
it until you get OOMs.
- The larger the GPU the more likely batching is going to be more interesting
- As soon as you enable batching, make sure you can handle OOMs nicely.
## Pipeline chunk batching
`zero-shot-classification` and `question-answering` are slightly specific in the sense, that a single input might yield
multiple forward pass of a model. Under normal circumstances, this would yield issues with `batch_size` argument.
In order to circumvent this issue, both of these pipelines are a bit specific, they are `ChunkPipeline` instead of
regular `Pipeline`. In short:
```python
preprocessed = pipe.preprocess(inputs)
model_outputs = pipe.forward(preprocessed)
outputs = pipe.postprocess(model_outputs)
```
Now becomes:
```python
all_model_outputs = []
for preprocessed in pipe.preprocess(inputs):
model_outputs = pipe.forward(preprocessed)
all_model_outputs.append(model_outputs)
outputs = pipe.postprocess(all_model_outputs)
```
This should be very transparent to your code because the pipelines are used in
the same way.
This is a simplified view, since the pipeline can handle automatically the batch to ! Meaning you don't have to care
about how many forward passes you inputs are actually going to trigger, you can optimize the `batch_size`
independently of the inputs. The caveats from the previous section still apply.
## Pipeline custom code
If you want to override a specific pipeline.
Don't hesitate to create an issue for your task at hand, the goal of the pipeline is to be easy to use and support most
cases, so `transformers` could maybe support your use case.
If you want to try simply you can:
- Subclass your pipeline of choice
```python
class MyPipeline(TextClassificationPipeline):
def postprocess():
# Your code goes here
scores = scores * 100
# And here
my_pipeline = MyPipeline(model=model, tokenizer=tokenizer, ...)
# or if you use *pipeline* function, then:
my_pipeline = pipeline(model="xxxx", pipeline_class=MyPipeline)
```
That should enable you to do all the custom code you want.
## Implementing a pipeline
[Implementing a new pipeline](../add_new_pipeline)
## Audio
Pipelines available for audio tasks include the following.
### AudioClassificationPipeline
[[autodoc]] AudioClassificationPipeline
- __call__
- all
### AutomaticSpeechRecognitionPipeline
[[autodoc]] AutomaticSpeechRecognitionPipeline
- __call__
- all
### TextToAudioPipeline
[[autodoc]] TextToAudioPipeline
- __call__
- all
### ZeroShotAudioClassificationPipeline
[[autodoc]] ZeroShotAudioClassificationPipeline
- __call__
- all
## Computer vision
Pipelines available for computer vision tasks include the following.
### DepthEstimationPipeline
[[autodoc]] DepthEstimationPipeline
- __call__
- all
### ImageClassificationPipeline
[[autodoc]] ImageClassificationPipeline
- __call__
- all
### ImageSegmentationPipeline
[[autodoc]] ImageSegmentationPipeline
- __call__
- all
### ImageToImagePipeline
[[autodoc]] ImageToImagePipeline
- __call__
- all
### ObjectDetectionPipeline
[[autodoc]] ObjectDetectionPipeline
- __call__
- all
### VideoClassificationPipeline
[[autodoc]] VideoClassificationPipeline
- __call__
- all
### ZeroShotImageClassificationPipeline
[[autodoc]] ZeroShotImageClassificationPipeline
- __call__
- all
### ZeroShotObjectDetectionPipeline
[[autodoc]] ZeroShotObjectDetectionPipeline
- __call__
- all
## Natural Language Processing
Pipelines available for natural language processing tasks include the following.
### ConversationalPipeline
[[autodoc]] Conversation
[[autodoc]] ConversationalPipeline
- __call__
- all
### FillMaskPipeline
[[autodoc]] FillMaskPipeline
- __call__
- all
### QuestionAnsweringPipeline
[[autodoc]] QuestionAnsweringPipeline
- __call__
- all
### SummarizationPipeline
[[autodoc]] SummarizationPipeline
- __call__
- all
### TableQuestionAnsweringPipeline
[[autodoc]] TableQuestionAnsweringPipeline
- __call__
### TextClassificationPipeline
[[autodoc]] TextClassificationPipeline
- __call__
- all
### TextGenerationPipeline
[[autodoc]] TextGenerationPipeline
- __call__
- all
### Text2TextGenerationPipeline
[[autodoc]] Text2TextGenerationPipeline
- __call__
- all
### TokenClassificationPipeline
[[autodoc]] TokenClassificationPipeline
- __call__
- all
### TranslationPipeline
[[autodoc]] TranslationPipeline
- __call__
- all
### ZeroShotClassificationPipeline
[[autodoc]] ZeroShotClassificationPipeline
- __call__
- all
## Multimodal
Pipelines available for multimodal tasks include the following.
