Upload model

README.md

@@ -0,0 +1,199 @@
+---
+library_name: transformers
+tags: []
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]

config.json

@@ -0,0 +1,308 @@
+{
+  "_commit_hash": null,
+  "_name_or_path": "/workspace/final-model/mini-internvl-chat-4b-v1_5/v2-20240705-133804/checkpoint-250-merged",
+  "architectures": [
+    "InternVLChatModel"
+  ],
+  "auto_map": {
+    "AutoConfig": "configuration_internvl_chat.InternVLChatConfig",
+    "AutoModel": "modeling_internvl_chat.InternVLChatModel",
+    "AutoModelForCausalLM": "modeling_internvl_chat.InternVLChatModel"
+  },
+  "downsample_ratio": 0.5,
+  "dynamic_image_size": true,
+  "force_image_size": 448,
+  "hidden_size": 3072,
+  "llm_config": {
+    "_name_or_path": "pretrained/Phi-3-mini-128k-instruct/",
+    "add_cross_attention": false,
+    "architectures": [
+      "Phi3ForCausalLM"
+    ],
+    "attention_dropout": 0.0,
+    "auto_map": {
+      "AutoConfig": "configuration_phi3.Phi3Config",
+      "AutoModel": "modeling_phi3.Phi3ForCausalLM",
+      "AutoModelForCausalLM": "modeling_phi3.Phi3ForCausalLM"
+    },
+    "bad_words_ids": null,
+    "begin_suppress_tokens": null,
+    "bos_token_id": 1,
+    "chunk_size_feed_forward": 0,
+    "cross_attention_hidden_size": null,
+    "decoder_start_token_id": null,
+    "diversity_penalty": 0.0,
+    "do_sample": false,
+    "early_stopping": false,
+    "embd_pdrop": 0.0,
+    "encoder_no_repeat_ngram_size": 0,
+    "eos_token_id": 32000,
+    "exponential_decay_length_penalty": null,
+    "finetuning_task": null,
+    "forced_bos_token_id": null,
+    "forced_eos_token_id": null,
+    "hidden_act": "silu",
+    "hidden_size": 3072,
+    "id2label": {
+      "0": "LABEL_0",
+      "1": "LABEL_1"
+    },
+    "initializer_range": 0.02,
+    "intermediate_size": 8192,
+    "is_decoder": false,
+    "is_encoder_decoder": false,
+    "label2id": {
+      "LABEL_0": 0,
+      "LABEL_1": 1
+    },
+    "length_penalty": 1.0,
+    "max_length": 20,
+    "max_position_embeddings": 131072,
+    "min_length": 0,
+    "model_type": "phi3",
+    "no_repeat_ngram_size": 0,
+    "num_attention_heads": 32,
+    "num_beam_groups": 1,
+    "num_beams": 1,
+    "num_hidden_layers": 32,
+    "num_key_value_heads": 32,
+    "num_return_sequences": 1,
+    "original_max_position_embeddings": 4096,
+    "output_attentions": false,
+    "output_hidden_states": false,
+    "output_scores": false,
+    "pad_token_id": 32000,
+    "prefix": null,
+    "problem_type": null,
+    "pruned_heads": {},
+    "remove_invalid_values": false,
+    "repetition_penalty": 1.0,
+    "resid_pdrop": 0.0,
+    "return_dict": true,
+    "return_dict_in_generate": false,
+    "rms_norm_eps": 1e-05,
+    "rope_scaling": {
+      "long_factor": [
+        1.0299999713897705,
+        1.0499999523162842,
+        1.0499999523162842,
+        1.0799999237060547,
+        1.2299998998641968,
+        1.2299998998641968,
+        1.2999999523162842,
+        1.4499999284744263,
+        1.5999999046325684,
+        1.6499998569488525,
+        1.8999998569488525,
+        2.859999895095825,
+        3.68999981880188,
+        5.419999599456787,
+        5.489999771118164,
+        5.489999771118164,
+        9.09000015258789,
+        11.579999923706055,
+        15.65999984741211,
+        15.769999504089355,
+        15.789999961853027,
+        18.360000610351562,
+        21.989999771118164,
+        23.079999923706055,
+        30.009998321533203,
+        32.35000228881836,
+        32.590003967285156,
+        35.56000518798828,
+        39.95000457763672,
+        53.840003967285156,
+        56.20000457763672,
+        57.95000457763672,
+        59.29000473022461,
+        59.77000427246094,
+        59.920005798339844,
+        61.190006256103516,
+        61.96000671386719,
+        62.50000762939453,
+        63.3700065612793,
+        63.48000717163086,
+        63.48000717163086,
+        63.66000747680664,
+        63.850006103515625,
+        64.08000946044922,
+        64.760009765625,
+        64.80001068115234,
+        64.81001281738281,
+        64.81001281738281
+      ],
+      "short_factor": [
+        1.05,
+        1.05,
+        1.05,
+        1.1,
+        1.1,
+        1.1500000000000001,
+        1.2000000000000002,
+        1.2500000000000002,
+        1.3000000000000003,
+        1.3500000000000003,
+        1.5000000000000004,
+        2.000000000000001,
+        2.000000000000001,
+        2.000000000000001,
+        2.000000000000001,
+        2.000000000000001,
+        2.000000000000001,
+        2.000000000000001,
+        2.000000000000001,
+        2.000000000000001,
+        2.000000000000001,
+        2.000000000000001,
+        2.000000000000001,
+        2.000000000000001,
+        2.000000000000001,
+        2.000000000000001,
+        2.000000000000001,
+        2.000000000000001,
+        2.000000000000001,
+        2.000000000000001,
+        2.000000000000001,
+        2.000000000000001,
+        2.0500000000000007,
+        2.0500000000000007,
+        2.0500000000000007,
+        2.1000000000000005,
+        2.1000000000000005,
+        2.1000000000000005,
+        2.1500000000000004,
+        2.1500000000000004,
+        2.3499999999999996,
+        2.549999999999999,
+        2.5999999999999988,
+        2.5999999999999988,
+        2.7499999999999982,
+        2.849999999999998,
+        2.849999999999998,
+        2.9499999999999975
+      ],
+      "type": "su"
+    },
+    "rope_theta": 10000.0,
+    "sep_token_id": null,
+    "sliding_window": 262144,
+    "suppress_tokens": null,
+    "task_specific_params": null,
+    "temperature": 1.0,
+    "tf_legacy_loss": false,
+    "tie_encoder_decoder": false,
+    "tie_word_embeddings": false,
+    "tokenizer_class": null,
+    "top_k": 50,
+    "top_p": 1.0,
+    "torch_dtype": "bfloat16",
+    "torchscript": false,
+    "transformers_version": "4.42.3",
+    "typical_p": 1.0,
+    "use_bfloat16": false,
+    "use_cache": true,
+    "vocab_size": 32020
+  },
+  "max_dynamic_patch": 12,
+  "max_position_embeddings": 131072,
+  "min_dynamic_patch": 1,
+  "model_type": "internvl_chat",
+  "pad2square": false,
+  "ps_version": "v2",
+  "select_layer": -1,
+  "template": "phi3-chat",
+  "torch_dtype": "bfloat16",
+  "transformers_version": null,
+  "use_backbone_lora": 0,
+  "use_llm_lora": 0,
+  "use_thumbnail": true,
+  "vision_config": {
+    "_name_or_path": "OpenGVLab/InternViT-300M-448px",
+    "add_cross_attention": false,
+    "architectures": [
+      "InternVisionModel"
+    ],
+    "attention_dropout": 0.0,
+    "auto_map": {
+      "AutoConfig": "configuration_intern_vit.InternVisionConfig",
+      "AutoModel": "modeling_intern_vit.InternVisionModel"
+    },
+    "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": 0.1,
+    "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": 1024,
+    "id2label": {
+      "0": "LABEL_0",
+      "1": "LABEL_1"
+    },
+    "image_size": 448,
+    "initializer_factor": 1.0,
