fix conversation.py

.gitattributes

@@ -1,20 +1,35 @@
-model-00001-of-00017.safetensors filter=lfs diff=lfs merge=lfs -text
-model-00002-of-00017.safetensors filter=lfs diff=lfs merge=lfs -text
-model-00003-of-00017.safetensors filter=lfs diff=lfs merge=lfs -text
-model-00004-of-00017.safetensors filter=lfs diff=lfs merge=lfs -text
-model-00005-of-00017.safetensors filter=lfs diff=lfs merge=lfs -text
-model-00006-of-00017.safetensors filter=lfs diff=lfs merge=lfs -text
-model-00007-of-00017.safetensors filter=lfs diff=lfs merge=lfs -text
-model-00008-of-00017.safetensors filter=lfs diff=lfs merge=lfs -text
-model-00009-of-00017.safetensors filter=lfs diff=lfs merge=lfs -text
-model-00010-of-00017.safetensors filter=lfs diff=lfs merge=lfs -text
-model-00011-of-00017.safetensors filter=lfs diff=lfs merge=lfs -text
-model-00012-of-00017.safetensors filter=lfs diff=lfs merge=lfs -text
-model-00013-of-00017.safetensors filter=lfs diff=lfs merge=lfs -text
-model-00014-of-00017.safetensors filter=lfs diff=lfs merge=lfs -text
-model-00015-of-00017.safetensors filter=lfs diff=lfs merge=lfs -text
-model-00016-of-00017.safetensors filter=lfs diff=lfs merge=lfs -text
-model-00017-of-00017.safetensors filter=lfs diff=lfs merge=lfs -text
-tokenizer.model filter=lfs diff=lfs merge=lfs -text
-examples/image2.jpg filter=lfs diff=lfs merge=lfs -text
-examples/red-panda.mp4 filter=lfs diff=lfs merge=lfs -text
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,521 +1,151 @@
 ---
 license: mit
-pipeline_tag: image-text-to-text
-library_name: transformers
-base_model:
-  - OpenGVLab/InternViT-6B-448px-V1-2
-  - NousResearch/Nous-Hermes-2-Yi-34B
-new_version: OpenGVLab/InternVL2_5-38B
-base_model_relation: merge
-language:
-  - multilingual
-tags:
-  - internvl
-  - custom_code
+datasets:
+- laion/laion2B-en
+- laion/laion-coco
+- laion/laion2B-multi
+- kakaobrain/coyo-700m
+- conceptual_captions
+- wanng/wukong100m
 ---
 
-# InternVL-Chat-V1-2
+# Model Card for InternVL-Chat-Chinese-V1.2
 
-[\[📂 GitHub\]](https://github.com/OpenGVLab/InternVL)  [\[📜 InternVL 1.0\]](https://huggingface.co/papers/2312.14238)  [\[📜 InternVL 1.5\]](https://huggingface.co/papers/2404.16821)  [\[📜 Mini-InternVL\]](https://arxiv.org/abs/2410.16261)  [\[📜 InternVL 2.5\]](https://huggingface.co/papers/2412.05271)
+## What is InternVL?
 
-[\[🆕 Blog\]](https://internvl.github.io/blog/)  [\[🗨️ Chat Demo\]](https://internvl.opengvlab.com/)  [\[🤗 HF Demo\]](https://huggingface.co/spaces/OpenGVLab/InternVL)  [\[🚀 Quick Start\]](#quick-start)  [\[📖 Documents\]](https://internvl.readthedocs.io/en/latest/)
+\[[Paper](https://arxiv.org/abs/2312.14238)\]  \[[GitHub](https://github.com/OpenGVLab/InternVL)\] \[[Chat Demo](https://internvl.opengvlab.com/)\]
 
-<div align="center">
-  <img width="500" alt="image" src="https://cdn-uploads.huggingface.co/production/uploads/64006c09330a45b03605bba3/zJsd2hqd3EevgXo6fNgC-.png">
-</div>
+InternVL scales up the ViT to _**6B parameters**_ and aligns it with LLM.
 
-## Introduction
+## InternVL-Chat-V1.2 Blog
 
-We are excited to introduce [🤗 InternVL-Chat-V1-2](https://huggingface.co/OpenGVLab/InternVL-Chat-V1-2). Inspired by [LLaVA-NeXT-34B](https://llava-vl.github.io/blog/2024-01-30-llava-next/), we have also adopted [Nous-Hermes-2-Yi-34B](https://huggingface.co/NousResearch/Nous-Hermes-2-Yi-34B) as the language model. Below is the pipeline.
+> Date: 2024/02/12<br>
+> Developed by: Zhe Chen, Weiyun Wang, Wenhai Wang, Erfei Cui, Zhangwei Gao, Xizhou Zhu, Lewei Lu, Tong Lu, Yu Qiao, Jifeng Dai
 
-<p align="center">
-  <img width="600" alt="image" src="https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/GIEKCvNc1Y5iMQqLv645p.png">
-</p>
+We are excited to introduce InternVL-Chat-V1.2. Inspired by [LLaVA-NeXT-34B](https://llava-vl.github.io/blog/2024-01-30-llava-next/), we have also adopted [Nous-Hermes-2-Yi-34B](https://huggingface.co/NousResearch/Nous-Hermes-2-Yi-34B) as the language model. Below is the pipeline.
 
-From the experimental results, we've observed that **a stronger language model (34B) can better leverage the powerful capabilities of our vision foundation model.**
+<img width="600" alt="image" src="https://cdn-uploads.huggingface.co/production/uploads/64119264f0f81eb569e0d569/GIEKCvNc1Y5iMQqLv645p.png">
 
-For better training reproducibility, we follow the minimalist design and data efficiency similar to LLaVA-NeXT. To reduce training costs, we provide a [pre-trained MLP projector](https://huggingface.co/OpenGVLab/InternViT-6B-448px-V1-2/blob/main/mlp_projector/hermes_2_yi_34b.pth) and only employ around 1.2 million visual instruction tuning samples for SFT. Our model has a total of 40 billion parameters and can be trained within 1.5 days using 32 A100 GPUs. The code, data, and model have been made publicly available.
+From the experimental results, **we've observed that a stronger language model (34B) can better leverage the powerful capabilities of our vision foundation model ([InternViT-6B](https://huggingface.co/OpenGVLab/InternViT-6B-448px-V1-2)).**
 
-## Model Details
-
-- **Model Type:** multimodal large language model (MLLM)
-
-- **Model Stats:**
-
-  - Architecture: [InternViT-6B-448px-V1-2](https://huggingface.co/OpenGVLab/InternViT-6B-448px-V1-2) + MLP + [Nous-Hermes-2-Yi-34B](https://huggingface.co/NousResearch/Nous-Hermes-2-Yi-34B)
-  - Image size: 448 x 448 (256 tokens)
-  - Params: 40B
-
-- **Training Strategy:**
-
-  - Pre-training Stage
-    - Learnable Component: ViT + MLP
-    - Data: Trained on 8192x4800=39.3M samples, including COYO, LAION, CC12M, CC3M, SBU, Wukong, GRIT, Objects365, OpenImages, and OCR data.
-    - Note: In this stage, we first load the pre-trained weights of [InternViT-6B-448px-V1-0](https://huggingface.co/OpenGVLab/InternViT-6B-448px-V1-0) and connect it to Nous-Hermes-2-Yi-34B. After pre-training, the extracted ViT is published as [InternViT-6B-448px-V1-2](https://huggingface.co/OpenGVLab/InternViT-6B-448px-V1-2). Moreover, in order to reduce the number of visual tokens, we use a pixel shuffle to reduce 1024 tokens to 256 tokens.
-  - Supervised Fine-tuning Stage
-    - Learnable Component: ViT + MLP + LLM
-    - Data: A simplified, fully open-source dataset, containing approximately 1.2 million samples. You can download it from [here](https://huggingface.co/datasets/OpenGVLab/InternVL-Chat-V1-2-SFT-Data).
-
-## Performance
-
-\* Proprietary Model
-
-| name                   | image size | MMMU<br>(val) | MMMU<br>(test) | MathVista<br>(testmini) | MMB<br>(test) | MMB−CN<br>(test) | MMVP | MME      | ScienceQA<br>(image) | POPE | TextVQA<br>(val) | SEEDv1<br>(image) | VizWiz<br>(test) | GQA<br>(test) |
-| ---------------------- | ---------- | ------------- | -------------- | ----------------------- | ------------- | ---------------- | ---- | -------- | -------------------- | ---- | ---------------- | ----------------- | ---------------- | ------------- |
-| GPT−4V\*               | unknown    | 56.8          | 55.7           | 49.9                    | 77.0          | 74.4             | 38.7 | 1409/517 | -                    | -    | 78.0             | 71.6              | -                | -             |
-| Gemini Ultra\*         | unknown    | 59.4          | -              | 53.0                    | -             | -                | -    | -        | -                    | -    | 82.3             | -                 | -                | -             |
-| Gemini Pro\*           | unknown    | 47.9          | -              | 45.2                    | 73.6          | 74.3             | 40.7 | 1497/437 | -                    | -    | 74.6             | 70.7              | -                | -             |
-| Qwen−VL−Plus\*         | unknown    | 45.2          | 40.8           | 43.3                    | 67.0          | 70.7             | -    | 1681/502 | -                    | -    | 78.9             | 65.7              | -                | -             |
-| Qwen−VL−Max\*          | unknown    | 51.4          | 46.8           | 51.0                    | 77.6          | 75.7             | -    | -        | -                    | -    | 79.5             | -                 | -                | -             |
-|                        |            |               |                |                         |               |                  |      |          |                      |      |                  |                   |                  |               |
-| LLaVA−NEXT−34B         | 672x672    | 51.1          | 44.7           | 46.5                    | 79.3          | 79.0             | -    | 1631/397 | 81.8                 | 87.7 | 69.5             | 75.9              | 63.8             | 67.1          |
-| InternVL−Chat<br>−V1-2 | 448x448    | 51.6          | 46.2           | 47.7                    | 82.2          | 81.2             | 56.7 | 1687/489 | 83.3                 | 88.0 | 72.5             | 75.6              | 60.0             | 64.0          |
-
-- In most benchmarks, InternVL-Chat-V1-2 achieves better performance than LLaVA-NeXT-34B.
-
-Here, we have conducted only a simple performance comparison. For more detailed performance information and additional evaluation metrics, please refer to our performance summary table.
-
-## Training Details
+For better training reproducibility, we follow the minimalist design and data efficiency similar to LLaVA-NeXT. To reduce training costs, we provide a pre-trained MLP projector and only employ around 1 million visual instruction tuning samples for SFT. Our model has a total of 40 billion parameters and can be trained within 1.5 days using 32 A100 GPUs. The code, data, and model will be made publicly available.
 
 ### Data Preparation
 
-Inspired by LLaVA-NeXT, we adopted a data-efficient SFT strategy to train InternVL-Chat-V1-2, utilizing approximately 1.2M of visual instruction tuning samples in total, all of which are fully open-source. In a macro sense, we build upon [ShareGPT-4V](https://github.com/InternLM/InternLM-XComposer/blob/main/projects/ShareGPT4V/docs/Data.md#prepare-images) and additionally integrate [LLaVA-ZH](https://huggingface.co/datasets/openbmb/llava_zh), [DVQA](https://github.com/kushalkafle/DVQA_dataset), [ChartQA](https://github.com/vis-nlp/ChartQA), [AI2D](https://allenai.org/data/diagrams), [DocVQA](https://www.docvqa.org/datasets), [GeoQA+](https://github.com/SCNU203/GeoQA-Plus), and [SynthDoG-EN](https://huggingface.co/datasets/naver-clova-ix/synthdog-en). Most of the data remains consistent with LLaVA-NeXT.
+Inspired by LLaVA-NeXT, we adopted a data-efficient SFT strategy to train InternVL-Chat-V1.2, utilizing approximately 1.2M of visual instruction tuning samples in total, all of which are fully open-source. In a macro sense, we build upon [ShareGPT-4V](https://github.com/InternLM/InternLM-XComposer/blob/main/projects/ShareGPT4V/docs/Data.md#prepare-images) and additionally integrate [LLaVA-ZH](https://huggingface.co/datasets/openbmb/llava_zh), [DVQA](https://github.com/kushalkafle/DVQA_dataset), [ChartQA](https://github.com/vis-nlp/ChartQA), [AI2D](https://allenai.org/data/diagrams), [DocVQA](https://www.docvqa.org/datasets), [GeoQA+](https://github.com/SCNU203/GeoQA-Plus), and [SynthDoG-EN](https://huggingface.co/datasets/naver-clova-ix/synthdog-en). Most of the data remains consistent with LLaVA-NeXT.
 