### DocumentQuestionAnsweringPipeline
[[autodoc]] DocumentQuestionAnsweringPipeline
- __call__
- all
### FeatureExtractionPipeline
[[autodoc]] FeatureExtractionPipeline
- __call__
- all
### ImageToTextPipeline
[[autodoc]] ImageToTextPipeline
- __call__
- all
### MaskGenerationPipeline
[[autodoc]] MaskGenerationPipeline
- __call__
- all
### VisualQuestionAnsweringPipeline
[[autodoc]] VisualQuestionAnsweringPipeline
- __call__
- all
## Parent class: `Pipeline`
[[autodoc]] Pipeline
| huggingface/transformers/blob/main/docs/source/en/main_classes/pipelines.md |
!--Copyright 2022 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# CLIPSeg
## Overview
The CLIPSeg model was proposed in [Image Segmentation Using Text and Image Prompts](https://arxiv.org/abs/2112.10003) by Timo Lüddecke
and Alexander Ecker. CLIPSeg adds a minimal decoder on top of a frozen [CLIP](clip) model for zero- and one-shot image segmentation.
The abstract from the paper is the following:
*Image segmentation is usually addressed by training a
model for a fixed set of object classes. Incorporating additional classes or more complex queries later is expensive
as it requires re-training the model on a dataset that encompasses these expressions. Here we propose a system
that can generate image segmentations based on arbitrary
prompts at test time. A prompt can be either a text or an
image. This approach enables us to create a unified model
(trained once) for three common segmentation tasks, which
come with distinct challenges: referring expression segmentation, zero-shot segmentation and one-shot segmentation.
We build upon the CLIP model as a backbone which we extend with a transformer-based decoder that enables dense
prediction. After training on an extended version of the
PhraseCut dataset, our system generates a binary segmentation map for an image based on a free-text prompt or on
an additional image expressing the query. We analyze different variants of the latter image-based prompts in detail.
This novel hybrid input allows for dynamic adaptation not
only to the three segmentation tasks mentioned above, but
to any binary segmentation task where a text or image query
can be formulated. Finally, we find our system to adapt well
to generalized queries involving affordances or properties*
<img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/model_doc/clipseg_architecture.png"
alt="drawing" width="600"/>
<small> CLIPSeg overview. Taken from the <a href="https://arxiv.org/abs/2112.10003">original paper.</a> </small>
This model was contributed by [nielsr](https://huggingface.co/nielsr).
The original code can be found [here](https://github.com/timojl/clipseg).
## Usage tips
- [`CLIPSegForImageSegmentation`] adds a decoder on top of [`CLIPSegModel`]. The latter is identical to [`CLIPModel`].
- [`CLIPSegForImageSegmentation`] can generate image segmentations based on arbitrary prompts at test time. A prompt can be either a text
(provided to the model as `input_ids`) or an image (provided to the model as `conditional_pixel_values`). One can also provide custom
conditional embeddings (provided to the model as `conditional_embeddings`).
## Resources
A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with CLIPSeg. If you're interested in submitting a resource to be included here, please feel free to open a Pull Request and we'll review it! The resource should ideally demonstrate something new instead of duplicating an existing resource.
<PipelineTag pipeline="image-segmentation"/>
- A notebook that illustrates [zero-shot image segmentation with CLIPSeg](https://github.com/NielsRogge/Transformers-Tutorials/blob/master/CLIPSeg/Zero_shot_image_segmentation_with_CLIPSeg.ipynb).