+    "initializer_range": 0.02,
+    "intermediate_size": 4096,
+    "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",
+    "no_repeat_ngram_size": 0,
+    "norm_type": "layer_norm",
+    "num_attention_heads": 16,
+    "num_beam_groups": 1,
+    "num_beams": 1,
+    "num_channels": 3,
+    "num_hidden_layers": 24,
+    "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": false,
+    "qkv_bias": true,
+    "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": "4.42.3",
+    "typical_p": 1.0,
+    "use_bfloat16": true,
+    "use_flash_attn": false
+  }
+}

configuration_intern_vit.py

@@ -0,0 +1,119 @@
+# --------------------------------------------------------
+# InternVL
+# Copyright (c) 2023 OpenGVLab
+# Licensed under The MIT License [see LICENSE for details]
+# --------------------------------------------------------
+import os
+from typing import Union
+
+from transformers.configuration_utils import PretrainedConfig
+from transformers.utils import logging
+
+logger = logging.get_logger(__name__)
+
+
+class InternVisionConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`InternVisionModel`]. It is used to
+    instantiate a vision encoder according to the specified arguments, defining the model architecture.
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+    Args:
+        num_channels (`int`, *optional*, defaults to 3):
+            Number of color channels in the input images (e.g., 3 for RGB).
+        patch_size (`int`, *optional*, defaults to 14):
+            The size (resolution) of each patch.
+        image_size (`int`, *optional*, defaults to 224):
+            The size (resolution) of each image.
+        qkv_bias (`bool`, *optional*, defaults to `False`):
+            Whether to add a bias to the queries and values in the self-attention layers.
+        hidden_size (`int`, *optional*, defaults to 3200):
+            Dimensionality of the encoder layers and the pooler layer.
+        num_attention_heads (`int`, *optional*, defaults to 25):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        intermediate_size (`int`, *optional*, defaults to 12800):
+            Dimensionality of the "intermediate" (i.e., feed-forward) layer in the Transformer encoder.
+        qk_normalization (`bool`, *optional*, defaults to `True`):
+            Whether to normalize the queries and keys in the self-attention layers.
+        num_hidden_layers (`int`, *optional*, defaults to 48):
+            Number of hidden layers in the Transformer encoder.
+        use_flash_attn (`bool`, *optional*, defaults to `True`):
+            Whether to use flash attention mechanism.
+        hidden_act (`str` or `function`, *optional*, defaults to `"gelu"`):
+            The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,
+            `"relu"`, `"selu"` and `"gelu_new"` ``"gelu"` are supported.
+        layer_norm_eps (`float`, *optional*, defaults to 1e-6):
+            The epsilon used by the layer normalization layers.
+        dropout (`float`, *optional*, defaults to 0.0):
+            The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.
+        drop_path_rate (`float`, *optional*, defaults to 0.0):
+            Dropout rate for stochastic depth.
+        attention_dropout (`float`, *optional*, defaults to 0.0):
+            The dropout ratio for the attention probabilities.
+        initializer_range (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        initializer_factor (`float`, *optional*, defaults to 0.1):
+            A factor for layer scale.
+    """
+
+    model_type = 'intern_vit_6b'
+
+    def __init__(
+            self,
+            num_channels=3,
+            patch_size=14,
+            image_size=224,
+            qkv_bias=False,
+            hidden_size=3200,
+            num_attention_heads=25,
+            intermediate_size=12800,
+            qk_normalization=True,
+            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,
+            attention_dropout=0.0,
+            initializer_range=0.02,
+            initializer_factor=0.1,
+            **kwargs,
+    ):
+        super().__init__(**kwargs)
+
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.dropout = dropout
+        self.drop_path_rate = drop_path_rate
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.num_channels = num_channels
+        self.patch_size = patch_size
+        self.image_size = image_size
+        self.initializer_range = initializer_range
+        self.initializer_factor = initializer_factor
+        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
+
+    @classmethod
+    def from_pretrained(cls, pretrained_model_name_or_path: Union[str, os.PathLike], **kwargs) -> 'PretrainedConfig':
+        config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs)
+
+        if 'vision_config' in config_dict:
+            config_dict = config_dict['vision_config']
+
+        if 'model_type' in config_dict and hasattr(cls, 'model_type') and config_dict['model_type'] != cls.model_type:
+            logger.warning(
+                f"You are using a model of type {config_dict['model_type']} to instantiate a model of type "
+                f'{cls.model_type}. This is not supported for all configurations of models and can yield errors.'
+            )
+
+        return cls.from_dict(config_dict, **kwargs)

configuration_internvl_chat.py

@@ -0,0 +1,99 @@
+# --------------------------------------------------------
+# InternVL
+# Copyright (c) 2023 OpenGVLab
+# Licensed under The MIT License [see LICENSE for details]
+# --------------------------------------------------------
+
+import copy
+
+from transformers import AutoConfig, LlamaConfig
+from transformers.configuration_utils import PretrainedConfig
+from transformers.utils import logging
+
+from .configuration_intern_vit import InternVisionConfig
+from .configuration_phi3 import Phi3Config
+
+logger = logging.get_logger(__name__)
+
+
+class InternVLChatConfig(PretrainedConfig):
+    model_type = 'internvl_chat'
+    is_composition = True
+
+    def __init__(
+            self,
+            vision_config=None,
+            llm_config=None,
+            use_backbone_lora=0,
+            use_llm_lora=0,
+            pad2square=False,
+            select_layer=-1,
+            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 = {}
+            logger.info('vision_config is None. Initializing the InternVisionConfig with default values.')
+
+        if llm_config is None:
+            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)
+        elif llm_config['architectures'][0] == 'Phi3ForCausalLM':
+            self.llm_config = Phi3Config(**llm_config)
+        else:
+            raise ValueError('Unsupported architecture: {}'.format(llm_config['architectures'][0]))
+        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
+
+        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):
+        """
+        Serializes this instance to a Python dictionary. Override the default [`~PretrainedConfig.to_dict`].