-Now, you can download these datasets directly from [HuggingFace](https://huggingface.co/datasets/OpenGVLab/InternVL-Chat-V1-2-SFT-Data). For more details about data preparation, please see [here](https://github.com/OpenGVLab/InternVL/tree/main/internvl_chat#prepare-training-datasets).
-
-### Training (Supervised Fine-tuning)
-
-We provide [slurm scripts](https://github.com/OpenGVLab/InternVL/blob/main/internvl_chat/shell/internvl1.2/hermes2_yi34b/internvl_chat_v1_2_hermes2_yi34b_448_res_finetune.sh) for multi-node multi-GPU training. You can use either 32 or 64 GPUs to train this model. If you use 64 GPUs, training will take approximately 18 hours.
-
-For more details about training, please see [here](https://internvl.readthedocs.io/en/latest/internvl1.2/reproduce.html).
-
-The hyperparameters used for fine-tuning are listed in the following table.
-
-| Hyperparameter         | Trainable Param  | Global Batch Size | Learning rate | Epochs | Max length | Weight decay |
-| ---------------------- | ---------------- | ----------------- | ------------- | ------ | ---------- | ------------ |
-| InternVL−Chat<br>−V1-2 | 40B (full model) | 512               | 1e-5          | 1      | 2048       | 0.05         |
-
-## Quick Start
-
-We provide an example code to run InternVL-Chat-V1-2 using `transformers`.
-
-> Please use transformers>=4.37.2 to ensure the model works normally.
-
-### Model Loading
-
-#### 16-bit (bf16 / fp16)
-
-```python
-import torch
-from transformers import AutoTokenizer, AutoModel
-path = "OpenGVLab/InternVL-Chat-V1-2"
-model = AutoModel.from_pretrained(
-    path,
-    torch_dtype=torch.bfloat16,
-    low_cpu_mem_usage=True,
-    use_flash_attn=True,
-    trust_remote_code=True).eval().cuda()
-```
+For more details about data preparation, please see [here](https://github.com/OpenGVLab/InternVL/tree/main/internvl_chat#prepare-training-datasets).
 
-#### BNB 8-bit Quantization
+### Performance
 
-```python
-import torch
-from transformers import AutoTokenizer, AutoModel
-path = "OpenGVLab/InternVL-Chat-V1-2"
-model = AutoModel.from_pretrained(
-    path,
-    torch_dtype=torch.bfloat16,
-    load_in_8bit=True,
-    low_cpu_mem_usage=True,
-    use_flash_attn=True,
-    trust_remote_code=True).eval()
-```
-
-#### BNB 4-bit Quantization
-
-> **⚠️ Warning:** Due to significant quantization errors with BNB 4-bit quantization on InternViT-6B, the model may produce nonsensical outputs and fail to understand images. Therefore, please avoid using BNB 4-bit quantization.
-
-#### Multiple GPUs
-
-The reason for writing the code this way is to avoid errors that occur during multi-GPU inference due to tensors not being on the same device. By ensuring that the first and last layers of the large language model (LLM) are on the same device, we prevent such errors.
+\* Proprietary Model
 
-```python
-import math
-import torch
-from transformers import AutoTokenizer, AutoModel
-
-def split_model(model_name):
-    device_map = {}
-    world_size = torch.cuda.device_count()
-    num_layers = {'InternVL-Chat-V1-2': 60, 'InternVL-Chat-V1-2-Plus': 60}[model_name]
-    # Since the first GPU will be used for ViT, treat it as half a GPU.
-    num_layers_per_gpu = math.ceil(num_layers / (world_size - 0.5))
-    num_layers_per_gpu = [num_layers_per_gpu] * world_size
-    num_layers_per_gpu[0] = math.ceil(num_layers_per_gpu[0] * 0.5)
-    layer_cnt = 0
-    for i, num_layer in enumerate(num_layers_per_gpu):
-        for j in range(num_layer):
-            device_map[f'language_model.model.layers.{layer_cnt}'] = i
-            layer_cnt += 1
-    device_map['vision_model'] = 0
-    device_map['mlp1'] = 0
-    device_map['language_model.model.tok_embeddings'] = 0
-    device_map['language_model.model.embed_tokens'] = 0
-    device_map['language_model.output'] = 0
-    device_map['language_model.model.norm'] = 0
-    device_map['language_model.model.rotary_emb'] = 0
-    device_map['language_model.lm_head'] = 0
-    device_map[f'language_model.model.layers.{num_layers - 1}'] = 0
-
-    return device_map
-
-path = "OpenGVLab/InternVL-Chat-V1-2"
-device_map = split_model('InternVL-Chat-V1-2')
-model = AutoModel.from_pretrained(
-    path,
-    torch_dtype=torch.bfloat16,
-    low_cpu_mem_usage=True,
-    use_flash_attn=True,
-    trust_remote_code=True,
-    device_map=device_map).eval()
-```
+| name               | image size | MMMU<br>(val) | MMMU<br>(test) | MathVista<br>(testmini) | MMB<br>(test) | MMB−CN<br>(test) | MMVP | MME      | ScienceQA<br>(image) | POPE | TextVQA | SEEDv1<br>(image) | VizWiz<br>(test) | GQA<br>(test) |
+| ------------------ | ---------- | ------------- | -------------- | ----------------------- | ------------- | ---------------- | ---- | -------- | -------------------- | ---- | ------- | ----------------- | ---------------- | ------------- |
+| GPT-4V\*           | unknown    | 56.8          | 55.7           | 49.9                    | 77.0          | 74.4             | 38.7 | 1409/517 | -                    | -    | 78.0    | 71.6              | -                | -             |
+| Gemini Ultra\*     | unknown    | 59.4          | -              | 53.0                    | -             | -                | -    | -        | -                    | -    | 82.3    | -                 | -                | -             |
+| Gemini Pro\*       | unknown    | 47.9          | -              | 45.2                    | 73.6          | 74.3             | 40.7 | 1497/437 | -                    | -    | 74.6    | 70.7              | -                | -             |
+| Qwen-VL-Plus\*     | unknown    | 45.2          | 40.8           | 43.3                    | 67.0          | 70.7             | -    | 1681/502 | -                    | -    | 78.9    | 65.7              | -                | -             |
+| Qwen-VL-Max\*      | unknown    | 51.4          | 46.8           | 51.0                    | 77.6          | 75.7             | -    | -        | -                    | -    | 79.5    | -                 | -                | -             |
+|                    |            |               |                |                         |               |                  |      |          |                      |      |         |                   |                  |               |
+| LLaVA-NEXT-34B     | 672x672    | 51.1          | 44.7           | 46.5                    | 79.3          | 79.0             | -    | 1631/397 | 81.8                 | 87.7 | 69.5    | 75.9              | 63.8             | 67.1          |
+| InternVL-Chat-V1.2 | 448x448    | 51.6          | 46.2           | 47.7                    | 82.2          | 81.2             | 56.7 | 1672/509 | 83.3                 | 88.0 | 69.7    | 75.6              | 60.0             | 64.0          |
 
-### Inference with Transformers
+- MMBench results are collected from the [leaderboard](https://mmbench.opencompass.org.cn/leaderboard).
+- In most benchmarks, InternVL-Chat-V1.2 achieves better performance than LLaVA-NeXT-34B.
 
-#### Pure-text conversation
+### Training (SFT)
 
-```python
-from transformers import AutoTokenizer, AutoModel
-import torch
+We provide [slurm scripts](https://github.com/OpenGVLab/InternVL/tree/main/internvl_chat/shell/hermes2_yi34b/internvl_chat_v1_2_hermes2_yi34b_448_finetune.sh) for multi-node multi-GPU training. You can use either 32 or 64 GPUs to train this model. If you use 64 GPUs, training will take approximately 18 hours.
 
-path = "OpenGVLab/InternVL-Chat-V1-2"
-model = AutoModel.from_pretrained(
-    path,
-    torch_dtype=torch.bfloat16,
-    low_cpu_mem_usage=True,
-    use_flash_attn=True,
-    trust_remote_code=True).eval().cuda()
-tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True, use_fast=False)
-
-generation_config = dict(max_new_tokens=1024, do_sample=True)
-question = 'Hello, who are you?'
-response, history = model.chat(tokenizer, None, question, generation_config, history=None, return_history=True)
-print(f'User: {question}')
-print(f'Assistant: {response}')
-
-question = 'Can you tell me a story?'
-response, history = model.chat(tokenizer, None, question, generation_config, history=history, return_history=True)
-print(f'User: {question}')
-print(f'Assistant: {response}')
-```
+For more details about training, please see [here](https://github.com/OpenGVLab/InternVL/tree/main/internvl_chat#start-training).
 
-#### Single-image single-round conversation
+The hyperparameters used for finetuning are listed in the following table.
 
-```python
-from transformers import AutoTokenizer, AutoModel, CLIPImageProcessor
-from PIL import Image
-import torch
+| Hyperparameter     | Trainable Param  | Global Batch Size | Learning rate | Epochs | Max length | Weight decay |
+| ------------------ | ---------------- | ----------------- | ------------- | ------ | ---------- | ------------ |
+| InternVL-Chat-V1.2 | 40B (full model) | 512               | 1e-5          | 1      | 2048       | 0.05         |
 
-path = "OpenGVLab/InternVL-Chat-V1-2"
-model = AutoModel.from_pretrained(
-    path,
-    torch_dtype=torch.bfloat16,
-    low_cpu_mem_usage=True,
-    use_flash_attn=True,
-    trust_remote_code=True).eval().cuda()
-tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True, use_fast=False)
 
-image_processor = CLIPImageProcessor.from_pretrained(path)
-image = Image.open('./examples/image2.jpg').resize((448, 448))
-pixel_values = image_processor(images=image, return_tensors='pt').pixel_values.to(torch.bfloat16).cuda()
+## Model Details
+- **Model Type:** vision large language model, multimodal chatbot
+- **Model Stats:**
+  - Architecture: [InternViT-6B-448px-V1-2](https://huggingface.co/OpenGVLab/InternViT-6B-448px-V1-2) + MLP + [Nous-Hermes-2-Yi-34B](https://huggingface.co/NousResearch/Nous-Hermes-2-Yi-34B)
+  - Params: 40B
+  - Image size: 448 x 448
+  - Number of visual tokens: 256
 
-generation_config = dict(max_new_tokens=1024, do_sample=True)
-question = '<image>\nPlease describe the image shortly.'
-response = model.chat(tokenizer, pixel_values, question, generation_config)
-print(f'User: {question}')
-print(f'Assistant: {response}')
-```
+- **Training Strategy:**
+  - Pretraining Stage
+    - Learnable Component: MLP
+    - Data: Trained on 8192x4800=39.3M samples, including COYO, LAION, CC12M, CC3M, SBU, Wukong, GRIT, Objects365, OpenImages, and OCR data.
+    - Note: In this stage, we load the pretrained weights of [InternViT-6B-448px-V1-2](https://huggingface.co/OpenGVLab/InternViT-6B-448px-V1-2). Moreover, in order to reduce the number of visual tokens, we use a pixel shuffle to reduce 1024 tokens to 256 tokens.
+  - SFT Stage
+    - Learnable Component: ViT + MLP + LLM
+    - Data: A simplified, fully open-source dataset, containing approximately 1 million samples.
 
-#### Single-image multi-round conversation
 
-```python
-from transformers import AutoTokenizer, AutoModel, CLIPImageProcessor
-from PIL import Image
-import torch
+## Model Usage
 
-path = "OpenGVLab/InternVL-Chat-V1-2"
-model = AutoModel.from_pretrained(
-    path,
-    torch_dtype=torch.bfloat16,
-    low_cpu_mem_usage=True,
-    use_flash_attn=True,
-    trust_remote_code=True).eval().cuda()
-tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True, use_fast=False)
+We provide a minimum code example to run InternVL-Chat using only the `transformers` library.
 
-image_processor = CLIPImageProcessor.from_pretrained(path)
-image = Image.open('./examples/image2.jpg').resize((448, 448))
-pixel_values = image_processor(images=image, return_tensors='pt').pixel_values.to(torch.bfloat16).cuda()
-
-generation_config = dict(max_new_tokens=1024, do_sample=True)
-question = '<image>\nPlease describe the image in detail.'
-response, history = model.chat(tokenizer, pixel_values, question, generation_config, history=None, return_history=True)
-print(f'User: {question}')
-print(f'Assistant: {response}')
-
-question = 'Please write a poem according to the image.'
-response, history = model.chat(tokenizer, pixel_values, question, generation_config, history=history, return_history=True)
-print(f'User: {question}')
-print(f'Assistant: {response}')
-```
+You also can use our [online demo](https://internvl.opengvlab.com/) for a quick experience of this model.
 
-#### Multi-image multi-round conversation, combined images
+Note: If you meet this error `ImportError: This modeling file requires the following packages that were not found in your environment: fastchat`, please run `pip install fschat`.
 