## CLIPSegConfig
[[autodoc]] CLIPSegConfig
- from_text_vision_configs
## CLIPSegTextConfig
[[autodoc]] CLIPSegTextConfig
## CLIPSegVisionConfig
[[autodoc]] CLIPSegVisionConfig
## CLIPSegProcessor
[[autodoc]] CLIPSegProcessor
## CLIPSegModel
[[autodoc]] CLIPSegModel
- forward
- get_text_features
- get_image_features
## CLIPSegTextModel
[[autodoc]] CLIPSegTextModel
- forward
## CLIPSegVisionModel
[[autodoc]] CLIPSegVisionModel
- forward
## CLIPSegForImageSegmentation
[[autodoc]] CLIPSegForImageSegmentation
- forward | huggingface/transformers/blob/main/docs/source/en/model_doc/clipseg.md |
!--Copyright 2020 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# BERT
<div class="flex flex-wrap space-x-1">
<a href="https://huggingface.co/models?filter=bert">
<img alt="Models" src="https://img.shields.io/badge/All_model_pages-bert-blueviolet">
</a>
<a href="https://huggingface.co/spaces/docs-demos/bert-base-uncased">
<img alt="Spaces" src="https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-blue">
</a>
</div>
## Overview
The BERT model was proposed in [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://arxiv.org/abs/1810.04805) by Jacob Devlin, Ming-Wei Chang, Kenton Lee and Kristina Toutanova. It's a
bidirectional transformer pretrained using a combination of masked language modeling objective and next sentence
prediction on a large corpus comprising the Toronto Book Corpus and Wikipedia.
The abstract from the paper is the following:
*We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations
from Transformers. Unlike recent language representation models, BERT is designed to pre-train deep bidirectional
representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result,
the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models
for a wide range of tasks, such as question answering and language inference, without substantial task-specific
architecture modifications.*
*BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural
language processing tasks, including pushing the GLUE score to 80.5% (7.7% point absolute improvement), MultiNLI
accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute
improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement).*
This model was contributed by [thomwolf](https://huggingface.co/thomwolf). The original code can be found [here](https://github.com/google-research/bert).
## Usage tips
- BERT is a model with absolute position embeddings so it's usually advised to pad the inputs on the right rather than
the left.
- BERT was trained with the masked language modeling (MLM) and next sentence prediction (NSP) objectives. It is
efficient at predicting masked tokens and at NLU in general, but is not optimal for text generation.
- Corrupts the inputs by using random masking, more precisely, during pretraining, a given percentage of tokens (usually 15%) is masked by:
* a special mask token with probability 0.8
* a random token different from the one masked with probability 0.1
* the same token with probability 0.1
- The model must predict the original sentence, but has a second objective: inputs are two sentences A and B (with a separation token in between). With probability 50%, the sentences are consecutive in the corpus, in the remaining 50% they are not related. The model has to predict if the sentences are consecutive or not.
## Resources
A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with BERT. If you're interested in submitting a resource to be included here, please feel free to open a Pull Request and we'll review it! The resource should ideally demonstrate something new instead of duplicating an existing resource.
<PipelineTag pipeline="text-classification"/>
- A blog post on [BERT Text Classification in a different language](https://www.philschmid.de/bert-text-classification-in-a-different-language).
- A notebook for [Finetuning BERT (and friends) for multi-label text classification](https://colab.research.google.com/github/NielsRogge/Transformers-Tutorials/blob/master/BERT/Fine_tuning_BERT_(and_friends)_for_multi_label_text_classification.ipynb).
- A notebook on how to [Finetune BERT for multi-label classification using PyTorch](https://colab.research.google.com/github/abhimishra91/transformers-tutorials/blob/master/transformers_multi_label_classification.ipynb). 🌎
- A notebook on how to [warm-start an EncoderDecoder model with BERT for summarization](https://colab.research.google.com/github/patrickvonplaten/notebooks/blob/master/BERT2BERT_for_CNN_Dailymail.ipynb).
- [`BertForSequenceClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/text-classification) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/text_classification.ipynb).
- [`TFBertForSequenceClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/tensorflow/text-classification) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/text_classification-tf.ipynb).
- [`FlaxBertForSequenceClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/flax/text-classification) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/text_classification_flax.ipynb).