+
+        Returns:
+            `Dict[str, any]`: Dictionary of all the attributes that make up this configuration instance,
+        """
+        output = copy.deepcopy(self.__dict__)
+        output['vision_config'] = self.vision_config.to_dict()
+        output['llm_config'] = self.llm_config.to_dict()
+        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

configuration_phi3.py

@@ -0,0 +1,211 @@
+# Copyright 2024 Microsoft and the HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License atd
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+""" Phi-3 model configuration"""
+
+
+from transformers.configuration_utils import PretrainedConfig
+from transformers.utils import logging
+
+logger = logging.get_logger(__name__)
+
+PHI3_PRETRAINED_CONFIG_ARCHIVE_MAP = {
+    'microsoft/Phi-3-mini-4k-instruct': 'https://huggingface.co/microsoft/Phi-3-mini-4k-instruct/resolve/main/config.json',
+    'microsoft/Phi-3-mini-128k-instruct': 'https://huggingface.co/microsoft/Phi-3-mini-128k-instruct/resolve/main/config.json',
+}
+
+
+class Phi3Config(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`Phi3Model`]. It is used to instantiate a Phi-3
+    model according to the specified arguments, defining the model architecture. Instantiating a configuration with the
+    defaults will yield a similar configuration to that of the
+    [microsoft/Phi-3-mini-4k-instruct](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct).
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+    Args:
+        vocab_size (`int`, *optional*, defaults to 32064):
+            Vocabulary size of the Phi-3 model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`Phi3Model`].
+        hidden_size (`int`, *optional*, defaults to 3072):
+            Dimension of the hidden representations.
+        intermediate_size (`int`, *optional*, defaults to 8192):
+            Dimension of the MLP representations.
+        num_hidden_layers (`int`, *optional*, defaults to 32):
+            Number of hidden layers in the Transformer decoder.
+        num_attention_heads (`int`, *optional*, defaults to 32):
+            Number of attention heads for each attention layer in the Transformer decoder.
+        num_key_value_heads (`int`, *optional*):
+            This is the number of key_value heads that should be used to implement Grouped Query Attention. If
+            `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if
+            `num_key_value_heads=1 the model will use Multi Query Attention (MQA) otherwise GQA is used. When
+            converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed
+            by meanpooling all the original heads within that group. For more details checkout [this
+            paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to
+            `num_attention_heads`.
+        resid_pdrop (`float`, *optional*, defaults to 0.0):
+            Dropout probability for mlp outputs.
+        embd_pdrop (`int`, *optional*, defaults to 0.0):
+            The dropout ratio for the embeddings.
+        attention_dropout (`float`, *optional*, defaults to 0.0):
+            The dropout ratio after computing the attention scores.
+        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
+            The non-linear activation function (function or string) in the decoder.
+        max_position_embeddings (`int`, *optional*, defaults to 4096):
+            The maximum sequence length that this model might ever be used with.
+        original_max_position_embeddings (`int`, *optional*, defaults to 4096):
+            The maximum sequence length that this model was trained with. This is used to determine the size of the
+            original RoPE embeddings when using long scaling.
+        initializer_range (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        rms_norm_eps (`float`, *optional*, defaults to 1e-05):
+            The epsilon value used for the RMSNorm.
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether or not the model should return the last key/values attentions (not used by all models). Only
+            relevant if `config.is_decoder=True`. Whether to tie weight embeddings or not.
+        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
+            Whether to tie weight embeddings
+        rope_theta (`float`, *optional*, defaults to 10000.0):
+            The base period of the RoPE embeddings.
+        rope_scaling (`dict`, *optional*):
+            The scaling strategy for the RoPE embeddings. If `None`, no scaling is applied. If a dictionary, it must
+            contain the following keys: `type`, `short_factor` and `long_factor`. The `type` must be either `su` or `yarn` and
+            the `short_factor` and `long_factor` must be lists of numbers with the same length as the hidden size
+            divided by the number of attention heads divided by 2.
+        bos_token_id (`int`, *optional*, defaults to 1):
+            The id of the "beginning-of-sequence" token.
+        eos_token_id (`int`, *optional*, defaults to 32000):
+            The id of the "end-of-sequence" token.
+        pad_token_id (`int`, *optional*, defaults to 32000):
+            The id of the padding token.
+        sliding_window (`int`, *optional*):
+            Sliding window attention window size. If `None`, no sliding window is applied.
+
+    Example:
+
+    ```python
+    >>> from transformers import Phi3Model, Phi3Config
+
+    >>> # Initializing a Phi-3 style configuration
+    >>> configuration = Phi3Config.from_pretrained("microsoft/Phi-3-mini-4k-instruct")
+
+    >>> # Initializing a model from the configuration
+    >>> model = Phi3Model(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = 'phi3'
+    keys_to_ignore_at_inference = ['past_key_values']
+
+    def __init__(
+        self,
+        vocab_size=32064,
+        hidden_size=3072,
+        intermediate_size=8192,
+        num_hidden_layers=32,
+        num_attention_heads=32,
+        num_key_value_heads=None,
+        resid_pdrop=0.0,
+        embd_pdrop=0.0,
+        attention_dropout=0.0,
+        hidden_act='silu',
+        max_position_embeddings=4096,
+        original_max_position_embeddings=4096,
+        initializer_range=0.02,
+        rms_norm_eps=1e-5,
+        use_cache=True,
+        tie_word_embeddings=False,
+        rope_theta=10000.0,
+        rope_scaling=None,
+        bos_token_id=1,
+        eos_token_id=32000,
+        pad_token_id=32000,
+        sliding_window=None,
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+
+        if num_key_value_heads is None:
+            num_key_value_heads = num_attention_heads
+
+        self.num_key_value_heads = num_key_value_heads
+        self.resid_pdrop = resid_pdrop
+        self.embd_pdrop = embd_pdrop
+        self.attention_dropout = attention_dropout
+        self.hidden_act = hidden_act
+        self.max_position_embeddings = max_position_embeddings
+        self.original_max_position_embeddings = original_max_position_embeddings
+        self.initializer_range = initializer_range
+        self.rms_norm_eps = rms_norm_eps
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.rope_scaling = rope_scaling
+        self._rope_scaling_validation()
+        self.sliding_window = sliding_window
+
+        super().__init__(
+            bos_token_id=bos_token_id,
+            eos_token_id=eos_token_id,
+            pad_token_id=pad_token_id,
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )
+
+    def _rope_scaling_validation(self):
+        """
+        Validate the `rope_scaling` configuration.