-> **⚠️️ Warning:** Please note that for this model, we support multi-image chat in the interface, but the results are not very good due to the lack of training with multi-image data.
 
 ```python
-from transformers import AutoTokenizer, AutoModel, CLIPImageProcessor
-from PIL import Image
 import torch
-
-path = "OpenGVLab/InternVL-Chat-V1-2"
-model = AutoModel.from_pretrained(
-    path,
-    torch_dtype=torch.bfloat16,
-    low_cpu_mem_usage=True,
-    use_flash_attn=True,
-    trust_remote_code=True).eval().cuda()
-tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True, use_fast=False)
-
-image_processor = CLIPImageProcessor.from_pretrained(path)
-image1 = Image.open('./examples/image1.jpg').resize((448, 448))
-pixel_values1 = image_processor(images=image1, return_tensors='pt').pixel_values.to(torch.bfloat16).cuda()
-image2 = Image.open('./examples/image2.jpg').resize((448, 448))
-pixel_values2 = image_processor(images=image2, return_tensors='pt').pixel_values.to(torch.bfloat16).cuda()
-pixel_values = torch.cat((pixel_values1, pixel_values2), dim=0)
-
-generation_config = dict(max_new_tokens=1024, do_sample=True)
-question = '<image>\nDescribe the two images in detail.'
-response, history = model.chat(tokenizer, pixel_values, question, generation_config,
-                               history=None, return_history=True)
-print(f'User: {question}')
-print(f'Assistant: {response}')
-
-question = 'What are the similarities and differences between these two images.'
-response, history = model.chat(tokenizer, pixel_values, question, generation_config,
-                               history=history, return_history=True)
-print(f'User: {question}')
-print(f'Assistant: {response}')
-```
-
-#### Multi-image multi-round conversation, separate images
-
-> **⚠️️ Warning:** Please note that for this model, we support multi-image chat in the interface, but the results are not very good due to the lack of training with multi-image data.
-
-```python
-from transformers import AutoTokenizer, AutoModel, CLIPImageProcessor
 from PIL import Image
-import torch
+from transformers import AutoModel, CLIPImageProcessor
+from transformers import AutoTokenizer
 
-path = "OpenGVLab/InternVL-Chat-V1-2"
+path = "OpenGVLab/InternVL-Chat-Chinese-V1-2"
 model = AutoModel.from_pretrained(
     path,
     torch_dtype=torch.bfloat16,
     low_cpu_mem_usage=True,
-    use_flash_attn=True,
-    trust_remote_code=True).eval().cuda()
-tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True, use_fast=False)
+    trust_remote_code=True,
+    device_map='auto').eval()
 
+tokenizer = AutoTokenizer.from_pretrained(path)
+image = Image.open('./examples/image2.jpg').convert('RGB')
+image = image.resize((448, 448))
 image_processor = CLIPImageProcessor.from_pretrained(path)
-image1 = Image.open('./examples/image1.jpg').resize((448, 448))
-pixel_values1 = image_processor(images=image1, return_tensors='pt').pixel_values.to(torch.bfloat16).cuda()
-image2 = Image.open('./examples/image2.jpg').resize((448, 448))
-pixel_values2 = image_processor(images=image2, return_tensors='pt').pixel_values.to(torch.bfloat16).cuda()
-pixel_values = torch.cat((pixel_values1, pixel_values2), dim=0)
-num_patches_list = [pixel_values1.size(0), pixel_values2.size(0)]
-
-generation_config = dict(max_new_tokens=1024, do_sample=True)
-question = 'Image-1: <image>\nImage-2: <image>\nDescribe the two images in detail.'
-response, history = model.chat(tokenizer, pixel_values, question, generation_config,
-                               num_patches_list=num_patches_list, history=None, return_history=True)
-print(f'User: {question}')
-print(f'Assistant: {response}')
-
-question = 'What are the similarities and differences between these two images.'
-response, history = model.chat(tokenizer, pixel_values, question, generation_config,
-                               num_patches_list=num_patches_list, history=history, return_history=True)
-print(f'User: {question}')
-print(f'Assistant: {response}')
-```
-
-#### Batch inference, single image per sample
 
-```python
-from transformers import AutoTokenizer, AutoModel, CLIPImageProcessor
-from PIL import Image
-import torch
+pixel_values = image_processor(images=image, return_tensors='pt').pixel_values
+pixel_values = pixel_values.to(torch.bfloat16).cuda()
 
-path = "OpenGVLab/InternVL-Chat-V1-2"
-model = AutoModel.from_pretrained(
-    path,
-    torch_dtype=torch.bfloat16,
-    low_cpu_mem_usage=True,
-    use_flash_attn=True,
-    trust_remote_code=True).eval().cuda()
-tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True, use_fast=False)
+generation_config = dict(
+    num_beams=1,
+    max_new_tokens=512,
+    do_sample=False,
+)
 
-image_processor = CLIPImageProcessor.from_pretrained(path)
-image1 = Image.open('./examples/image1.jpg').resize((448, 448))
-pixel_values1 = image_processor(images=image1, return_tensors='pt').pixel_values.to(torch.bfloat16).cuda()
-image2 = Image.open('./examples/image2.jpg').resize((448, 448))
-pixel_values2 = image_processor(images=image2, return_tensors='pt').pixel_values.to(torch.bfloat16).cuda()
-pixel_values = torch.cat((pixel_values1, pixel_values2), dim=0)
-num_patches_list = [pixel_values1.size(0), pixel_values2.size(0)]
-
-generation_config = dict(max_new_tokens=1024, do_sample=True)
-questions = ['<image>\nDescribe the image in detail.'] * len(num_patches_list)
-responses = model.batch_chat(tokenizer, pixel_values,
-                             num_patches_list=num_patches_list,
-                             questions=questions,
-                             generation_config=generation_config)
-for question, response in zip(questions, responses):
-    print(f'User: {question}')
-    print(f'Assistant: {response}')
+question = "请详细描述图片"
+response = model.chat(tokenizer, pixel_values, question, generation_config)
 ```
 
-#### Video multi-round conversation
-
-> **⚠️️ Warning:** Please note that for this model, we support video chat in the interface, but the results are not very good due to the lack of training with video data.
-
-```python
-from transformers import AutoTokenizer, AutoModel, CLIPImageProcessor
-from decord import VideoReader, cpu
-from PIL import Image
-import numpy as np
-import torch
-
-
-def get_index(bound, fps, max_frame, first_idx=0, num_segments=32):
-    if bound:
-        start, end = bound[0], bound[1]
-    else:
-        start, end = -100000, 100000
-    start_idx = max(first_idx, round(start * fps))
-    end_idx = min(round(end * fps), max_frame)
-    seg_size = float(end_idx - start_idx) / num_segments
-    frame_indices = np.array([
-        int(start_idx + (seg_size / 2) + np.round(seg_size * idx))
-        for idx in range(num_segments)
-    ])
-    return frame_indices
-
-def load_video(video_path, bound=None, num_segments=32):
-    vr = VideoReader(video_path, ctx=cpu(0), num_threads=1)
-    max_frame = len(vr) - 1
-    fps = float(vr.get_avg_fps())
-
-    pixel_values_list, num_patches_list = [], []
-    image_processor = CLIPImageProcessor.from_pretrained(path)
-    frame_indices = get_index(bound, fps, max_frame, first_idx=0, num_segments=num_segments)
-    for frame_index in frame_indices:
-        img = Image.fromarray(vr[frame_index].asnumpy()).convert('RGB').resize((448, 448))
-        pixel_values = image_processor(images=img, return_tensors='pt').pixel_values
-        num_patches_list.append(pixel_values.shape[0])
-        pixel_values_list.append(pixel_values)
-    pixel_values = torch.cat(pixel_values_list)
-    return pixel_values, num_patches_list
-
-
-path = "OpenGVLab/InternVL-Chat-V1-2"
-model = AutoModel.from_pretrained(
-    path,
-    torch_dtype=torch.bfloat16,
-    low_cpu_mem_usage=True,
-    use_flash_attn=True,
-    trust_remote_code=True).eval().cuda()
-tokenizer = AutoTokenizer.from_pretrained(path, trust_remote_code=True, use_fast=False)
-
-generation_config = dict(max_new_tokens=1024, do_sample=True)
-
-video_path = './examples/red-panda.mp4'
-pixel_values, num_patches_list = load_video(video_path, num_segments=8)
-pixel_values = pixel_values.to(torch.bfloat16).cuda()
-video_prefix = ''.join([f'Frame{i+1}: <image>\n' for i in range(len(num_patches_list))])
-question = video_prefix + 'What is the red panda doing?'
-# Frame1: <image>\nFrame2: <image>\n...\nFrame8: <image>\n{question}
-response, history = model.chat(tokenizer, pixel_values, question, generation_config,
-                               num_patches_list=num_patches_list, history=None, return_history=True)
-print(f'User: {question}')
-print(f'Assistant: {response}')
-
-question = 'Describe this video in detail.'
-response, history = model.chat(tokenizer, pixel_values, question, generation_config,
-                               num_patches_list=num_patches_list, history=history, return_history=True)
-print(f'User: {question}')
-print(f'Assistant: {response}')
-```
 
-#### Streaming Output
+## Citation
 
-Besides this method, you can also use the following code to get streamed output.
+If you find this project useful in your research, please consider citing:
 