- [Text classification task guide](../tasks/sequence_classification)
<PipelineTag pipeline="token-classification"/>
- A blog post on how to use [Hugging Face Transformers with Keras: Fine-tune a non-English BERT for Named Entity Recognition](https://www.philschmid.de/huggingface-transformers-keras-tf).
- A notebook for [Finetuning BERT for named-entity recognition](https://colab.research.google.com/github/NielsRogge/Transformers-Tutorials/blob/master/Custom_Named_Entity_Recognition_with_BERT_only_first_wordpiece.ipynb) using only the first wordpiece of each word in the word label during tokenization. To propagate the label of the word to all wordpieces, see this [version](https://github.com/NielsRogge/Transformers-Tutorials/blob/master/BERT/Custom_Named_Entity_Recognition_with_BERT.ipynb) of the notebook instead.
- [`BertForTokenClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/token-classification) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/token_classification.ipynb).
- [`TFBertForTokenClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/tensorflow/token-classification) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/token_classification-tf.ipynb).
- [`FlaxBertForTokenClassification`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/flax/token-classification).
- [Token classification](https://huggingface.co/course/chapter7/2?fw=pt) chapter of the 🤗 Hugging Face Course.
- [Token classification task guide](../tasks/token_classification)
<PipelineTag pipeline="fill-mask"/>
- [`BertForMaskedLM`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/language-modeling#robertabertdistilbert-and-masked-language-modeling) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/language_modeling.ipynb).
- [`TFBertForMaskedLM`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/tensorflow/language-modeling#run_mlmpy) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/language_modeling-tf.ipynb).
- [`FlaxBertForMaskedLM`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/flax/language-modeling#masked-language-modeling) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/masked_language_modeling_flax.ipynb).
- [Masked language modeling](https://huggingface.co/course/chapter7/3?fw=pt) chapter of the 🤗 Hugging Face Course.
- [Masked language modeling task guide](../tasks/masked_language_modeling)
<PipelineTag pipeline="question-answering"/>
- [`BertForQuestionAnswering`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/question-answering) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/question_answering.ipynb).
- [`TFBertForQuestionAnswering`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/tensorflow/question-answering) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/question_answering-tf.ipynb).
- [`FlaxBertForQuestionAnswering`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/flax/question-answering).
- [Question answering](https://huggingface.co/course/chapter7/7?fw=pt) chapter of the 🤗 Hugging Face Course.
- [Question answering task guide](../tasks/question_answering)
**Multiple choice**
- [`BertForMultipleChoice`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/pytorch/multiple-choice) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/multiple_choice.ipynb).
- [`TFBertForMultipleChoice`] is supported by this [example script](https://github.com/huggingface/transformers/tree/main/examples/tensorflow/multiple-choice) and [notebook](https://colab.research.google.com/github/huggingface/notebooks/blob/main/examples/multiple_choice-tf.ipynb).
- [Multiple choice task guide](../tasks/multiple_choice)
⚡️ **Inference**
- A blog post on how to [Accelerate BERT inference with Hugging Face Transformers and AWS Inferentia](https://huggingface.co/blog/bert-inferentia-sagemaker).
- A blog post on how to [Accelerate BERT inference with DeepSpeed-Inference on GPUs](https://www.philschmid.de/bert-deepspeed-inference).
⚙️ **Pretraining**
- A blog post on [Pre-Training BERT with Hugging Face Transformers and Habana Gaudi](https://www.philschmid.de/pre-training-bert-habana).
🚀 **Deploy**
- A blog post on how to [Convert Transformers to ONNX with Hugging Face Optimum](https://www.philschmid.de/convert-transformers-to-onnx).
- A blog post on how to [Setup Deep Learning environment for Hugging Face Transformers with Habana Gaudi on AWS](https://www.philschmid.de/getting-started-habana-gaudi#conclusion).
- A blog post on [Autoscaling BERT with Hugging Face Transformers, Amazon SageMaker and Terraform module](https://www.philschmid.de/terraform-huggingface-amazon-sagemaker-advanced).
- A blog post on [Serverless BERT with HuggingFace, AWS Lambda, and Docker](https://www.philschmid.de/serverless-bert-with-huggingface-aws-lambda-docker).