+        """
+        if self.rope_scaling is None:
+            return
+
+        if not isinstance(self.rope_scaling, dict) or len(self.rope_scaling) != 3:
+            raise ValueError(
+                '`rope_scaling` must be a dictionary with three fields, `type`, `short_factor` and `long_factor`, '
+                f'got {self.rope_scaling}'
+            )
+        rope_scaling_type = self.rope_scaling.get('type', None)
+        rope_scaling_short_factor = self.rope_scaling.get('short_factor', None)
+        rope_scaling_long_factor = self.rope_scaling.get('long_factor', None)
+        if rope_scaling_type is None or rope_scaling_type not in ['su', 'yarn']:
+            raise ValueError(f"`rope_scaling`'s type field must be one of ['su', 'yarn'], got {rope_scaling_type}")
+        if not (
+            isinstance(rope_scaling_short_factor, list)
+            and all(isinstance(x, (int, float)) for x in rope_scaling_short_factor)
+        ):
+            raise ValueError(
+                f"`rope_scaling`'s short_factor field must be a list of numbers, got {rope_scaling_short_factor}"
+            )
+        if not len(rope_scaling_short_factor) == self.hidden_size // self.num_attention_heads // 2:
+            raise ValueError(
+                f"`rope_scaling`'s short_factor field must have length {self.hidden_size // self.num_attention_heads // 2}, got {len(rope_scaling_short_factor)}"
+            )
+        if not (
+            isinstance(rope_scaling_long_factor, list)
+            and all(isinstance(x, (int, float)) for x in rope_scaling_long_factor)
+        ):
+            raise ValueError(
+                f"`rope_scaling`'s long_factor field must be a list of numbers, got {rope_scaling_long_factor}"
+            )
+        if not len(rope_scaling_long_factor) == self.hidden_size // self.num_attention_heads // 2:
+            raise ValueError(
+                f"`rope_scaling`'s long_factor field must have length {self.hidden_size // self.num_attention_heads // 2}, got {len(rope_scaling_long_factor)}"
+            )

conversation.py

@@ -0,0 +1,1293 @@
+"""
+Conversation prompt templates.
+
+We kindly request that you import fastchat instead of copying this file if you wish to use it.
+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
+from enum import IntEnum, auto
+from typing import Any, Dict, List, Tuple, Union
+
+
+class SeparatorStyle(IntEnum):
+    """Separator styles."""
+
+    ADD_COLON_SINGLE = auto()
+    ADD_COLON_TWO = auto()
+    ADD_COLON_SPACE_SINGLE = auto()
+    NO_COLON_SINGLE = auto()
+    NO_COLON_TWO = auto()
+    ADD_NEW_LINE_SINGLE = auto()
+    LLAMA2 = auto()
+    CHATGLM = auto()
+    CHATML = auto()
+    CHATINTERN = auto()
+    DOLLY = auto()
+    RWKV = auto()
+    PHOENIX = auto()
+    ROBIN = auto()
+    FALCON_CHAT = auto()
+    CHATGLM3 = auto()
+    INTERNVL_ZH = auto()
+    MPT = auto()
+
+
+@dataclasses.dataclass
+class Conversation:
+    """A class that manages prompt templates and keeps all conversation history."""
+
+    # The name of this template
+    name: str
+    # The template of the system prompt
+    system_template: str = '{system_message}'
+    # The system message
+    system_message: str = ''
+    # The names of two roles
+    roles: Tuple[str] = ('USER', 'ASSISTANT')
+    # All messages. Each item is (role, message).
+    messages: List[List[str]] = ()
+    # The number of few shot examples
+    offset: int = 0
+    # The separator style and configurations
+    sep_style: SeparatorStyle = SeparatorStyle.ADD_COLON_SINGLE
+    sep: str = '\n'
+    sep2: str = None
+    # Stop criteria (the default one is EOS token)
+    stop_str: Union[str, List[str]] = None
+    # Stops generation if meeting any token in this list
+    stop_token_ids: List[int] = None
+
+    def get_prompt(self) -> str:
+        """Get the prompt for generation."""
+        system_prompt = self.system_template.format(system_message=self.system_message)
+        if self.sep_style == SeparatorStyle.ADD_COLON_SINGLE:
+            ret = system_prompt + self.sep
+            for role, message in self.messages:
+                if message:
+                    ret += role + ': ' + message + self.sep
+                else:
+                    ret += role + ':'
+            return ret
+        elif self.sep_style == SeparatorStyle.ADD_COLON_TWO:
+            seps = [self.sep, self.sep2]
+            ret = system_prompt + seps[0]
+            for i, (role, message) in enumerate(self.messages):
+                if message:
+                    ret += role + ': ' + message + seps[i % 2]
+                else:
+                    ret += role + ':'
+            return ret
+        elif self.sep_style == SeparatorStyle.ADD_COLON_SPACE_SINGLE:
+            ret = system_prompt + self.sep
+            for role, message in self.messages:
+                if message:
+                    ret += role + ': ' + message + self.sep
+                else:
+                    ret += role + ': '  # must be end with a space
+            return ret
+        elif self.sep_style == SeparatorStyle.ADD_NEW_LINE_SINGLE:
+            ret = '' if system_prompt == '' else system_prompt + self.sep
+            for role, message in self.messages:
+                if message:
+                    ret += role + '\n' + message + self.sep
+                else:
+                    ret += role + '\n'
+            return ret
+        elif self.sep_style == SeparatorStyle.NO_COLON_SINGLE:
+            ret = system_prompt
+            for role, message in self.messages:
+                if message:
+                    ret += role + message + self.sep
+                else:
+                    ret += role
+            return ret
+        elif self.sep_style == SeparatorStyle.NO_COLON_TWO:
+            seps = [self.sep, self.sep2]
+            ret = system_prompt
+            for i, (role, message) in enumerate(self.messages):
+                if message:
+                    ret += role + message + seps[i % 2]
+                else:
+                    ret += role
+            return ret
+        elif self.sep_style == SeparatorStyle.RWKV:
+            ret = system_prompt
+            for i, (role, message) in enumerate(self.messages):
+                if message:
+                    ret += (
+                        role
+                        + ': '
+                        + message.replace('\r\n', '\n').replace('\n\n', '\n')
+                    )
+                    ret += '\n\n'
+                else:
+                    ret += role + ':'
+            return ret
+        elif self.sep_style == SeparatorStyle.LLAMA2:
+            seps = [self.sep, self.sep2]
+            if self.system_message:
+                ret = system_prompt
+            else:
+                ret = '[INST] '
+            for i, (role, message) in enumerate(self.messages):
+                tag = self.roles[i % 2]
+                if message:
+                    if i == 0:
+                        ret += message + ' '