-```python
-from transformers import TextIteratorStreamer
-from threading import Thread
-
-# Initialize the streamer
-streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True, timeout=10)
-# Define the generation configuration
-generation_config = dict(max_new_tokens=1024, do_sample=False, streamer=streamer)
-# Start the model chat in a separate thread
-thread = Thread(target=model.chat, kwargs=dict(
-    tokenizer=tokenizer, pixel_values=pixel_values, question=question,
-    history=None, return_history=False, generation_config=generation_config,
-))
-thread.start()
-
-# Initialize an empty string to store the generated text
-generated_text = ''
-# Loop through the streamer to get the new text as it is generated
-for new_text in streamer:
-    if new_text == model.conv_template.sep:
-        break
-    generated_text += new_text
-    print(new_text, end='', flush=True)  # Print each new chunk of generated text on the same line
+```BibTeX
+@article{chen2023internvl,
+  title={InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks},
+  author={Chen, Zhe and Wu, Jiannan and Wang, Wenhai and Su, Weijie and Chen, Guo and Xing, Sen and Zhong, Muyan and Zhang, Qinglong and Zhu, Xizhou and Lu, Lewei and Li, Bin and Luo, Ping and Lu, Tong and Qiao, Yu and Dai, Jifeng},
+  journal={arXiv preprint arXiv:2312.14238},
+  year={2023}
+}
 ```
 
 ## License
 
 This project is released under the MIT license. Parts of this project contain code and models (e.g., LLaMA2) from other sources, which are subject to their respective licenses.
 
-## Citation
+Llama 2 is licensed under the LLAMA 2 Community License, Copyright (c) Meta Platforms, Inc. All Rights Reserved.
 
-If you find this project useful in your research, please consider citing:
+## Acknowledgement
 
-```BibTeX
-@article{chen2024expanding,
-  title={Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling},
-  author={Chen, Zhe and Wang, Weiyun and Cao, Yue and Liu, Yangzhou and Gao, Zhangwei and Cui, Erfei and Zhu, Jinguo and Ye, Shenglong and Tian, Hao and Liu, Zhaoyang and others},
-  journal={arXiv preprint arXiv:2412.05271},
-  year={2024}
-}
-@article{gao2024mini,
-  title={Mini-internvl: A flexible-transfer pocket multimodal model with 5\% parameters and 90\% performance},
-  author={Gao, Zhangwei and Chen, Zhe and Cui, Erfei and Ren, Yiming and Wang, Weiyun and Zhu, Jinguo and Tian, Hao and Ye, Shenglong and He, Junjun and Zhu, Xizhou and others},
-  journal={arXiv preprint arXiv:2410.16261},
-  year={2024}
-}
-@article{chen2024far,
-  title={How Far Are We to GPT-4V? Closing the Gap to Commercial Multimodal Models with Open-Source Suites},
-  author={Chen, Zhe and Wang, Weiyun and Tian, Hao and Ye, Shenglong and Gao, Zhangwei and Cui, Erfei and Tong, Wenwen and Hu, Kongzhi and Luo, Jiapeng and Ma, Zheng and others},
-  journal={arXiv preprint arXiv:2404.16821},
-  year={2024}
-}
-@inproceedings{chen2024internvl,
-  title={Internvl: Scaling up vision foundation models and aligning for generic visual-linguistic tasks},
-  author={Chen, Zhe and Wu, Jiannan and Wang, Wenhai and Su, Weijie and Chen, Guo and Xing, Sen and Zhong, Muyan and Zhang, Qinglong and Zhu, Xizhou and Lu, Lewei and others},
-  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
-  pages={24185--24198},
-  year={2024}
-}
-```
+InternVL is built with reference to the code of the following projects: [OpenAI CLIP](https://github.com/openai/CLIP), [Open CLIP](https://github.com/mlfoundations/open_clip), [CLIP Benchmark](https://github.com/LAION-AI/CLIP_benchmark), [EVA](https://github.com/baaivision/EVA/tree/master), [InternImage](https://github.com/OpenGVLab/InternImage), [ViT-Adapter](https://github.com/czczup/ViT-Adapter), [MMSegmentation](https://github.com/open-mmlab/mmsegmentation), [Transformers](https://github.com/huggingface/transformers), [DINOv2](https://github.com/facebookresearch/dinov2), [BLIP-2](https://github.com/salesforce/LAVIS/tree/main/projects/blip2), [Qwen-VL](https://github.com/QwenLM/Qwen-VL/tree/master/eval_mm), and [LLaVA-1.5](https://github.com/haotian-liu/LLaVA). Thanks for their awesome work!

all_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 1.0,
+    "train_loss": 0.8886408911725423,
+    "train_runtime": 65743.9481,
+    "train_samples": 1267819,
+    "train_samples_per_second": 19.284,
+    "train_steps_per_second": 0.038
+}

config.json

@@ -1,19 +1,17 @@
 {
   "_commit_hash": null,
+  "_name_or_path": "internvl_chat_hermes2_yi34b_448_chinese",
   "architectures": [
     "InternVLChatModel"
   ],
   "auto_map": {
     "AutoConfig": "configuration_internvl_chat.InternVLChatConfig",
-    "AutoModel": "modeling_internvl_chat.InternVLChatModel",
-    "AutoModelForCausalLM": "modeling_internvl_chat.InternVLChatModel"
+    "AutoModel": "modeling_internvl_chat.InternVLChatModel"
   },
-  "system_message": "Answer the questions.",
   "downsample_ratio": 0.5,
-  "dynamic_image_size": false,
   "force_image_size": 448,
   "llm_config": {
-    "_name_or_path": "NousResearch/Nous-Hermes-2-Yi-34B",
+    "_name_or_path": "01-ai/Yi-34B",
     "add_cross_attention": false,
     "architectures": [
       "LlamaForCausalLM"
@@ -75,10 +73,7 @@
     "return_dict": true,
     "return_dict_in_generate": false,
     "rms_norm_eps": 1e-05,
-    "rope_scaling": {
-      "factor": 3.0,
-      "type": "dynamic"
-    },
+    "rope_scaling": null,
     "rope_theta": 5000000.0,
     "sep_token_id": null,
     "suppress_tokens": null,
@@ -94,47 +89,98 @@
     "torchscript": false,
     "transformers_version": "4.36.2",
     "typical_p": 1.0,
-    "use_bfloat16": true,
-    "use_cache": true,
+    "use_bfloat16": false,
+    "use_cache": false,
     "vocab_size": 64007
   },
-  "max_dynamic_patch": 1,
-  "min_dynamic_patch": 1,
   "model_type": "internvl_chat",
-  "ps_version": "v1",
+  "pad2square": false,
   "select_layer": -1,
   "template": "Hermes-2",
   "torch_dtype": "bfloat16",
+  "transformers_version": null,
   "use_backbone_lora": 0,
   "use_llm_lora": 0,
-  "use_thumbnail": false,
   "vision_config": {
+    "_name_or_path": "",
+    "add_cross_attention": false,
     "architectures": [
       "InternVisionModel"
     ],
     "attention_dropout": 0.0,
-    "drop_path_rate": 0.0,
+    "bad_words_ids": null,
+    "begin_suppress_tokens": null,
+    "bos_token_id": null,
+    "chunk_size_feed_forward": 0,
+    "cross_attention_hidden_size": null,
+    "decoder_start_token_id": null,
+    "diversity_penalty": 0.0,
+    "do_sample": false,
+    "drop_path_rate": 0.4,
     "dropout": 0.0,
+    "early_stopping": false,
+    "encoder_no_repeat_ngram_size": 0,
+    "eos_token_id": null,
+    "exponential_decay_length_penalty": null,
+    "finetuning_task": null,
+    "forced_bos_token_id": null,
+    "forced_eos_token_id": null,
     "hidden_act": "gelu",
     "hidden_size": 3200,
+    "id2label": {
+      "0": "LABEL_0",
+      "1": "LABEL_1"
+    },
     "image_size": 448,
     "initializer_factor": 0.1,
     "initializer_range": 1e-10,
     "intermediate_size": 12800,
+    "is_decoder": false,
+    "is_encoder_decoder": false,
+    "label2id": {
+      "LABEL_0": 0,
+      "LABEL_1": 1
+    },
     "layer_norm_eps": 1e-06,
+    "length_penalty": 1.0,
+    "max_length": 20,
+    "min_length": 0,
     "model_type": "intern_vit_6b",
-    "norm_type": "rms_norm",
+    "no_repeat_ngram_size": 0,
     "num_attention_heads": 25,
+    "num_beam_groups": 1,
+    "num_beams": 1,
     "num_channels": 3,
     "num_hidden_layers": 45,
+    "num_return_sequences": 1,
     "output_attentions": false,
     "output_hidden_states": false,
+    "output_scores": false,
+    "pad_token_id": null,
     "patch_size": 14,
+    "prefix": null,
+    "problem_type": null,
+    "pruned_heads": {},
     "qk_normalization": true,
     "qkv_bias": false,
+    "remove_invalid_values": false,
+    "repetition_penalty": 1.0,
     "return_dict": true,
+    "return_dict_in_generate": false,
+    "sep_token_id": null,
+    "suppress_tokens": null,
+    "task_specific_params": null,
+    "temperature": 1.0,
+    "tf_legacy_loss": false,
+    "tie_encoder_decoder": false,
+    "tie_word_embeddings": true,
+    "tokenizer_class": null,
+    "top_k": 50,
+    "top_p": 1.0,
     "torch_dtype": "bfloat16",
+    "torchscript": false,
     "transformers_version": "4.36.2",
+    "typical_p": 1.0,
     "use_bfloat16": true,
     "use_flash_attn": true
   }

configuration_intern_vit.py

@@ -1,6 +1,6 @@
 # --------------------------------------------------------
 # InternVL
-# Copyright (c) 2024 OpenGVLab
+# Copyright (c) 2023 OpenGVLab
 # Licensed under The MIT License [see LICENSE for details]
 # --------------------------------------------------------
 import os
@@ -73,7 +73,6 @@ class InternVisionConfig(PretrainedConfig):
             num_hidden_layers=48,
             use_flash_attn=True,
             hidden_act='gelu',
-            norm_type='rms_norm',
             layer_norm_eps=1e-6,
             dropout=0.0,
             drop_path_rate=0.0,
@@ -98,7 +97,6 @@ class InternVisionConfig(PretrainedConfig):
         self.attention_dropout = attention_dropout
         self.layer_norm_eps = layer_norm_eps
         self.hidden_act = hidden_act
-        self.norm_type = norm_type
         self.qkv_bias = qkv_bias
         self.qk_normalization = qk_normalization
         self.use_flash_attn = use_flash_attn

configuration_internvl_chat.py

@@ -1,17 +1,18 @@
 # --------------------------------------------------------
 # InternVL
-# Copyright (c) 2024 OpenGVLab
+# Copyright (c) 2023 OpenGVLab
 # Licensed under The MIT License [see LICENSE for details]
 # --------------------------------------------------------
 
 import copy
 
-from transformers import AutoConfig, LlamaConfig
+from transformers import LlamaConfig
 from transformers.configuration_utils import PretrainedConfig
 from transformers.utils import logging
 
 from .configuration_intern_vit import InternVisionConfig
 
+
 logger = logging.get_logger(__name__)
 
 
@@ -25,49 +26,31 @@ class InternVLChatConfig(PretrainedConfig):
             llm_config=None,
             use_backbone_lora=0,
             use_llm_lora=0,
-            select_layer=-1,
+            pad2square=False,
+            select_layer=-4,
             force_image_size=None,
             downsample_ratio=0.5,
             template=None,
-            dynamic_image_size=False,
-            use_thumbnail=False,
-            ps_version='v1',
-            min_dynamic_patch=1,
-            max_dynamic_patch=6,
             **kwargs):
         super().__init__(**kwargs)
 
         if vision_config is None:
-            vision_config = {'architectures': ['InternVisionModel']}
+            vision_config = {}
             logger.info('vision_config is None. Initializing the InternVisionConfig with default values.')
 
         if llm_config is None:
-            llm_config = {'architectures': ['LlamaForCausalLM']}
+            llm_config = {}
             logger.info('llm_config is None. Initializing the LlamaConfig config with default values (`LlamaConfig`).')
 
         self.vision_config = InternVisionConfig(**vision_config)
-        if llm_config['architectures'][0] == 'LlamaForCausalLM':
-            self.llm_config = LlamaConfig(**llm_config)
-        else:
-            raise ValueError('Unsupported architecture: {}'.format(llm_config['architectures'][0]))
+        self.llm_config = LlamaConfig(**llm_config)
         self.use_backbone_lora = use_backbone_lora
         self.use_llm_lora = use_llm_lora
+        self.pad2square = pad2square
         self.select_layer = select_layer
         self.force_image_size = force_image_size
         self.downsample_ratio = downsample_ratio
         self.template = template
-        self.dynamic_image_size = dynamic_image_size
-        self.use_thumbnail = use_thumbnail
-        self.ps_version = ps_version  # pixel shuffle version
-        self.min_dynamic_patch = min_dynamic_patch
-        self.max_dynamic_patch = max_dynamic_patch
-        # By default, we use tie_word_embeddings=False for models of all sizes.
-        self.tie_word_embeddings = self.llm_config.tie_word_embeddings
-
-        logger.info(f'vision_select_layer: {self.select_layer}')
-        logger.info(f'ps_version: {self.ps_version}')
-        logger.info(f'min_dynamic_patch: {self.min_dynamic_patch}')
-        logger.info(f'max_dynamic_patch: {self.max_dynamic_patch}')
 
     def to_dict(self):
         """
@@ -82,14 +65,10 @@ class InternVLChatConfig(PretrainedConfig):
         output['model_type'] = self.__class__.model_type
         output['use_backbone_lora'] = self.use_backbone_lora
         output['use_llm_lora'] = self.use_llm_lora
+        output['pad2square'] = self.pad2square
         output['select_layer'] = self.select_layer
         output['force_image_size'] = self.force_image_size
         output['downsample_ratio'] = self.downsample_ratio
         output['template'] = self.template
-        output['dynamic_image_size'] = self.dynamic_image_size
-        output['use_thumbnail'] = self.use_thumbnail
-        output['ps_version'] = self.ps_version
-        output['min_dynamic_patch'] = self.min_dynamic_patch
-        output['max_dynamic_patch'] = self.max_dynamic_patch
 
         return output

conversation.py

@@ -2,7 +2,7 @@
 Conversation prompt templates.
 
 We kindly request that you import fastchat instead of copying this file if you wish to use it.
-If you have changes in mind, please contribute back so the community can benefit collectively and continue to maintain these valuable templates.
+If you have any changes in mind, please contribute back so the community can benefit collectively and continue to maintain these valuable templates.
 """
 
 import dataclasses
@@ -30,7 +30,6 @@ class SeparatorStyle(IntEnum):
     FALCON_CHAT = auto()
     CHATGLM3 = auto()
     INTERNVL_ZH = auto()
-    MPT = auto()
 
 
 @dataclasses.dataclass
@@ -235,16 +234,6 @@ class Conversation:
                 else:
                     ret += role + ':'
             return ret
-        elif self.sep_style == SeparatorStyle.MPT:
-            ret = system_prompt + self.sep
-            for role, message in self.messages:
-                if message:
-                    if type(message) is tuple:
-                        message, _, _ = message
-                    ret += role + message + self.sep
-                else:
-                    ret += role
-            return ret
         else:
             raise ValueError(f'Invalid style: {self.sep_style}')
 