- A blog post on [Hugging Face Transformers BERT fine-tuning using Amazon SageMaker and Training Compiler](https://www.philschmid.de/huggingface-amazon-sagemaker-training-compiler).
- A blog post on [Task-specific knowledge distillation for BERT using Transformers & Amazon SageMaker](https://www.philschmid.de/knowledge-distillation-bert-transformers).
## BertConfig
[[autodoc]] BertConfig
- all
## BertTokenizer
[[autodoc]] BertTokenizer
- build_inputs_with_special_tokens
- get_special_tokens_mask
- create_token_type_ids_from_sequences
- save_vocabulary
<frameworkcontent>
<pt>
## BertTokenizerFast
[[autodoc]] BertTokenizerFast
</pt>
<tf>
## TFBertTokenizer
[[autodoc]] TFBertTokenizer
</tf>
</frameworkcontent>
## Bert specific outputs
[[autodoc]] models.bert.modeling_bert.BertForPreTrainingOutput
[[autodoc]] models.bert.modeling_tf_bert.TFBertForPreTrainingOutput
[[autodoc]] models.bert.modeling_flax_bert.FlaxBertForPreTrainingOutput
<frameworkcontent>
<pt>
## BertModel
[[autodoc]] BertModel
- forward
## BertForPreTraining
[[autodoc]] BertForPreTraining
- forward
## BertLMHeadModel
[[autodoc]] BertLMHeadModel
- forward
## BertForMaskedLM
[[autodoc]] BertForMaskedLM
- forward
## BertForNextSentencePrediction
[[autodoc]] BertForNextSentencePrediction
- forward
## BertForSequenceClassification
[[autodoc]] BertForSequenceClassification
- forward
## BertForMultipleChoice
[[autodoc]] BertForMultipleChoice
- forward
## BertForTokenClassification
[[autodoc]] BertForTokenClassification
- forward
## BertForQuestionAnswering
[[autodoc]] BertForQuestionAnswering
- forward
</pt>
<tf>
## TFBertModel
[[autodoc]] TFBertModel
- call
## TFBertForPreTraining
[[autodoc]] TFBertForPreTraining
- call
## TFBertModelLMHeadModel
[[autodoc]] TFBertLMHeadModel
- call
## TFBertForMaskedLM
[[autodoc]] TFBertForMaskedLM
- call
## TFBertForNextSentencePrediction
[[autodoc]] TFBertForNextSentencePrediction
- call
## TFBertForSequenceClassification
[[autodoc]] TFBertForSequenceClassification
- call
## TFBertForMultipleChoice
[[autodoc]] TFBertForMultipleChoice
- call
## TFBertForTokenClassification
[[autodoc]] TFBertForTokenClassification
- call
## TFBertForQuestionAnswering
[[autodoc]] TFBertForQuestionAnswering
- call
</tf>
<jax>
## FlaxBertModel
[[autodoc]] FlaxBertModel
- __call__
## FlaxBertForPreTraining
[[autodoc]] FlaxBertForPreTraining
- __call__
## FlaxBertForCausalLM
[[autodoc]] FlaxBertForCausalLM
- __call__
## FlaxBertForMaskedLM
[[autodoc]] FlaxBertForMaskedLM
- __call__
## FlaxBertForNextSentencePrediction
[[autodoc]] FlaxBertForNextSentencePrediction
- __call__
## FlaxBertForSequenceClassification
[[autodoc]] FlaxBertForSequenceClassification
- __call__
## FlaxBertForMultipleChoice
[[autodoc]] FlaxBertForMultipleChoice
- __call__
## FlaxBertForTokenClassification
[[autodoc]] FlaxBertForTokenClassification
- __call__
## FlaxBertForQuestionAnswering
[[autodoc]] FlaxBertForQuestionAnswering
- __call__
</jax>
</frameworkcontent>
| huggingface/transformers/blob/main/docs/source/en/model_doc/bert.md |
!--Copyright 2022 The HuggingFace Team. All rights reserved.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
specific language governing permissions and limitations under the License.
⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
rendered properly in your Markdown viewer.