+                    else:
+                        ret += tag + ' ' + message + seps[i % 2]
+                else:
+                    ret += tag
+            return ret
+        elif self.sep_style == SeparatorStyle.CHATGLM:
+            # source: https://huggingface.co/THUDM/chatglm-6b/blob/1d240ba371910e9282298d4592532d7f0f3e9f3e/modeling_chatglm.py#L1302-L1308
+            # source2: https://huggingface.co/THUDM/chatglm2-6b/blob/e186c891cf64310ac66ef10a87e6635fa6c2a579/modeling_chatglm.py#L926
+            round_add_n = 1 if self.name == 'chatglm2' else 0
+            if system_prompt:
+                ret = system_prompt + self.sep
+            else:
+                ret = ''
+
+            for i, (role, message) in enumerate(self.messages):
+                if i % 2 == 0:
+                    ret += f'[Round {i//2 + round_add_n}]{self.sep}'
+
+                if message:
+                    ret += f'{role}：{message}{self.sep}'
+                else:
+                    ret += f'{role}：'
+            return ret
+        elif self.sep_style == SeparatorStyle.CHATML:
+            ret = '' if system_prompt == '' else system_prompt + self.sep + '\n'
+            for role, message in self.messages:
+                if message:
+                    ret += role + '\n' + message + self.sep + '\n'
+                else:
+                    ret += role + '\n'
+            return ret
+        elif self.sep_style == SeparatorStyle.CHATGLM3:
+            ret = ''
+            if self.system_message:
+                ret += system_prompt
+            for role, message in self.messages:
+                if message:
+                    ret += role + '\n' + ' ' + message
+                else:
+                    ret += role
+            return ret
+        elif self.sep_style == SeparatorStyle.CHATINTERN:
+            # source: https://huggingface.co/internlm/internlm-chat-7b-8k/blob/bd546fa984b4b0b86958f56bf37f94aa75ab8831/modeling_internlm.py#L771
+            seps = [self.sep, self.sep2]
+            ret = system_prompt
+            for i, (role, message) in enumerate(self.messages):
+                # if i % 2 == 0:
+                #     ret += "<s>"
+                if message:
+                    ret += role + ':' + message + seps[i % 2] + '\n'
+                else:
+                    ret += role + ':'
+            return ret
+        elif self.sep_style == SeparatorStyle.DOLLY:
+            seps = [self.sep, self.sep2]
+            ret = system_prompt
+            for i, (role, message) in enumerate(self.messages):
+                if message:
+                    ret += role + ':\n' + message + seps[i % 2]
+                    if i % 2 == 1:
+                        ret += '\n\n'
+                else:
+                    ret += role + ':\n'
+            return ret
+        elif self.sep_style == SeparatorStyle.PHOENIX:
+            ret = system_prompt
+            for role, message in self.messages:
+                if message:
+                    ret += role + ': ' + '<s>' + message + '</s>'
+                else:
+                    ret += role + ': ' + '<s>'
+            return ret
+        elif self.sep_style == SeparatorStyle.ROBIN:
+            ret = system_prompt + self.sep
+            for role, message in self.messages:
+                if message:
+                    ret += role + ':\n' + message + self.sep
+                else:
+                    ret += role + ':\n'
+            return ret
+        elif self.sep_style == SeparatorStyle.FALCON_CHAT:
+            ret = ''
+            if self.system_message:
+                ret += system_prompt + self.sep
+            for role, message in self.messages:
+                if message:
+                    ret += role + ': ' + message + self.sep
+                else:
+                    ret += role + ':'
+
+            return ret
+        elif self.sep_style == SeparatorStyle.INTERNVL_ZH:
+            seps = [self.sep, self.sep2]
+            ret = self.system_message + seps[0]
+            for i, (role, message) in enumerate(self.messages):
+                if message:
+                    ret += role + ': ' + message + seps[i % 2]
+                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}')
+
+    def set_system_message(self, system_message: str):
+        """Set the system message."""
+        self.system_message = system_message
+
+    def append_message(self, role: str, message: str):
+        """Append a new message."""
+        self.messages.append([role, message])
+
+    def update_last_message(self, message: str):
+        """Update the last output.
+
+        The last message is typically set to be None when constructing the prompt,
+        so we need to update it in-place after getting the response from a model.
+        """
+        self.messages[-1][1] = message
+
+    def to_gradio_chatbot(self):
+        """Convert the conversation to gradio chatbot format."""
+        ret = []
+        for i, (role, msg) in enumerate(self.messages[self.offset :]):
+            if i % 2 == 0:
+                ret.append([msg, None])
+            else:
+                ret[-1][-1] = msg
+        return ret
+
+    def to_openai_api_messages(self):
+        """Convert the conversation to OpenAI chat completion format."""
+        ret = [{'role': 'system', 'content': self.system_message}]
+
+        for i, (_, msg) in enumerate(self.messages[self.offset :]):
+            if i % 2 == 0:
+                ret.append({'role': 'user', 'content': msg})
+            else:
+                if msg is not None:
+                    ret.append({'role': 'assistant', 'content': msg})
+        return ret
+
+    def copy(self):
+        return Conversation(
+            name=self.name,
+            system_template=self.system_template,
+            system_message=self.system_message,
+            roles=self.roles,
+            messages=[[x, y] for x, y in self.messages],
+            offset=self.offset,
+            sep_style=self.sep_style,
+            sep=self.sep,
+            sep2=self.sep2,
+            stop_str=self.stop_str,
+            stop_token_ids=self.stop_token_ids,
+        )
+
+    def dict(self):
+        return {
+            'template_name': self.name,
+            'system_message': self.system_message,
+            'roles': self.roles,
+            'messages': self.messages,
+            'offset': self.offset,
+        }
+
+
+# A global registry for all conversation templates
+conv_templates: Dict[str, Conversation] = {}
+
+
+def register_conv_template(template: Conversation, override: bool = False):
+    """Register a new conversation template."""
+    if not override:
+        assert (
+            template.name not in conv_templates
+        ), f'{template.name} has been registered.'
+
+    conv_templates[template.name] = template
+
+
+def get_conv_template(name: str) -> Conversation:
+    """Get a conversation template."""
+    return conv_templates[name].copy()
+
+
+# An empty template for raw conversation.