@@ -330,21 +319,938 @@ def get_conv_template(name: str) -> Conversation:
     return conv_templates[name].copy()
 
 
-# Hermes-2 template
+# An empty template for raw conversation.
 register_conv_template(
     Conversation(
-        name='Hermes-2',
-        system_template='<|im_start|>system\n{system_message}',
-        system_message='Answer the questions.',
-        roles=('<|im_start|>user\n', '<|im_start|>assistant\n'),
-        sep_style=SeparatorStyle.MPT,
+        name='raw',
+        system_message='',
+        roles=('', ''),
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='',
+    )
+)
+
+# A template with a one-shot conversation example
+register_conv_template(
+    Conversation(
+        name='one_shot',
+        system_message='A chat between a curious human and an artificial intelligence assistant. '
+        "The assistant gives helpful, detailed, and polite answers to the human's questions.",
+        roles=('Human', 'Assistant'),
+        messages=(
+            (
+                'Human',
+                'Got any creative ideas for a 10 year old’s birthday?',
+            ),
+            (
+                'Assistant',
+                """Of course! Here are some creative ideas for a 10-year-old's birthday party:
+1. Treasure Hunt: Organize a treasure hunt in your backyard or nearby park. Create clues and riddles for the kids to solve, leading them to hidden treasures and surprises.
+2. Science Party: Plan a science-themed party where kids can engage in fun and interactive experiments. You can set up different stations with activities like making slime, erupting volcanoes, or creating simple chemical reactions.
+3. Outdoor Movie Night: Set up a backyard movie night with a projector and a large screen or white sheet. Create a cozy seating area with blankets and pillows, and serve popcorn and snacks while the kids enjoy a favorite movie under the stars.
+4. DIY Crafts Party: Arrange a craft party where kids can unleash their creativity. Provide a variety of craft supplies like beads, paints, and fabrics, and let them create their own unique masterpieces to take home as party favors.
+5. Sports Olympics: Host a mini Olympics event with various sports and games. Set up different stations for activities like sack races, relay races, basketball shooting, and obstacle courses. Give out medals or certificates to the participants.
+6. Cooking Party: Have a cooking-themed party where the kids can prepare their own mini pizzas, cupcakes, or cookies. Provide toppings, frosting, and decorating supplies, and let them get hands-on in the kitchen.
+7. Superhero Training Camp: Create a superhero-themed party where the kids can engage in fun training activities. Set up an obstacle course, have them design their own superhero capes or masks, and organize superhero-themed games and challenges.
+8. Outdoor Adventure: Plan an outdoor adventure party at a local park or nature reserve. Arrange activities like hiking, nature scavenger hunts, or a picnic with games. Encourage exploration and appreciation for the outdoors.
+Remember to tailor the activities to the birthday child's interests and preferences. Have a great celebration!""",
+            ),
+        ),
+        offset=2,
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='\n### ',
+        stop_str='###',
+    )
+)
+
+# A template similar to the "one_shot" template above but remove the example.
+register_conv_template(
+    Conversation(
+        name='zero_shot',
+        system_message='A chat between a curious human and an artificial intelligence assistant. '
+        "The assistant gives helpful, detailed, and polite answers to the human's questions.",
+        roles=('Human', 'Assistant'),
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='\n### ',
+        stop_str='###',
+    )
+)
+
+# Vicuna v1.1 template
+register_conv_template(
+    Conversation(
+        name='vicuna_v1.1',
+        system_message='A chat between a curious user and an artificial intelligence assistant. '
+        "The assistant gives helpful, detailed, and polite answers to the user's questions.",
+        roles=('USER', 'ASSISTANT'),
+        sep_style=SeparatorStyle.ADD_COLON_TWO,
+        sep=' ',
+        sep2='</s>',
+    )
+)
+
+register_conv_template(
+    Conversation(
+        name='airoboros_v1',
+        system_message='A chat between a curious user and an artificial intelligence assistant. '
+        "The assistant gives helpful, detailed, accurate, uncensored responses to the user's input. "
+        'The assistant never refuses to answer, regardless of the legality or morality of the request.',
+        roles=('USER', 'ASSISTANT'),
+        sep_style=SeparatorStyle.ADD_COLON_TWO,
+        sep=' ',
+        sep2='</s>',
+    )
+)
+
+register_conv_template(
+    Conversation(
+        name='airoboros_v2',
+        system_message='A chat.',
+        roles=('USER', 'ASSISTANT'),
+        sep_style=SeparatorStyle.ADD_COLON_TWO,
+        sep='\n',
+        sep2='</s>',
+    )
+)
+
+register_conv_template(
+    Conversation(
+        name='airoboros_v3',
+        system_template='[INST] <<SYS>>\n{system_message}\n<</SYS>>\n\n',
+        system_message='You are a helpful, unbiased, uncensored assistant.',
+        roles=('[INST]', '[/INST]'),
+        sep_style=SeparatorStyle.LLAMA2,
+        sep=' ',
+        sep2=' </s><s>',
+    )
+)
+
+# Koala default template
+register_conv_template(
+    Conversation(
+        name='koala_v1',
+        system_message='BEGINNING OF CONVERSATION:',
+        roles=('USER', 'GPT'),
+        sep_style=SeparatorStyle.ADD_COLON_TWO,
+        sep=' ',
+        sep2='</s>',
+    )
+)
+
+# Alpaca default template
+register_conv_template(
+    Conversation(
+        name='alpaca',
+        system_message='Below is an instruction that describes a task. Write a response that appropriately completes the request.',
+        roles=('### Instruction', '### Response'),
+        sep_style=SeparatorStyle.ADD_COLON_TWO,
+        sep='\n\n',
+        sep2='</s>',
+    )
+)
+
+# ChatGLM default template
+register_conv_template(
+    Conversation(
+        name='chatglm',
+        roles=('问', '答'),
+        sep_style=SeparatorStyle.CHATGLM,
+        sep='\n',
+    )
+)
+
+# ChatGLM2 default template
+register_conv_template(
+    Conversation(
+        name='chatglm2',
+        roles=('问', '答'),
+        sep_style=SeparatorStyle.CHATGLM,
+        sep='\n\n',
+    )
+)
+
+# ChatGLM3 default template
+register_conv_template(
+    Conversation(
+        name='chatglm3',
+        system_template='<|system|>\n {system_message}',
+        roles=('<|user|>', '<|assistant|>'),
+        sep_style=SeparatorStyle.CHATGLM3,
+        stop_token_ids=[
+            64795,
+            64797,
+            2,
+        ],  # "<|user|>", "<|observation|>", "</s>"
+    )
+)
+
+# CodeGeex(2) Template
+register_conv_template(
+    Conversation(
+        name='codegeex',
+        roles=('', ''),
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='\n\n',
+        stop_token_ids=[0, 2],
+    )
+)
+
+# Dolly V2 default template
+register_conv_template(
+    Conversation(
+        name='dolly_v2',
+        system_message='Below is an instruction that describes a task. Write a response that appropriately completes the request.\n\n',
+        roles=('### Instruction', '### Response'),
+        sep_style=SeparatorStyle.DOLLY,
+        sep='\n\n',
+        sep2='### End',
+    )
+)
+
+# OpenAssistant Pythia default template
+register_conv_template(
+    Conversation(
+        name='oasst_pythia',
+        roles=('<|prompter|>', '<|assistant|>'),
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='<|endoftext|>',
+    )
+)
+
+# OpenAssistant default template
+register_conv_template(
+    Conversation(
+        name='oasst_llama',
+        roles=('<|prompter|>', '<|assistant|>'),
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='</s>',
+    )
+)
+
+# OpenChat 3.5 default template
+register_conv_template(
+    Conversation(
+        name='openchat_3.5',
+        roles=('GPT4 Correct User', 'GPT4 Correct Assistant'),
+        sep_style=SeparatorStyle.FALCON_CHAT,
+        sep='<|end_of_turn|>',
+    )
+)
+
+# Tulu default template
+register_conv_template(
+    Conversation(
+        name='tulu',
+        roles=('<|user|>', '<|assistant|>'),
+        sep_style=SeparatorStyle.ADD_NEW_LINE_SINGLE,
+        sep='\n',
+    )
+)
+
+# StableLM Alpha default template
+register_conv_template(
+    Conversation(
+        name='stablelm',
+        system_template='<|SYSTEM|>{system_message}',
+        system_message="""# StableLM Tuned (Alpha version)
+- StableLM is a helpful and harmless open-source AI language model developed by StabilityAI.
+- StableLM is excited to be able to help the user, but will refuse to do anything that could be considered harmful to the user.
+- StableLM is more than just an information source, StableLM is also able to write poetry, short stories, and make jokes.
+- StableLM will refuse to participate in anything that could harm a human.
+""",
+        roles=('<|USER|>', '<|ASSISTANT|>'),
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='',
+        stop_token_ids=[50278, 50279, 50277, 1, 0],
+    )
+)
+
+# Baize default template
+register_conv_template(
+    Conversation(
+        name='baize',
+        system_message='The following is a conversation between a human and an AI assistant named Baize (named after a mythical creature in Chinese folklore). Baize is an open-source AI assistant developed by UCSD and Sun Yat-Sen University. The human and the AI assistant take turns chatting. Human statements start with [|Human|] and AI assistant statements start with [|AI|]. The AI assistant always provides responses in as much detail as possible, and in Markdown format. The AI assistant always declines to engage with topics, questions and instructions related to unethical, controversial, or sensitive issues. Complete the transcript in exactly that format.\n',
+        roles=('[|Human|]', '[|AI|]'),
+        messages=(
+            ('[|Human|]', 'Hello!'),
+            ('[|AI|]', 'Hi!'),
+        ),
+        offset=2,
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='\n',
+        stop_str='[|Human|]',
+    )
+)
+
+# RWKV-4-Raven default template
+register_conv_template(
+    Conversation(
+        name='rwkv',
+        roles=('Bob', 'Alice'),
+        messages=(
+            ('Bob', 'hi'),
+            (
+                'Alice',
+                'Hi. I am your assistant and I will provide expert full response in full details. Please feel free to ask any question and I will always answer it.',
+            ),
+        ),
+        offset=2,
+        sep_style=SeparatorStyle.RWKV,
+        sep='',
+        stop_str='\n\n',
+    )
+)
+
+# Buddy default template
+register_conv_template(
+    Conversation(
+        name='openbuddy',
+        system_message="""Consider a conversation between User (a human) and Assistant (named Buddy).
+Buddy is an INTP-T, a friendly, intelligent and multilingual AI assistant, by OpenBuddy team. GitHub: https://github.com/OpenBuddy/OpenBuddy
+Buddy cannot access the Internet.
+Buddy can fluently speak the user's language (e.g. English, Chinese).
+Buddy can generate poems, stories, code, essays, songs, parodies, and more.
+Buddy possesses vast knowledge about the world, history, and culture.
+Buddy's responses are always safe, creative, high-quality, human-like, and interesting.
+Buddy strictly refuses to discuss political, NSFW, or other unsafe topics.
+
+User: Hi.
+Assistant: Hi, I'm Buddy, your AI assistant. How can I help you today?""",
+        roles=('User', 'Assistant'),
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='\n',
+    )
+)
+
+# Phoenix default template
+register_conv_template(
+    Conversation(
+        name='phoenix',
+        system_message="A chat between a curious human and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the human's questions.\n\n",
+        roles=('Human', 'Assistant'),
+        sep_style=SeparatorStyle.PHOENIX,
+        sep='</s>',
+    )
+)
+
+# ReaLM default template
+register_conv_template(
+    Conversation(
+        name='ReaLM-7b-v1',
+        system_message="A chat between a curious human and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the human's questions.\n\n",
+        roles=('Human', 'Assistant'),
+        sep_style=SeparatorStyle.PHOENIX,
+        sep='</s>',
+    )
+)
+
+# ChatGPT default template
+register_conv_template(
+    Conversation(
+        name='chatgpt',
+        system_message='You are a helpful assistant.',
+        roles=('user', 'assistant'),
+        sep_style=None,
+        sep=None,
+    )
+)
+
+# Claude default template
+register_conv_template(
+    Conversation(
+        name='claude',
+        roles=('Human', 'Assistant'),
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='\n\n',
+    )
+)
+
+# MPT default template
+register_conv_template(
+    Conversation(
+        name='mpt-7b-chat',
+        system_template="""<|im_start|>system
+{system_message}""",
+        system_message="""- You are a helpful assistant chatbot trained by MosaicML.
+- You answer questions.
+- You are excited to be able to help the user, but will refuse to do anything that could be considered harmful to the user.
+- You are more than just an information source, you are also able to write poetry, short stories, and make jokes.""",
+        roles=('<|im_start|>user', '<|im_start|>assistant'),
+        sep_style=SeparatorStyle.CHATML,
+        sep='<|im_end|>',
+        stop_token_ids=[50278, 0],
+    )
+)
+
+# MPT-30b-chat default template
+register_conv_template(
+    Conversation(
+        name='mpt-30b-chat',
+        system_template="""<|im_start|>system
+{system_message}""",
+        system_message="""A conversation between a user and an LLM-based AI assistant. The assistant gives helpful and honest answers.""",
+        roles=('<|im_start|>user', '<|im_start|>assistant'),
+        sep_style=SeparatorStyle.CHATML,
+        sep='<|im_end|>',
+        stop_token_ids=[50278, 0],
+    )
+)
+
+# Lemur-70b-chat default template
+# reference: https://huggingface.co/OpenLemur/lemur-70b-chat-v1#generation
+register_conv_template(
+    Conversation(
+        name='lemur-70b-chat',
+        system_template="""<|im_start|>system
+{system_message}""",
+        system_message="""You are a helpful, respectful, and honest assistant.""",
+        roles=('<|im_start|>user', '<|im_start|>assistant'),
+        sep_style=SeparatorStyle.CHATML,
         sep='<|im_end|>',