-->
# Export to TorchScript
<Tip>
This is the very beginning of our experiments with TorchScript and we are still
exploring its capabilities with variable-input-size models. It is a focus of interest to
us and we will deepen our analysis in upcoming releases, with more code examples, a more
flexible implementation, and benchmarks comparing Python-based codes with compiled
TorchScript.
</Tip>
According to the [TorchScript documentation](https://pytorch.org/docs/stable/jit.html):
> TorchScript is a way to create serializable and optimizable models from PyTorch code.
There are two PyTorch modules, [JIT and
TRACE](https://pytorch.org/docs/stable/jit.html), that allow developers to export their
models to be reused in other programs like efficiency-oriented C++ programs.
We provide an interface that allows you to export 🤗 Transformers models to TorchScript
so they can be reused in a different environment than PyTorch-based Python programs.
Here, we explain how to export and use our models using TorchScript.
Exporting a model requires two things:
- model instantiation with the `torchscript` flag
- a forward pass with dummy inputs
These necessities imply several things developers should be careful about as detailed
below.
## TorchScript flag and tied weights
The `torchscript` flag is necessary because most of the 🤗 Transformers language models
have tied weights between their `Embedding` layer and their `Decoding` layer.
TorchScript does not allow you to export models that have tied weights, so it is
necessary to untie and clone the weights beforehand.
Models instantiated with the `torchscript` flag have their `Embedding` layer and
`Decoding` layer separated, which means that they should not be trained down the line.
Training would desynchronize the two layers, leading to unexpected results.
This is not the case for models that do not have a language model head, as those do not
have tied weights. These models can be safely exported without the `torchscript` flag.
## Dummy inputs and standard lengths
The dummy inputs are used for a models forward pass. While the inputs' values are
propagated through the layers, PyTorch keeps track of the different operations executed
on each tensor. These recorded operations are then used to create the *trace* of the
model.
The trace is created relative to the inputs' dimensions. It is therefore constrained by
the dimensions of the dummy input, and will not work for any other sequence length or
batch size. When trying with a different size, the following error is raised:
```
`The expanded size of the tensor (3) must match the existing size (7) at non-singleton dimension 2`
```
We recommended you trace the model with a dummy input size at least as large as the
largest input that will be fed to the model during inference. Padding can help fill the
missing values. However, since the model is traced with a larger input size, the
dimensions of the matrix will also be large, resulting in more calculations.
Be careful of the total number of operations done on each input and follow the
performance closely when exporting varying sequence-length models.
## Using TorchScript in Python
This section demonstrates how to save and load models as well as how to use the trace
for inference.
### Saving a model
To export a `BertModel` with TorchScript, instantiate `BertModel` from the `BertConfig`
class and then save it to disk under the filename `traced_bert.pt`:
```python
from transformers import BertModel, BertTokenizer, BertConfig
import torch
enc = BertTokenizer.from_pretrained("bert-base-uncased")
# Tokenizing input text
text = "[CLS] Who was Jim Henson ? [SEP] Jim Henson was a puppeteer [SEP]"
tokenized_text = enc.tokenize(text)
# Masking one of the input tokens
masked_index = 8
tokenized_text[masked_index] = "[MASK]"
indexed_tokens = enc.convert_tokens_to_ids(tokenized_text)
segments_ids = [0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1]
# Creating a dummy input
tokens_tensor = torch.tensor([indexed_tokens])
segments_tensors = torch.tensor([segments_ids])
dummy_input = [tokens_tensor, segments_tensors]