+register_conv_template(
+    Conversation(
+        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],
+    )
+)
+
+
+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,
+        sep='<|im_end|>',
+        stop_token_ids=[
+            2,
+            6,
+            7,
+            8,
+        ],  # "<|endoftext|>", "<|im_start|>", "<|im_end|>", "<|im_sep|>"
+        stop_str='<|endoftext|>',
+    )
+)
+
+
+register_conv_template(
+    Conversation(
+        name='internlm2-chat',
+        system_template='<|im_start|>system\n{system_message}',
+        system_message='You are an AI assistant whose name is InternLM (书生·浦语).',
+        roles=('<|im_start|>user\n', '<|im_start|>assistant\n'),
+        sep_style=SeparatorStyle.MPT,
+        sep='<|im_end|>',
+        stop_token_ids=[
+            2,
+            92543,
+            92542
+        ]
+    )
+)
+
+
+register_conv_template(
+    Conversation(
+        name='llama3-chat',
+        system_template='<|system|>\n{system_message}',
+        system_message='You are an AI assistant whose name is InternVL.',
+        roles=('<|user|>\n', '<|assistant|>\n'),
+        sep_style=SeparatorStyle.MPT,
+        sep='<|end|>',
+        stop_token_ids=[
+            128259,
+            128001
+        ]
+    )
+)
+
+
+register_conv_template(
+    Conversation(
+        name='phi3-chat',
+        system_template='<|system|>\n{system_message}',
+        system_message='You are an AI assistant whose name is Phi-3.',
+        roles=('<|user|>\n', '<|assistant|>\n'),
+        sep_style=SeparatorStyle.MPT,
+        sep='<|end|>',
+        stop_token_ids=[
+            2,
+            32000,
+            32007
+        ]
+    )
+)
+
+# 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>',
+    )
+)

generation_config.json

@@ -0,0 +1,10 @@
+{
+  "do_sample": true,
+  "eos_token_id": 2,
+  "max_new_tokens": 2048,
+  "pad_token_id": 2,
+  "temperature": 0.3,
+  "top_k": 20,
+  "top_p": 0.7,
+  "transformers_version": "4.42.3"
+}

modeling_intern_vit.py

@@ -0,0 +1,434 @@
+# --------------------------------------------------------
+# InternVL
+# Copyright (c) 2023 OpenGVLab
+# Licensed under The MIT License [see LICENSE for details]
+# --------------------------------------------------------
+from typing import Optional, Tuple, Union
+
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+from einops import rearrange
+from timm.models.layers import DropPath
+from torch import nn
+from transformers.activations import ACT2FN
+from transformers.modeling_outputs import (BaseModelOutput,
+                                           BaseModelOutputWithPooling)
+from transformers.modeling_utils import PreTrainedModel
+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
+
+    has_flash_attn = True
+except:
+    print('FlashAttention is not installed.')
+    has_flash_attn = False
+
+logger = logging.get_logger(__name__)
+
+
+class FlashAttention(nn.Module):
+    """Implement the scaled dot product attention with softmax.
+    Arguments
+    ---------
+        softmax_scale: The temperature to use for the softmax attention.
+                      (default: 1/sqrt(d_keys) where d_keys is computed at
+                      runtime)
+        attention_dropout: The dropout rate to apply to the attention
+                           (default: 0.0)
+    """
+
+    def __init__(self, softmax_scale=None, attention_dropout=0.0, device=None, dtype=None):
+        super().__init__()
+        self.softmax_scale = softmax_scale
+        self.dropout_p = attention_dropout
+
+    def forward(self, qkv, key_padding_mask=None, causal=False, cu_seqlens=None,
+                max_s=None, need_weights=False):
+        """Implements the multihead softmax attention.
+        Arguments
+        ---------
+            qkv: The tensor containing the query, key, and value. (B, S, 3, H, D) if key_padding_mask is None
+                if unpadded: (nnz, 3, h, d)
+            key_padding_mask: a bool tensor of shape (B, S)
+        """
+        assert not need_weights
+        assert qkv.dtype in [torch.float16, torch.bfloat16]
+        assert qkv.is_cuda
+
+        if cu_seqlens is None:
+            batch_size = qkv.shape[0]
+            seqlen = qkv.shape[1]
+            if key_padding_mask is None:
+                qkv = rearrange(qkv, 'b s ... -> (b s) ...')
+                max_s = seqlen
+                cu_seqlens = torch.arange(0, (batch_size + 1) * seqlen, step=seqlen, dtype=torch.int32,
+                                          device=qkv.device)
+                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
+                )
+                output = rearrange(output, '(b s) ... -> b s ...', b=batch_size)
+            else:
+                nheads = qkv.shape[-2]
+                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_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
+                )
+                output = rearrange(pad_input(rearrange(output_unpad, 'nnz h d -> nnz (h d)'),
+                                             indices, batch_size, seqlen),
+                                   'b s (h d) -> b s h d', h=nheads)
+        else:
+            assert max_s is not None
+            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
+            )
+
+        return output, None
+
+
+class InternRMSNorm(nn.Module):
+    def __init__(self, hidden_size, eps=1e-6):
+        super().__init__()
+        self.weight = nn.Parameter(torch.ones(hidden_size))
+        self.variance_epsilon = eps
+
+    def forward(self, hidden_states):
+        input_dtype = hidden_states.dtype
+        hidden_states = hidden_states.to(torch.float32)
+        variance = hidden_states.pow(2).mean(-1, keepdim=True)
+        hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
+        return self.weight * hidden_states.to(input_dtype)
+
+
+try:
+    from apex.normalization import FusedRMSNorm
+
+    InternRMSNorm = FusedRMSNorm  # noqa
+
+    logger.info('Discovered apex.normalization.FusedRMSNorm - will use it instead of InternRMSNorm')
+except ImportError:
+    # using the normal InternRMSNorm
+    pass
+except Exception:
+    logger.warning('discovered apex but it failed to load, falling back to InternRMSNorm')
+    pass
+
+
+NORM2FN = {
+    'rms_norm': InternRMSNorm,
+    'layer_norm': nn.LayerNorm,
+}
+
+
+class InternVisionEmbeddings(nn.Module):
+    def __init__(self, config: InternVisionConfig):
+        super().__init__()
+        self.config = config
+        self.embed_dim = config.hidden_size
+        self.image_size = config.image_size
+        self.patch_size = config.patch_size
+
+        self.class_embedding = nn.Parameter(
+            torch.randn(1, 1, self.embed_dim),
+        )
+
+        self.patch_embedding = nn.Conv2d(
+            in_channels=3, out_channels=self.embed_dim, kernel_size=self.patch_size, stride=self.patch_size
+        )
+
+        self.num_patches = (self.image_size // self.patch_size) ** 2
+        self.num_positions = self.num_patches + 1
+
+        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:
+        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 = 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)
+        return embeddings
+
+
+class InternAttention(nn.Module):
+    """Multi-headed attention from 'Attention Is All You Need' paper"""
+
+    def __init__(self, config: InternVisionConfig):
+        super().__init__()
+        self.config = config
+        self.embed_dim = config.hidden_size
+        self.num_heads = config.num_attention_heads
+        self.use_flash_attn = config.use_flash_attn and has_flash_attn
+        if config.use_flash_attn and not has_flash_attn:
+            print('Warning: Flash Attention is not available, use_flash_attn is set to False.')
+        self.head_dim = self.embed_dim // self.num_heads
+        if self.head_dim * self.num_heads != self.embed_dim:
+            raise ValueError(
+                f'embed_dim must be divisible by num_heads (got `embed_dim`: {self.embed_dim} and `num_heads`:'
+                f' {self.num_heads}).'