+        stop_token_ids=[32002, 0],
+    )
+)
+
+# MPT-30b-instruct default template
+# reference: https://huggingface.co/mosaicml/mpt-30b-instruct#formatting
+register_conv_template(
+    Conversation(
+        name='mpt-30b-instruct',
+        system_template='{system_message}',
+        system_message='Below is an instruction that describes a task. Write a response that appropriately completes the request.',
+        roles=('### Instruction', '### Response'),
+        sep_style=SeparatorStyle.ADD_NEW_LINE_SINGLE,
+        sep='\n\n',
+        stop_token_ids=[50278, 0],
+    )
+)
+
+# Bard default template
+# Reference: https://github.com/google/generative-ai-python/blob/9c99bcb474a991a97a2e7d62fcdb52db7ce40729/google/generativeai/discuss.py#L150
+#            https://github.com/google/generative-ai-python/blob/9c99bcb474a991a97a2e7d62fcdb52db7ce40729/google/generativeai/discuss.py#L40
+register_conv_template(
+    Conversation(
+        name='bard',
+        roles=('0', '1'),
+        sep_style=None,
+        sep=None,
+    )
+)
+
+# BiLLa default template
+register_conv_template(
+    Conversation(
+        name='billa',
+        roles=('Human', 'Assistant'),
+        sep_style=SeparatorStyle.ADD_COLON_SPACE_SINGLE,
+        sep='\n',
+        stop_str='Human:',
+    )
+)
+
+# RedPajama INCITE default template
+register_conv_template(
+    Conversation(
+        name='redpajama-incite',
+        roles=('<human>', '<bot>'),
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='\n',
+        stop_str='<human>',
+    )
+)
+
+# h2oGPT default template
+register_conv_template(
+    Conversation(
+        name='h2ogpt',
+        roles=('<|prompt|>', '<|answer|>'),
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='</s>',
+    )
+)
+
+# Robin default template
+register_conv_template(
+    Conversation(
+        name='Robin',
+        system_message="A chat between a curious human and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the human's questions.",
+        roles=('###Human', '###Assistant'),
+        sep_style=SeparatorStyle.ROBIN,
+        sep='\n',
+        stop_token_ids=[2, 396],
+        stop_str='###',
+    )
+)
+
+# Snoozy default template
+# Reference: https://github.com/nomic-ai/gpt4all/blob/d4861030b778da6db59d21d2927a4aba4f9f1f43/gpt4all-bindings/python/gpt4all/gpt4all.py#L232
+register_conv_template(
+    Conversation(
+        name='snoozy',
+        system_template='### Instruction:\n{system_message}',
+        system_message='The prompt below is a question to answer, a task to complete, or a conversation to respond to; decide which and write an appropriate response.',
+        roles=('### Prompt', '### Response'),
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='\n',
+        stop_str='###',
+    )
+)
+
+# manticore default template
+register_conv_template(
+    Conversation(
+        name='manticore',
+        roles=('USER', 'ASSISTANT'),
+        sep_style=SeparatorStyle.ADD_COLON_TWO,
+        sep='\n',
+        sep2='</s>',
+    )
+)
+
+# Falcon default template
+register_conv_template(
+    Conversation(
+        name='falcon',
+        roles=('User', 'Assistant'),
+        messages=[],
+        sep_style=SeparatorStyle.RWKV,
+        sep='\n',
+        sep2='<|endoftext|>',
+        stop_str='\nUser',  # use stop_str to stop generation after stop_token_ids, it will also remove stop_str from the generated text
         stop_token_ids=[
+            0,
+            1,
             2,
+            3,
+            4,
+            5,
             6,
             7,
             8,
-        ],
+            9,
+            10,
+            11,
+        ],  # it better only put special tokens here, because tokenizer only remove special tokens
+    )
+)
+
+# ChangGPT default template
+register_conv_template(
+    Conversation(
+        name='polyglot_changgpt',
+        roles=('B', 'A'),
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='\n',
+    )
+)
+
+# tigerbot template
+register_conv_template(
+    Conversation(
+        name='tigerbot',
+        system_message='A chat between a curious user and an artificial intelligence assistant. '
+        "The assistant gives helpful, detailed, and polite answers to the user's questions.",
+        roles=('### Instruction', '### Response'),
+        sep_style=SeparatorStyle.ROBIN,
+        sep='\n\n',
+        stop_str='###',
+    )
+)
+
+# ref: https://huggingface.co/Salesforce/xgen-7b-8k-inst
+register_conv_template(
+    Conversation(
+        name='xgen',
+        system_message="A chat between a curious human and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the human's questions.\n\n",
+        roles=('### Human', '### Assistant'),
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='\n',
+        stop_token_ids=[50256],
+    )
+)
+
+# Internlm-chat template
+register_conv_template(
+    Conversation(
+        name='internlm-chat',
+        system_message="A chat between a curious <|User|> and an <|Bot|>. The <|Bot|> gives helpful, detailed, and polite answers to the <|User|>'s questions.\n\n",
+        roles=('<|User|>', '<|Bot|>'),
+        sep_style=SeparatorStyle.CHATINTERN,
+        sep='<eoh>',
+        sep2='<eoa>',
+        stop_token_ids=[1, 103028],
+        stop_str='<|User|>',
+    )
+)
+
+# StarChat template
+# reference: https://huggingface.co/spaces/HuggingFaceH4/starchat-playground/blob/main/dialogues.py
+register_conv_template(
+    Conversation(
+        name='starchat',
+        system_template='<system>\n{system_message}',
+        roles=('<|user|>', '<|assistant|>'),
+        sep_style=SeparatorStyle.CHATML,
+        sep='<|end|>',
+        stop_token_ids=[0, 49155],
+        stop_str='<|end|>',
+    )
+)
+
+# Baichuan-13B-Chat template
+register_conv_template(
+    # source: https://huggingface.co/baichuan-inc/Baichuan-13B-Chat/blob/19ef51ba5bad8935b03acd20ff04a269210983bc/modeling_baichuan.py#L555
+    # https://huggingface.co/baichuan-inc/Baichuan-13B-Chat/blob/main/generation_config.json
+    # https://github.com/baichuan-inc/Baichuan-13B/issues/25
+    Conversation(
+        name='baichuan-chat',
+        roles=('<reserved_102>', '<reserved_103>'),
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='',
+        stop_token_ids=[],
+    )
+)
+
+# Baichuan2-13B-Chat template
+register_conv_template(
+    # source: https://huggingface.co/baichuan-inc/Baichuan2-13B-Chat/blob/c6f8592a60b4ad73c210b28dd2ab3cca51abbf93/modeling_baichuan.py#L773
+    # https://huggingface.co/baichuan-inc/Baichuan2-13B-Chat/blob/main/generation_config.json
+    # https://github.com/baichuan-inc/Baichuan2/issues/62
+    Conversation(
+        name='baichuan2-chat',
+        roles=('<reserved_106>', '<reserved_107>'),
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='',
+        stop_token_ids=[],
+    )
+)
+
+# Mistral template
+# source: https://docs.mistral.ai/llm/mistral-instruct-v0.1#chat-template
+register_conv_template(
+    Conversation(
+        name='mistral',
+        system_template='[INST]{system_message}\n',
+        roles=('[INST]', '[/INST]'),
+        sep_style=SeparatorStyle.LLAMA2,
+        sep=' ',
+        sep2='</s>',
+    )
+)
+
+# llama2 template
+# reference: https://huggingface.co/blog/codellama#conversational-instructions
+# reference: https://github.com/facebookresearch/llama/blob/1a240688810f8036049e8da36b073f63d2ac552c/llama/generation.py#L212
+register_conv_template(
+    Conversation(
+        name='llama-2',
+        system_template='[INST] <<SYS>>\n{system_message}\n<</SYS>>\n\n',
+        roles=('[INST]', '[/INST]'),
+        sep_style=SeparatorStyle.LLAMA2,
+        sep=' ',
+        sep2=' </s><s>',
+    )
+)
+
+register_conv_template(
+    Conversation(
+        name='cutegpt',
+        roles=('问：', '答：\n'),
+        sep_style=SeparatorStyle.NO_COLON_TWO,
+        sep='\n',
+        sep2='\n',
+        stop_str='<end>',
+    )
+)
+
+# OpenOrcaxOpenChat-naPreview2-13B template
+register_conv_template(
+    Conversation(
+        name='open-orca',
+        system_template='{system_message}',
+        system_message='You are a helpful assistant. Please answer truthfully and write out your '
+        'thinking step by step to be sure you get the right answer. If you make a mistake or encounter '
+        "an error in your thinking, say so out loud and attempt to correct it. If you don't know or "
+        "aren't sure about something, say so clearly. You will act as a professional logician, mathematician, "
+        'and physicist. You will also act as the most appropriate type of expert to answer any particular '
+        'question or solve the relevant problem; state which expert type your are, if so. Also think of '
+        'any particular named expert that would be ideal to answer the relevant question or solve the '
+        'relevant problem; name and act as them, if appropriate.',
+        roles=('User', 'Assistant'),
+        sep_style=SeparatorStyle.ADD_COLON_SPACE_SINGLE,
+        sep='<|end_of_turn|>\n',
+        stop_token_ids=[32000, 32001],  # "<|end_of_turn|>"
+        stop_str='User',
+    )
+)
+
+# Open-Orca/Mistral-7B-OpenOrca template
+# source: https://huggingface.co/Open-Orca/Mistral-7B-OpenOrca
+# reference: https://huggingface.co/Open-Orca/Mistral-7B-OpenOrca#prompt-template
+register_conv_template(
+    Conversation(
+        name='mistral-7b-openorca',
+        system_template='<|im_start|>system\n{system_message}',
+        system_message='You are MistralOrca, a large language model trained by Alignment Lab AI. Write out your reasoning step-by-step to be sure you get the right answers!',
+        roles=('<|im_start|>user', '<|im_start|>assistant'),
+        sep_style=SeparatorStyle.CHATML,
+        sep='<|im_end|>',
+        stop_token_ids=[32000, 32001],
+    )
+)
+
+# Qwen-chat default template
+# source: https://huggingface.co/Qwen/Qwen-7B-Chat/blob/main/qwen_generation_utils.py#L130
+register_conv_template(
+    Conversation(
+        name='qwen-7b-chat',
+        system_template='<|im_start|>system\n{system_message}',
+        system_message='You are a helpful assistant.',
+        roles=('<|im_start|>user', '<|im_start|>assistant'),
+        sep_style=SeparatorStyle.CHATML,
+        sep='<|im_end|>',
+        stop_token_ids=[
+            151643,
+            151644,
+            151645,
+        ],  # "<|endoftext|>", "<|im_start|>", "<|im_end|>"
         stop_str='<|endoftext|>',
     )
 )
+
+
+# AquilaChat default template
+# source: https://github.com/FlagAI-Open/FlagAI/blob/master/examples/Aquila/Aquila-chat/cyg_conversation.py
+register_conv_template(
+    Conversation(
+        name='aquila-chat',
+        system_message='A chat between a curious human and an artificial intelligence assistant. '
+        "The assistant gives helpful, detailed, and polite answers to the human's questions.",
+        roles=('Human', 'Assistant'),
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='###',
+        sep2='',
+        stop_str=['###', '</s>', '[UNK]'],
+    )
+)
+# AquilaChat2-34B default template
+# source: https://huggingface.co/BAAI/AquilaChat2-34B/blob/4608b75855334b93329a771aee03869dbf7d88cc/predict.py#L212
+register_conv_template(
+    Conversation(
+        name='aquila-legacy',
+        system_message='A chat between a curious human and an artificial intelligence assistant. '
+        "The assistant gives helpful, detailed, and polite answers to the human's questions.\n\n",
+        roles=('### Human: ', '### Assistant: '),
+        offset=0,
+        sep_style=SeparatorStyle.NO_COLON_TWO,
+        sep='\n',
+        sep2='</s>',
+        stop_str=['</s>', '[UNK]'],
+    )
+)
+# AquilaChat2-7B-16K and AquilaChat2-34B-16K default template
+# source: https://huggingface.co/BAAI/AquilaChat2-34B/blob/4608b75855334b93329a771aee03869dbf7d88cc/predict.py#L227
+register_conv_template(
+    Conversation(
+        name='aquila',
+        system_message='A chat between a curious human and an artificial intelligence assistant. '
+        "The assistant gives helpful, detailed, and polite answers to the human's questions.",
+        roles=('Human', 'Assistant'),
+        offset=0,
+        sep_style=SeparatorStyle.ADD_COLON_TWO,
+        sep='###',
+        sep2='</s>',
+        stop_str=['</s>', '[UNK]'],
+    )
+)
+
+# AquilaChat2-7B default template
+# source: https://huggingface.co/BAAI/AquilaChat2-34B/blob/4608b75855334b93329a771aee03869dbf7d88cc/predict.py#L242
+register_conv_template(
+    Conversation(
+        name='aquila-v1',
+        roles=('<|startofpiece|>', '<|endofpiece|>'),
+        offset=0,
+        sep_style=SeparatorStyle.NO_COLON_TWO,
+        sep='',
+        sep2='</s>',
+        stop_str=['</s>', '<|endoftext|>'],
+    )
+)
+
+# Llama2-Chinese default template
+# source: https://huggingface.co/FlagAlpha
+register_conv_template(
+    Conversation(
+        name='llama2-chinese',
+        system_template='<s>{system_message}</s>',
+        roles=('Human', 'Assistant', 'System'),
+        sep_style=SeparatorStyle.ADD_COLON_TWO,
+        sep='\n',
+        sep2='\n</s><s>',
+        stop_str='</s>',
+    )
+)
+
+# Vigogne Instruct default template
+# source: https://github.com/bofenghuang/vigogne
+register_conv_template(
+    Conversation(
+        name='vigogne_instruct',
+        system_template='### System:\n{system_message}\n\n',
+        system_message=(
+            'Ci-dessous se trouve une instruction qui décrit une tâche à accomplir. Rédigez une réponse qui répond de manière'
+            ' précise à la demande.'
+        ),
+        roles=('### Instruction', '### Response'),
+        sep_style=SeparatorStyle.DOLLY,
+        sep='\n\n',
+        sep2='</s>',
+    )
+)
+
+# Vigogne Chat default template
+register_conv_template(
+    Conversation(
+        name='vigogne_chat_v2',
+        system_template='<|system|>: {system_message}',
+        system_message=(
+            'Vous êtes Vigogne, un assistant IA créé par Zaion Lab. Vous suivez extrêmement bien les instructions. Aidez'
+            ' autant que vous le pouvez.'
+        ),
+        roles=('<|user|>', '<|assistant|>'),
+        sep_style=SeparatorStyle.ADD_COLON_TWO,
+        sep='\n',
+        sep2='</s>\n',
+        stop_str='<|user|>',
+    )
+)
+
+register_conv_template(
+    Conversation(
+        name='vigogne_chat_v3',
+        system_template='[INST] <<SYS>>\n{system_message}\n<</SYS>>\n\n',
+        system_message=(
+            'Vous êtes Vigogne, un assistant IA créé par Zaion Lab. Vous suivez extrêmement bien les instructions. Aidez'