# Initializing the model with the torchscript flag
# Flag set to True even though it is not necessary as this model does not have an LM Head.
config = BertConfig(
vocab_size_or_config_json_file=32000,
hidden_size=768,
num_hidden_layers=12,
num_attention_heads=12,
intermediate_size=3072,
torchscript=True,
)
# Instantiating the model
model = BertModel(config)
# The model needs to be in evaluation mode
model.eval()
# If you are instantiating the model with *from_pretrained* you can also easily set the TorchScript flag
model = BertModel.from_pretrained("bert-base-uncased", torchscript=True)
# Creating the trace
traced_model = torch.jit.trace(model, [tokens_tensor, segments_tensors])
torch.jit.save(traced_model, "traced_bert.pt")
```
### Loading a model
Now you can load the previously saved `BertModel`, `traced_bert.pt`, from disk and use
it on the previously initialised `dummy_input`:
```python
loaded_model = torch.jit.load("traced_bert.pt")
loaded_model.eval()
all_encoder_layers, pooled_output = loaded_model(*dummy_input)
```
### Using a traced model for inference
Use the traced model for inference by using its `__call__` dunder method:
```python
traced_model(tokens_tensor, segments_tensors)
```
## Deploy Hugging Face TorchScript models to AWS with the Neuron SDK
AWS introduced the [Amazon EC2 Inf1](https://aws.amazon.com/ec2/instance-types/inf1/)
instance family for low cost, high performance machine learning inference in the cloud.
The Inf1 instances are powered by the AWS Inferentia chip, a custom-built hardware
accelerator, specializing in deep learning inferencing workloads. [AWS
Neuron](https://awsdocs-neuron.readthedocs-hosted.com/en/latest/#) is the SDK for
Inferentia that supports tracing and optimizing transformers models for deployment on
Inf1. The Neuron SDK provides:
1. Easy-to-use API with one line of code change to trace and optimize a TorchScript
model for inference in the cloud.
2. Out of the box performance optimizations for [improved
cost-performance](https://awsdocs-neuron.readthedocs-hosted.com/en/latest/neuron-guide/benchmark/>).
3. Support for Hugging Face transformers models built with either
[PyTorch](https://awsdocs-neuron.readthedocs-hosted.com/en/latest/src/examples/pytorch/bert_tutorial/tutorial_pretrained_bert.html)
or
[TensorFlow](https://awsdocs-neuron.readthedocs-hosted.com/en/latest/src/examples/tensorflow/huggingface_bert/huggingface_bert.html).
### Implications
Transformers models based on the [BERT (Bidirectional Encoder Representations from
Transformers)](https://huggingface.co/docs/transformers/main/model_doc/bert)
architecture, or its variants such as
[distilBERT](https://huggingface.co/docs/transformers/main/model_doc/distilbert) and
[roBERTa](https://huggingface.co/docs/transformers/main/model_doc/roberta) run best on
Inf1 for non-generative tasks such as extractive question answering, sequence
classification, and token classification. However, text generation tasks can still be
adapted to run on Inf1 according to this [AWS Neuron MarianMT
tutorial](https://awsdocs-neuron.readthedocs-hosted.com/en/latest/src/examples/pytorch/transformers-marianmt.html).
More information about models that can be converted out of the box on Inferentia can be
found in the [Model Architecture
Fit](https://awsdocs-neuron.readthedocs-hosted.com/en/latest/neuron-guide/models/models-inferentia.html#models-inferentia)
section of the Neuron documentation.
### Dependencies
Using AWS Neuron to convert models requires a [Neuron SDK
environment](https://awsdocs-neuron.readthedocs-hosted.com/en/latest/neuron-guide/neuron-frameworks/pytorch-neuron/index.html#installation-guide)
which comes preconfigured on [AWS Deep Learning
AMI](https://docs.aws.amazon.com/dlami/latest/devguide/tutorial-inferentia-launching.html).
### Converting a model for AWS Neuron
Convert a model for AWS NEURON using the same code from [Using TorchScript in
Python](torchscript#using-torchscript-in-python) to trace a `BertModel`. Import the
`torch.neuron` framework extension to access the components of the Neuron SDK through a
Python API:
```python
from transformers import BertModel, BertTokenizer, BertConfig
import torch
import torch.neuron
```
You only need to modify the following line:
```diff
- torch.jit.trace(model, [tokens_tensor, segments_tensors])
+ torch.neuron.trace(model, [token_tensor, segments_tensors])
```
This enables the Neuron SDK to trace the model and optimize it for Inf1 instances.
To learn more about AWS Neuron SDK features, tools, example tutorials and latest
updates, please see the [AWS NeuronSDK
documentation](https://awsdocs-neuron.readthedocs-hosted.com/en/latest/index.html).
| huggingface/transformers/blob/main/docs/source/en/torchscript.md |