+            )
+
+        self.scale = self.head_dim ** -0.5
+        self.qkv = nn.Linear(self.embed_dim, 3 * self.embed_dim, bias=config.qkv_bias)
+        self.attn_drop = nn.Dropout(config.attention_dropout)
+        self.proj_drop = nn.Dropout(config.dropout)
+
+        self.qk_normalization = config.qk_normalization
+
+        if self.qk_normalization:
+            self.q_norm = InternRMSNorm(self.embed_dim, eps=config.layer_norm_eps)
+            self.k_norm = InternRMSNorm(self.embed_dim, eps=config.layer_norm_eps)
+
+        if self.use_flash_attn:
+            self.inner_attn = FlashAttention(attention_dropout=config.attention_dropout)
+        self.proj = nn.Linear(self.embed_dim, self.embed_dim)
+
+    def _naive_attn(self, x):
+        B, N, C = x.shape
+        qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4)
+        q, k, v = qkv.unbind(0)  # make torchscript happy (cannot use tensor as tuple)
+
+        if self.qk_normalization:
+            B_, H_, N_, D_ = q.shape
+            q = self.q_norm(q.transpose(1, 2).flatten(-2, -1)).view(B_, N_, H_, D_).transpose(1, 2)
+            k = self.k_norm(k.transpose(1, 2).flatten(-2, -1)).view(B_, N_, H_, D_).transpose(1, 2)
+
+        attn = ((q * self.scale) @ k.transpose(-2, -1))
+        attn = attn.softmax(dim=-1)
+        attn = self.attn_drop(attn)
+
+        x = (attn @ v).transpose(1, 2).reshape(B, N, C)
+        x = self.proj(x)
+        x = self.proj_drop(x)
+        return x
+
+    def _flash_attn(self, x, key_padding_mask=None, need_weights=False):
+        qkv = self.qkv(x)
+        qkv = rearrange(qkv, 'b s (three h d) -> b s three h d', three=3, h=self.num_heads)
+
+        if self.qk_normalization:
+            q, k, v = qkv.unbind(2)
+            q = self.q_norm(q.flatten(-2, -1)).view(q.shape)
+            k = self.k_norm(k.flatten(-2, -1)).view(k.shape)
+            qkv = torch.stack([q, k, v], dim=2)
+
+        context, _ = self.inner_attn(
+            qkv, key_padding_mask=key_padding_mask, need_weights=need_weights, causal=False
+        )
+        outs = self.proj(rearrange(context, 'b s h d -> b s (h d)'))
+        outs = self.proj_drop(outs)
+        return outs
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        x = self._naive_attn(hidden_states) if not self.use_flash_attn else self._flash_attn(hidden_states)
+        return x
+
+
+class InternMLP(nn.Module):
+    def __init__(self, config: InternVisionConfig):
+        super().__init__()
+        self.config = config
+        self.act = ACT2FN[config.hidden_act]
+        self.fc1 = nn.Linear(config.hidden_size, config.intermediate_size)
+        self.fc2 = nn.Linear(config.intermediate_size, config.hidden_size)
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        hidden_states = self.fc1(hidden_states)
+        hidden_states = self.act(hidden_states)
+        hidden_states = self.fc2(hidden_states)
+        return hidden_states
+
+
+class InternVisionEncoderLayer(nn.Module):
+    def __init__(self, config: InternVisionConfig, drop_path_rate: float):
+        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.ls1 = nn.Parameter(config.initializer_factor * torch.ones(self.embed_dim))
+        self.ls2 = nn.Parameter(config.initializer_factor * torch.ones(self.embed_dim))
+        self.drop_path1 = DropPath(drop_path_rate) if drop_path_rate > 0. else nn.Identity()
+        self.drop_path2 = DropPath(drop_path_rate) if drop_path_rate > 0. else nn.Identity()
+
+    def forward(
+            self,
+            hidden_states: torch.Tensor,
+    ) -> Tuple[torch.FloatTensor, Optional[torch.FloatTensor], Optional[Tuple[torch.FloatTensor]]]:
+        """
+        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)) * self.ls1)
+
+        hidden_states = hidden_states + self.drop_path2(self.mlp(self.norm2(hidden_states)) * self.ls2)
+
+        return hidden_states
+
+
+class InternVisionEncoder(nn.Module):
+    """
+    Transformer encoder consisting of `config.num_hidden_layers` self attention layers. Each layer is a
+    [`InternEncoderLayer`].
+
+    Args:
+        config (`InternConfig`):
+            The corresponding vision configuration for the `InternEncoder`.
+    """
+
+    def __init__(self, config: InternVisionConfig):
+        super().__init__()
+        self.config = config
+        # stochastic depth decay rule
+        dpr = [x.item() for x in torch.linspace(0, config.drop_path_rate, config.num_hidden_layers)]
+        self.layers = nn.ModuleList([
+            InternVisionEncoderLayer(config, dpr[idx]) for idx in range(config.num_hidden_layers)])
+        self.gradient_checkpointing = True
+
+    def forward(
+            self,
+            inputs_embeds,
+            output_hidden_states: Optional[bool] = None,
+            return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, BaseModelOutput]:
+        r"""
+        Args:
+            inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`):
+                Embedded representation of the inputs. Should be float, not int tokens.
+            output_hidden_states (`bool`, *optional*):
+                Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors
+                for more detail.