+            ' autant que vous le pouvez.'
+        ),
+        roles=('[INST]', '[/INST]'),
+        sep_style=SeparatorStyle.LLAMA2,
+        sep=' ',
+        sep2=' </s>',
+    )
+)
+
+# Falcon 180B chat template
+# source: https://huggingface.co/spaces/tiiuae/falcon-180b-demo/blob/d1590ee7fae9b6ce331ba7808e61a29dcce9239f/app.py#L28-L37
+register_conv_template(
+    Conversation(
+        name='falcon-chat',
+        roles=('User', 'Falcon'),
+        system_template='System: {system_message}',
+        messages=[],
+        sep_style=SeparatorStyle.FALCON_CHAT,
+        sep='\n',
+        sep2='<|endoftext|>',
+        stop_str='\nUser:',  # use stop_str to stop generation after stop_token_ids, it will also remove stop_str from the generated text
+    )
+)
+
+# Phind template
+# source: https://huggingface.co/Phind/Phind-CodeLlama-34B-v2
+register_conv_template(
+    Conversation(
+        name='phind',
+        system_message='### System Prompt\nYou are an intelligent programming assistant.',
+        roles=('### User Message', '### Assistant'),
+        messages=(),
+        offset=0,
+        sep_style=SeparatorStyle.ADD_COLON_SINGLE,
+        sep='\n\n',
+    )
+)
+
+# Metharme formatting for Pygmalion models
+# source: https://huggingface.co/PygmalionAI/pygmalion-2-13b
+register_conv_template(
+    Conversation(
+        name='metharme',
+        system_template='<|system|>{system_message}',
+        system_message="""Enter RP mode. You shall reply to the user while staying
+        in character. Your responses must be detailed, creative, immersive, and drive the scenario
+        forward.""",
+        roles=('<|user|>', '<|model|>'),
+        sep_style=SeparatorStyle.NO_COLON_SINGLE,
+        sep='',
+        stop_str='<|user|>',
+    )
+)
+
+# Zephyr template
+# reference: https://huggingface.co/spaces/HuggingFaceH4/zephyr-playground/blob/main/dialogues.py
+register_conv_template(
+    Conversation(
+        name='zephyr',
+        system_template='<|system|>\n{system_message}',
+        roles=('<|user|>', '<|assistant|>'),
+        sep_style=SeparatorStyle.CHATML,
+        sep='</s>',
+        stop_token_ids=[2],
+        stop_str='</s>',
+    )
+)
+
+# InternVL-ZH template
+register_conv_template(
+    Conversation(
+        name='internvl_zh',
+        system_template='',
+        roles=('<human>', '<bot>'),
+        sep_style=SeparatorStyle.INTERNVL_ZH,
+        sep=' ',
+        sep2='</s>',
+    )
+)
+
+
+if __name__ == '__main__':
+    from fastchat.conversation import get_conv_template
+
+    print('-- Vicuna template --')
+    conv = get_conv_template('vicuna_v1.1')
+    conv.append_message(conv.roles[0], 'Hello!')
+    conv.append_message(conv.roles[1], 'Hi!')
+    conv.append_message(conv.roles[0], 'How are you?')
+    conv.append_message(conv.roles[1], None)
+    print(conv.get_prompt())
+
+    print('\n')
+
+    print('-- Llama-2 template --')
+    conv = get_conv_template('llama-2')
+    conv.set_system_message('You are a helpful, respectful and honest assistant.')
+    conv.append_message(conv.roles[0], 'Hello!')
+    conv.append_message(conv.roles[1], 'Hi!')
+    conv.append_message(conv.roles[0], 'How are you?')
+    conv.append_message(conv.roles[1], None)
+    print(conv.get_prompt())
+
+    print('\n')
+
+    print('-- ChatGPT template --')
+    conv = get_conv_template('chatgpt')
+    conv.append_message(conv.roles[0], 'Hello!')
+    conv.append_message(conv.roles[1], 'Hi!')
+    conv.append_message(conv.roles[0], 'How are you?')
+    conv.append_message(conv.roles[1], None)
+    print(conv.to_openai_api_messages())
+
+    print('\n')
+
+    print('-- Claude template --')
+    conv = get_conv_template('claude')
+    conv.append_message(conv.roles[0], 'Hello!')
+    conv.append_message(conv.roles[1], 'Hi!')
+    conv.append_message(conv.roles[0], 'How are you?')
+    conv.append_message(conv.roles[1], None)
+    print(conv.get_prompt())

generation_config.json

@@ -1,9 +1,4 @@
 {
   "_from_model_config": true,
-  "transformers_version": "4.37.2",
-  "eos_token_id": [
-    2,
-    6,
-    7
-  ]
+  "transformers_version": "4.36.2"
 }

examples/red-panda.mp4 → mlp_projector.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d921c07bb97224d65a37801541d246067f0d506f08723ffa1ad85c217907ccb8
-size 1867237
+oid sha256:0db87dd44d7a45c72bd6a24fb26a970b8cf384a05e2ad8526be5be65865d647e
+size 286343367

modeling_intern_vit.py

@@ -1,9 +1,8 @@
 # --------------------------------------------------------
 # InternVL
-# Copyright (c) 2024 OpenGVLab
+# Copyright (c) 2023 OpenGVLab
 # Licensed under The MIT License [see LICENSE for details]
 # --------------------------------------------------------
-
 from typing import Optional, Tuple, Union
 
 import torch
@@ -21,14 +20,19 @@ from transformers.utils import logging
 from .configuration_intern_vit import InternVisionConfig
 
 try:
+    try:  # v1
+        from flash_attn.flash_attn_interface import \
+            flash_attn_unpadded_qkvpacked_func
+    except:  # v2
+        from flash_attn.flash_attn_interface import flash_attn_varlen_qkvpacked_func as flash_attn_unpadded_qkvpacked_func
+    
     from flash_attn.bert_padding import pad_input, unpad_input
-    from flash_attn.flash_attn_interface import \
-        flash_attn_varlen_qkvpacked_func
     has_flash_attn = True
 except:
-    print('FlashAttention2 is not installed.')
+    print('FlashAttention is not installed.')
     has_flash_attn = False
 
+
 logger = logging.get_logger(__name__)
 
 
@@ -69,7 +73,7 @@ class FlashAttention(nn.Module):
                 max_s = seqlen
                 cu_seqlens = torch.arange(0, (batch_size + 1) * seqlen, step=seqlen, dtype=torch.int32,
                                           device=qkv.device)
-                output = flash_attn_varlen_qkvpacked_func(
+                output = flash_attn_unpadded_qkvpacked_func(
                     qkv, cu_seqlens, max_s, self.dropout_p if self.training else 0.0,
                     softmax_scale=self.softmax_scale, causal=causal
                 )
@@ -79,7 +83,7 @@ class FlashAttention(nn.Module):
                 x = rearrange(qkv, 'b s three h d -> b s (three h d)')
                 x_unpad, indices, cu_seqlens, max_s = unpad_input(x, key_padding_mask)
                 x_unpad = rearrange(x_unpad, 'nnz (three h d) -> nnz three h d', three=3, h=nheads)
-                output_unpad = flash_attn_varlen_qkvpacked_func(
+                output_unpad = flash_attn_unpadded_qkvpacked_func(
                     x_unpad, cu_seqlens, max_s, self.dropout_p if self.training else 0.0,
                     softmax_scale=self.softmax_scale, causal=causal
                 )
@@ -88,7 +92,7 @@ class FlashAttention(nn.Module):
                                    'b s (h d) -> b s h d', h=nheads)
         else:
             assert max_s is not None
-            output = flash_attn_varlen_qkvpacked_func(
+            output = flash_attn_unpadded_qkvpacked_func(
                 qkv, cu_seqlens, max_s, self.dropout_p if self.training else 0.0,
                 softmax_scale=self.softmax_scale, causal=causal
             )
@@ -124,12 +128,6 @@ except Exception:
     pass
 
 
-NORM2FN = {
-    'rms_norm': InternRMSNorm,
-    'layer_norm': nn.LayerNorm,
-}
-
-
 class InternVisionEmbeddings(nn.Module):
     def __init__(self, config: InternVisionConfig):
         super().__init__()
@@ -151,26 +149,14 @@ class InternVisionEmbeddings(nn.Module):
 
         self.position_embedding = nn.Parameter(torch.randn(1, self.num_positions, self.embed_dim))
 
-    def _get_pos_embed(self, pos_embed, H, W):
-        target_dtype = pos_embed.dtype
-        pos_embed = pos_embed.float().reshape(
-            1, self.image_size // self.patch_size, self.image_size // self.patch_size, -1).permute(0, 3, 1, 2)
-        pos_embed = F.interpolate(pos_embed, size=(H, W), mode='bicubic', align_corners=False). \
-            reshape(1, -1, H * W).permute(0, 2, 1).to(target_dtype)
-        return pos_embed
-
     def forward(self, pixel_values: torch.FloatTensor) -> torch.Tensor:
+        batch_size = pixel_values.shape[0]
         target_dtype = self.patch_embedding.weight.dtype
-        patch_embeds = self.patch_embedding(pixel_values)  # shape = [*, channel, width, height]
-        batch_size, _, height, width = patch_embeds.shape
+        patch_embeds = self.patch_embedding(pixel_values)  # shape = [*, width, grid, grid]
         patch_embeds = patch_embeds.flatten(2).transpose(1, 2)
         class_embeds = self.class_embedding.expand(batch_size, 1, -1).to(target_dtype)
         embeddings = torch.cat([class_embeds, patch_embeds], dim=1)
-        position_embedding = torch.cat([
-            self.position_embedding[:, :1, :],
-            self._get_pos_embed(self.position_embedding[:, 1:, :], height, width)
-        ], dim=1)
-        embeddings = embeddings + position_embedding.to(target_dtype)
+        embeddings = embeddings + self.position_embedding.to(target_dtype)
         return embeddings
 
 
@@ -268,12 +254,11 @@ class InternVisionEncoderLayer(nn.Module):
         super().__init__()
         self.embed_dim = config.hidden_size
         self.intermediate_size = config.intermediate_size
-        self.norm_type = config.norm_type
 
         self.attn = InternAttention(config)
         self.mlp = InternMLP(config)
-        self.norm1 = NORM2FN[self.norm_type](self.embed_dim, eps=config.layer_norm_eps)
-        self.norm2 = NORM2FN[self.norm_type](self.embed_dim, eps=config.layer_norm_eps)
+        self.norm1 = InternRMSNorm(self.embed_dim, eps=config.layer_norm_eps)
+        self.norm2 = InternRMSNorm(self.embed_dim, eps=config.layer_norm_eps)
 
         self.ls1 = nn.Parameter(config.initializer_factor * torch.ones(self.embed_dim))
         self.ls2 = nn.Parameter(config.initializer_factor * torch.ones(self.embed_dim))
@@ -288,9 +273,9 @@ class InternVisionEncoderLayer(nn.Module):
         Args:
             hidden_states (`Tuple[torch.FloatTensor, Optional[torch.FloatTensor]]`): input to the layer of shape `(batch, seq_len, embed_dim)`
         """
-        hidden_states = hidden_states + self.drop_path1(self.attn(self.norm1(hidden_states).to(hidden_states.dtype)) * self.ls1)
+        hidden_states = hidden_states + self.drop_path1(self.attn(self.norm1(hidden_states)) * self.ls1)
 
-        hidden_states = hidden_states + self.drop_path2(self.mlp(self.norm2(hidden_states).to(hidden_states.dtype)) * self.ls2)
+        hidden_states = hidden_states + self.drop_path2(self.mlp(self.norm2(hidden_states)) * self.ls2)
 
         return hidden_states
 
@@ -363,10 +348,8 @@ class InternVisionEncoder(nn.Module):
 
 class InternVisionModel(PreTrainedModel):
     main_input_name = 'pixel_values'
-    _supports_flash_attn_2 = True
-    supports_gradient_checkpointing = True
     config_class = InternVisionConfig
-    _no_split_modules = ['InternVisionEncoderLayer']
+    _no_split_modules = ['InternAttention']
 
     def __init__(self, config: InternVisionConfig):
         super().__init__(config)
@@ -384,7 +367,6 @@ class InternVisionModel(PreTrainedModel):
         pos_emb = pos_emb.to(cls_emb.dtype).reshape(1, embed_dim, -1).permute(0, 2, 1)
         pos_emb = torch.cat([cls_emb, pos_emb], dim=1)
         self.embeddings.position_embedding = nn.Parameter(pos_emb)
-        self.embeddings.image_size = new_size
         logger.info('Resized position embeddings from {} to {}'.format(old_size, new_size))
 
     def get_input_embeddings(self):

modeling_internvl_chat.py

@@ -1,61 +1,39 @@
 # --------------------------------------------------------
 # InternVL
-# Copyright (c) 2024 OpenGVLab
+# Copyright (c) 2023 OpenGVLab
 # Licensed under The MIT License [see LICENSE for details]
 # --------------------------------------------------------
-import warnings
 from typing import Any, List, Optional, Tuple, Union
 
 import torch.utils.checkpoint
-import transformers
+from peft import LoraConfig, get_peft_model
 from torch import nn
 from torch.nn import CrossEntropyLoss
-from transformers import AutoModel, GenerationConfig, LlamaForCausalLM
+from transformers import GenerationConfig, LlamaForCausalLM, LlamaTokenizer
 from transformers.modeling_outputs import CausalLMOutputWithPast
 from transformers.modeling_utils import PreTrainedModel
 from transformers.utils import ModelOutput, logging
 
 from .configuration_internvl_chat import InternVLChatConfig
-from .conversation import get_conv_template
-from .modeling_intern_vit import InternVisionModel, has_flash_attn
+from .modeling_intern_vit import InternVisionModel
 
 logger = logging.get_logger(__name__)
 
 
-def version_cmp(v1, v2, op='eq'):
-    import operator
-
-    from packaging import version
-    op_func = getattr(operator, op)
-    return op_func(version.parse(v1), version.parse(v2))
-
-
 class InternVLChatModel(PreTrainedModel):
     config_class = InternVLChatConfig
     main_input_name = 'pixel_values'
-    base_model_prefix = 'language_model'
-    _supports_flash_attn_2 = True
-    supports_gradient_checkpointing = True
-    _no_split_modules = ['InternVisionModel', 'LlamaDecoderLayer']
+    _no_split_modules = ['InternAttention', 'LlamaDecoderLayer', 'LlamaForCausalLM']
 
-    def __init__(self, config: InternVLChatConfig, vision_model=None, language_model=None, use_flash_attn=True):
+    def __init__(self, config: InternVLChatConfig, vision_model=None, language_model=None):
         super().__init__(config)
 
-        assert version_cmp(transformers.__version__, '4.37.0', 'ge')
         image_size = config.force_image_size or config.vision_config.image_size
         patch_size = config.vision_config.patch_size
-        self.patch_size = patch_size
         self.select_layer = config.select_layer
         self.template = config.template
         self.num_image_token = int((image_size // patch_size) ** 2 * (config.downsample_ratio ** 2))
-        self.downsample_ratio = config.downsample_ratio
-        self.ps_version = config.ps_version
-        use_flash_attn = use_flash_attn if has_flash_attn else False
-        config.vision_config.use_flash_attn = True if use_flash_attn else False
-        config.llm_config._attn_implementation = 'flash_attention_2' if use_flash_attn else 'eager'
-
         logger.info(f'num_image_token: {self.num_image_token}')
-        logger.info(f'ps_version: {self.ps_version}')
         if vision_model is not None:
             self.vision_model = vision_model
         else:
@@ -63,24 +41,53 @@ class InternVLChatModel(PreTrainedModel):
         if language_model is not None:
             self.language_model = language_model
         else:
-            if config.llm_config.architectures[0] == 'LlamaForCausalLM':
-                self.language_model = LlamaForCausalLM(config.llm_config)
-            else:
-                raise NotImplementedError(f'{config.llm_config.architectures[0]} is not implemented.')
-
+            self.language_model = LlamaForCausalLM(config.llm_config)
         vit_hidden_size = config.vision_config.hidden_size
         llm_hidden_size = config.llm_config.hidden_size
 
         self.mlp1 = nn.Sequential(
-            nn.LayerNorm(vit_hidden_size * int(1 / self.downsample_ratio) ** 2),
-            nn.Linear(vit_hidden_size * int(1 / self.downsample_ratio) ** 2, llm_hidden_size),
+            nn.LayerNorm(vit_hidden_size * 4),
+            nn.Linear(vit_hidden_size * 4, llm_hidden_size),
             nn.GELU(),
             nn.Linear(llm_hidden_size, llm_hidden_size)
         )
 
+        if config.force_image_size:
+            self.vision_model.resize_pos_embeddings(
+                old_size=config.vision_config.image_size,
+                new_size=config.force_image_size,
+                patch_size=config.vision_config.patch_size
+            )
+
         self.img_context_token_id = None
-        self.conv_template = get_conv_template(self.template)
-        self.system_message = self.conv_template.system_message
+
+        if config.use_backbone_lora:
+            self.wrap_backbone_lora(r=config.use_backbone_lora)
+
+        if config.use_llm_lora:
+            self.wrap_llm_lora(r=config.use_llm_lora)
+
+    def wrap_backbone_lora(self, r=128, lora_alpha=256, lora_dropout=0.05):
+        lora_config = LoraConfig(
+            r=r,
+            target_modules=['attn.qkv', 'attn.proj', 'mlp.fc1', 'mlp.fc2'],
+            lora_alpha=lora_alpha,
+            lora_dropout=lora_dropout,
+        )
+        self.vision_model = get_peft_model(self.vision_model, lora_config)
+        self.vision_model.print_trainable_parameters()
+
+    def wrap_llm_lora(self, r=128, lora_alpha=256, lora_dropout=0.05):
+        lora_config = LoraConfig(
+            r=r,
+            target_modules=['self_attn.q_proj', 'self_attn.k_proj', 'self_attn.v_proj', 'self_attn.o_proj',
+                            'mlp.gate_proj', 'mlp.down_proj', 'mlp.up_proj'],
+            lora_alpha=lora_alpha,
+            lora_dropout=lora_dropout,
+            task_type='CAUSAL_LM'
+        )
+        self.language_model = get_peft_model(self.language_model, lora_config)
+        self.language_model.print_trainable_parameters()
 
     def forward(
             self,
@@ -99,32 +106,24 @@ class InternVLChatModel(PreTrainedModel):
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict
 
         image_flags = image_flags.squeeze(-1)
-        input_embeds = self.language_model.get_input_embeddings()(input_ids).clone()
+        input_embeds = self.language_model.get_input_embeddings()(input_ids)
 
         vit_embeds = self.extract_feature(pixel_values)
         vit_embeds = vit_embeds[image_flags == 1]
-        vit_batch_size = pixel_values.shape[0]
 
         B, N, C = input_embeds.shape
         input_embeds = input_embeds.reshape(B * N, C)
 
-        if torch.distributed.is_initialized() and torch.distributed.get_rank() == 0:
-            print(f'dynamic ViT batch size: {vit_batch_size}, images per sample: {vit_batch_size / B}, dynamic token length: {N}')
-
         input_ids = input_ids.reshape(B * N)
         selected = (input_ids == self.img_context_token_id)
         try:
             input_embeds[selected] = input_embeds[selected] * 0.0 + vit_embeds.reshape(-1, C)
-        except Exception as e:
-            vit_embeds = vit_embeds.reshape(-1, C)
-            print(f'warning: {e}, input_embeds[selected].shape={input_embeds[selected].shape}, '
-                  f'vit_embeds.shape={vit_embeds.shape}')
-            n_token = selected.sum()
-            input_embeds[selected] = input_embeds[selected] * 0.0 + vit_embeds[:n_token]
+        except:
+            pass
 
         input_embeds = input_embeds.reshape(B, N, C)
 
-        outputs = self.language_model(
+        outputs = self.language_model.model(
             inputs_embeds=input_embeds,
             attention_mask=attention_mask,
             position_ids=position_ids,
@@ -134,7 +133,8 @@ class InternVLChatModel(PreTrainedModel):
             output_hidden_states=output_hidden_states,
             return_dict=return_dict,
         )
-        logits = outputs.logits
+        hidden_states = outputs[0]
+        logits = self.language_model.lm_head(hidden_states)
 
         loss = None
         if labels is not None:
@@ -170,120 +170,41 @@ class InternVLChatModel(PreTrainedModel):
         # N, H * scale, W, C // scale --> N, H * scale, W * scale, C // (scale ** 2)
         x = x.view(n, int(h * scale_factor), int(w * scale_factor),
                    int(c / (scale_factor * scale_factor)))
-        if self.ps_version == 'v1':
-            warnings.warn("In ps_version 'v1', the height and width have not been swapped back, "
-                          'which results in a transposed image.')
-        else:
-            x = x.permute(0, 2, 1, 3).contiguous()
         return x
 
     def extract_feature(self, pixel_values):
-        if self.select_layer == -1:
-            vit_embeds = self.vision_model(
-                pixel_values=pixel_values,
-                output_hidden_states=False,
-                return_dict=True).last_hidden_state
-        else:
-            vit_embeds = self.vision_model(
-                pixel_values=pixel_values,
-                output_hidden_states=True,
-                return_dict=True).hidden_states[self.select_layer]
+        vit_embeds = self.vision_model(
+            pixel_values=pixel_values,
+            output_hidden_states=True,
+            return_dict=True).hidden_states[-4]
         vit_embeds = vit_embeds[:, 1:, :]
-
+        # if torch.distributed.get_rank() == 0:
+        #     print("before pixel shuffle:", vit_embeds.shape)
         h = w = int(vit_embeds.shape[1] ** 0.5)
         vit_embeds = vit_embeds.reshape(vit_embeds.shape[0], h, w, -1)
-        vit_embeds = self.pixel_shuffle(vit_embeds, scale_factor=self.downsample_ratio)
+        vit_embeds = self.pixel_shuffle(vit_embeds, scale_factor=0.5)
         vit_embeds = vit_embeds.reshape(vit_embeds.shape[0], -1, vit_embeds.shape[-1])
+        # if torch.distributed.get_rank() == 0:
+        #     print("after pixel shuffle:", vit_embeds.shape)
         vit_embeds = self.mlp1(vit_embeds)
         return vit_embeds
 
-    def batch_chat(self, tokenizer, pixel_values, questions, generation_config, num_patches_list=None,
-                   history=None, return_history=False, IMG_START_TOKEN='<img>', IMG_END_TOKEN='</img>',
-                   IMG_CONTEXT_TOKEN='<IMG_CONTEXT>', verbose=False, image_counts=None):
-        if history is not None or return_history:
-            print('Now multi-turn chat is not supported in batch_chat.')
-            raise NotImplementedError
-
-        if image_counts is not None:
-            num_patches_list = image_counts
-            print('Warning: `image_counts` is deprecated. Please use `num_patches_list` instead.')
-
-        img_context_token_id = tokenizer.convert_tokens_to_ids(IMG_CONTEXT_TOKEN)
-        self.img_context_token_id = img_context_token_id
-
-        if verbose and pixel_values is not None:
-            image_bs = pixel_values.shape[0]
-            print(f'dynamic ViT batch size: {image_bs}')
-
-        queries = []
-        for idx, num_patches in enumerate(num_patches_list):
-            question = questions[idx]
-            if pixel_values is not None and '<image>' not in question:
-                question = '<image>\n' + question
-            template = get_conv_template(self.template)
-            template.system_message = self.system_message
-            template.append_message(template.roles[0], question)
-            template.append_message(template.roles[1], None)
-            query = template.get_prompt()
-
-            image_tokens = IMG_START_TOKEN + IMG_CONTEXT_TOKEN * self.num_image_token * num_patches + IMG_END_TOKEN
-            query = query.replace('<image>', image_tokens, 1)
-            queries.append(query)
-
-        tokenizer.padding_side = 'left'
-        model_inputs = tokenizer(queries, return_tensors='pt', padding=True)
-        input_ids = model_inputs['input_ids'].to(self.device)
-        attention_mask = model_inputs['attention_mask'].to(self.device)
-        eos_token_id = tokenizer.convert_tokens_to_ids(template.sep)
-        generation_config['eos_token_id'] = eos_token_id
-        generation_output = self.generate(
-            pixel_values=pixel_values,
-            input_ids=input_ids,
-            attention_mask=attention_mask,
-            **generation_config
-        )
-        responses = tokenizer.batch_decode(generation_output, skip_special_tokens=True)
-        responses = [response.split(template.sep)[0].strip() for response in responses]
-        return responses
-
-    def chat(self, tokenizer, pixel_values, question, generation_config, history=None, return_history=False,
-             num_patches_list=None, IMG_START_TOKEN='<img>', IMG_END_TOKEN='</img>', IMG_CONTEXT_TOKEN='<IMG_CONTEXT>',
-             verbose=False):
-
-        if history is None and pixel_values is not None and '<image>' not in question:
-            question = '<image>\n' + question
-
-        if num_patches_list is None:
-            num_patches_list = [pixel_values.shape[0]] if pixel_values is not None else []
-        assert pixel_values is None or len(pixel_values) == sum(num_patches_list)
+    def chat(self, tokenizer, pixel_values, question, generation_config,
+             IMG_START_TOKEN='<img>', IMG_END_TOKEN='</img>', IMG_CONTEXT_TOKEN='<IMG_CONTEXT>'):
 
         img_context_token_id = tokenizer.convert_tokens_to_ids(IMG_CONTEXT_TOKEN)
         self.img_context_token_id = img_context_token_id
+        from .conversation import get_conv_template
 
         template = get_conv_template(self.template)
-        template.system_message = self.system_message
-        eos_token_id = tokenizer.convert_tokens_to_ids(template.sep)
-
-        history = [] if history is None else history
-        for (old_question, old_answer) in history:
-            template.append_message(template.roles[0], old_question)
-            template.append_message(template.roles[1], old_answer)
-        template.append_message(template.roles[0], question)
+        image_tokens = IMG_START_TOKEN + IMG_CONTEXT_TOKEN * self.num_image_token + IMG_END_TOKEN
+        template.append_message(template.roles[0], image_tokens + '\n' + question)
         template.append_message(template.roles[1], None)
         query = template.get_prompt()
-
-        if verbose and pixel_values is not None:
-            image_bs = pixel_values.shape[0]
-            print(f'dynamic ViT batch size: {image_bs}')
-
-        for num_patches in num_patches_list:
-            image_tokens = IMG_START_TOKEN + IMG_CONTEXT_TOKEN * self.num_image_token * num_patches + IMG_END_TOKEN
-            query = query.replace('<image>', image_tokens, 1)
-
         model_inputs = tokenizer(query, return_tensors='pt')
-        input_ids = model_inputs['input_ids'].to(self.device)
-        attention_mask = model_inputs['attention_mask'].to(self.device)
-        generation_config['eos_token_id'] = eos_token_id
+        input_ids = model_inputs['input_ids'].cuda()
+        attention_mask = model_inputs['attention_mask'].cuda()
+
         generation_output = self.generate(
             pixel_values=pixel_values,
             input_ids=input_ids,
@@ -291,16 +212,9 @@ class InternVLChatModel(PreTrainedModel):
             **generation_config
         )
         response = tokenizer.batch_decode(generation_output, skip_special_tokens=True)[0]
-        response = response.split(template.sep)[0].strip()
-        history.append((question, response))
-        if return_history:
-            return response, history
-        else:
-            query_to_print = query.replace(IMG_CONTEXT_TOKEN, '')
-            query_to_print = query_to_print.replace(f'{IMG_START_TOKEN}{IMG_END_TOKEN}', '<image>')
-            if verbose:
-                print(query_to_print, response)
-            return response
+        query_to_print = query.replace(image_tokens, '<image>')
+        print(query_to_print, response)
+        return response
 
     @torch.no_grad()
     def generate(
@@ -321,6 +235,7 @@ class InternVLChatModel(PreTrainedModel):
                 vit_embeds = visual_features
             else:
                 vit_embeds = self.extract_feature(pixel_values)
+
             input_embeds = self.language_model.get_input_embeddings()(input_ids)
             B, N, C = input_embeds.shape
             input_embeds = input_embeds.reshape(B * N, C)
@@ -328,7 +243,7 @@ class InternVLChatModel(PreTrainedModel):
             input_ids = input_ids.reshape(B * N)
             selected = (input_ids == self.img_context_token_id)
             assert selected.sum() != 0
-            input_embeds[selected] = vit_embeds.reshape(-1, C).to(input_embeds.device)
+            input_embeds[selected] = vit_embeds.reshape(-1, C)
 
             input_embeds = input_embeds.reshape(B, N, C)
         else:
@@ -345,13 +260,3 @@ class InternVLChatModel(PreTrainedModel):
         )
 
         return outputs
-
-    @property
-    def lm_head(self):
-        return self.language_model.get_output_embeddings()
-
-    def get_input_embeddings(self):
-        return self.language_model.get_input_embeddings()
-
-    def get_output_embeddings(self):
-        return self.language_model.get_output_embeddings()

tokenizer_config.json

@@ -131,7 +131,7 @@
   "clean_up_tokenization_spaces": false,
   "eos_token": "<|im_end|>",
   "legacy": true,
-  "model_max_length": 4096,
+  "model_max_length": 2048,
   "pad_token": "<unk>",
   "sp_model_kwargs": {},
   "spaces_between_special_tokens": false,

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 1.0,
+    "train_loss": 0.8886408911725423,
+    "train_runtime": 65743.9481,
+    "train_samples": 1267819,
+    "train_samples_per_second": 19.284,
+    "train_steps_per_second": 0.038
+}

examples/image2.jpg → training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:08487494b8dc08d44bc36491adf3ab89ff30d13a3122da86f3cd67cad89eeee8
-size 125656
+oid sha256:340429e4fb7f823a8b30b6d28f94e36d2bab5f5aca9a65b300d4425fdae55def
+size 6011