+            return_dict (`bool`, *optional*):
+                Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
+        """
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        encoder_states = () if output_hidden_states else None
+        hidden_states = inputs_embeds
+
+        for idx, encoder_layer in enumerate(self.layers):
+            if output_hidden_states:
+                encoder_states = encoder_states + (hidden_states,)
+            if self.gradient_checkpointing and self.training:
+                layer_outputs = torch.utils.checkpoint.checkpoint(
+                    encoder_layer,
+                    hidden_states)
+            else:
+                layer_outputs = encoder_layer(
+                    hidden_states,
+                )
+            hidden_states = layer_outputs
+
+        if output_hidden_states:
+            encoder_states = encoder_states + (hidden_states,)
+
+        if not return_dict:
+            return tuple(v for v in [hidden_states, encoder_states] if v is not None)
+        return BaseModelOutput(
+            last_hidden_state=hidden_states, hidden_states=encoder_states
+        )
+
+
+class InternVisionModel(PreTrainedModel):
+    main_input_name = 'pixel_values'
+    config_class = InternVisionConfig
+    _no_split_modules = ['InternVisionEncoderLayer']
+
+    def __init__(self, config: InternVisionConfig):
+        super().__init__(config)
+        self.config = config
+
+        self.embeddings = InternVisionEmbeddings(config)
+        self.encoder = InternVisionEncoder(config)
+
+    def resize_pos_embeddings(self, old_size, new_size, patch_size):
+        pos_emb = self.embeddings.position_embedding
+        _, num_positions, embed_dim = pos_emb.shape
+        cls_emb = pos_emb[:, :1, :]
+        pos_emb = pos_emb[:, 1:, :].reshape(1, old_size // patch_size, old_size // patch_size, -1).permute(0, 3, 1, 2)
+        pos_emb = F.interpolate(pos_emb.float(), size=new_size // patch_size, mode='bicubic', align_corners=False)
+        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):
+        return self.embeddings
+
+    def forward(
+            self,
+            pixel_values: Optional[torch.FloatTensor] = None,
+            output_hidden_states: Optional[bool] = None,
+            return_dict: Optional[bool] = None,
+            pixel_embeds: Optional[torch.FloatTensor] = None,
+    ) -> Union[Tuple, BaseModelOutputWithPooling]:
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        if pixel_values is None and pixel_embeds is None:
+            raise ValueError('You have to specify pixel_values or pixel_embeds')
+
+        if pixel_embeds is not None:
+            hidden_states = pixel_embeds
+        else:
+            if len(pixel_values.shape) == 4:
+                hidden_states = self.embeddings(pixel_values)
+            else:
+                raise ValueError(f'wrong pixel_values size: {pixel_values.shape}')
+        encoder_outputs = self.encoder(
+            inputs_embeds=hidden_states,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+        last_hidden_state = encoder_outputs.last_hidden_state
+        pooled_output = last_hidden_state[:, 0, :]
+
+        if not return_dict:
+            return (last_hidden_state, pooled_output) + encoder_outputs[1:]
+
+        return BaseModelOutputWithPooling(
+            last_hidden_state=last_hidden_state,
+            pooler_output=pooled_output,
+            hidden_states=encoder_outputs.hidden_states,
+            attentions=encoder_outputs.attentions,
+        )

modeling_internvl_chat.py

@@ -0,0 +1,358 @@
+# --------------------------------------------------------
+# InternVL
+# 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
+from peft import LoraConfig, get_peft_model
+from torch import nn
+from torch.nn import CrossEntropyLoss
+from transformers import (AutoModel, 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 .modeling_intern_vit import InternVisionModel
+from .modeling_phi3 import Phi3ForCausalLM
+
+logger = logging.get_logger(__name__)
+
+
+class InternVLChatModel(PreTrainedModel):
+    config_class = InternVLChatConfig
+    main_input_name = 'pixel_values'
+    _no_split_modules = ['InternVisionEncoderLayer', 'LlamaDecoderLayer', 'Phi3DecoderLayer']
+
+    def __init__(self, config: InternVLChatConfig, vision_model=None, language_model=None):
+        super().__init__(config)
+
+        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
+
+        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:
+            self.vision_model = InternVisionModel(config.vision_config)
+        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)
+            elif config.llm_config.architectures[0] == 'Phi3ForCausalLM':
+                self.language_model = Phi3ForCausalLM(config.llm_config)
+            else:
+                raise NotImplementedError(f'{config.llm_config.architectures[0]} is not implemented.')
+
+        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.GELU(),
+            nn.Linear(llm_hidden_size, llm_hidden_size)
+        )
+
+        # if config.force_image_size != config.vision_config.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.neftune_alpha = None
+
+        if config.use_backbone_lora:
+            self.wrap_backbone_lora(r=config.use_backbone_lora, lora_alpha=2 * config.use_backbone_lora)
+
+        if config.use_llm_lora:
+            self.wrap_llm_lora(r=config.use_llm_lora, lora_alpha=2 * 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.enable_input_require_grads()
+        self.language_model.print_trainable_parameters()
+
+    def forward(
+            self,
+            pixel_values: torch.FloatTensor,
+            input_ids: torch.LongTensor = None,
+            attention_mask: Optional[torch.Tensor] = None,
+            position_ids: Optional[torch.LongTensor] = None,
+            image_flags: Optional[torch.LongTensor] = None,
+            past_key_values: Optional[List[torch.FloatTensor]] = None,
+            labels: Optional[torch.LongTensor] = None,
+            use_cache: Optional[bool] = None,
+            output_attentions: Optional[bool] = None,
+            output_hidden_states: Optional[bool] = None,
+            return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, CausalLMOutputWithPast]:
+        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)
+
+        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.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]
+
+        input_embeds = input_embeds.reshape(B, N, C)
+
+        outputs = self.language_model(
+            inputs_embeds=input_embeds,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+        logits = outputs.logits
+
+        loss = None
+        if labels is not None:
+            # Shift so that tokens < n predict n
+            shift_logits = logits[..., :-1, :].contiguous()
+            shift_labels = labels[..., 1:].contiguous()
+            # Flatten the tokens
+            loss_fct = CrossEntropyLoss()
+            shift_logits = shift_logits.view(-1, self.language_model.config.vocab_size)
+            shift_labels = shift_labels.view(-1)
+            # Enable model parallelism
+            shift_labels = shift_labels.to(shift_logits.device)
+            loss = loss_fct(shift_logits, shift_labels)
+
+        if not return_dict:
+            output = (logits,) + outputs[1:]
+            return (loss,) + output if loss is not None else output
+
+        return CausalLMOutputWithPast(
+            loss=loss,
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+    def pixel_shuffle(self, x, scale_factor=0.5):
+        n, w, h, c = x.size()
+        # N, W, H, C --> N, W, H * scale, C // scale
+        x = x.view(n, w, int(h * scale_factor), int(c / scale_factor))
+        # N, W, H * scale, C // scale --> N, H * scale, W, C // scale
+        x = x.permute(0, 2, 1, 3).contiguous()
+        # 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 noised_embed(self, vit_embeds, noise_alpha=5):
+        dims = torch.tensor(vit_embeds.size(1) * vit_embeds.size(2))
+        mag_norm = noise_alpha / torch.sqrt(dims)
+        noise = torch.zeros_like(vit_embeds).uniform_(-mag_norm, mag_norm)
+        return vit_embeds + noise
+
+    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 = vit_embeds[:, 1:, :]
+
+        if self.training and self.neftune_alpha is not None:
+            vit_embeds = self.noised_embed(vit_embeds, self.neftune_alpha)
+
+        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 = vit_embeds.reshape(vit_embeds.shape[0], -1, vit_embeds.shape[-1])
+        vit_embeds = self.mlp1(vit_embeds)
+        return vit_embeds
+
+    def batch_chat(self, tokenizer, pixel_values, image_counts, questions, generation_config, history=None,
+                         return_history=False, IMG_START_TOKEN='<img>', IMG_END_TOKEN='</img>',
+                         IMG_CONTEXT_TOKEN='<IMG_CONTEXT>'):
+        if history is not None or return_history:
+            print('Now multi-turn chat is not supported in batch_chat.')
+            raise NotImplementedError
+        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
+
+        queries = []
+        image_bs = pixel_values.shape[0]
+        # print(f'dynamic ViT batch size: {image_bs}, image_counts: {image_counts}')
+        for idx, image_count in enumerate(image_counts):
+            image_token = IMG_START_TOKEN + IMG_CONTEXT_TOKEN * self.num_image_token * image_count + IMG_END_TOKEN
+            question = image_token + '\n' + questions[idx]
+            template = get_conv_template(self.template)
+            template.append_message(template.roles[0], question)
+            template.append_message(template.roles[1], None)
+            query = template.get_prompt()
+            queries.append(query)
+        tokenizer.padding_side = 'left'
+        model_inputs = tokenizer(queries, return_tensors='pt', padding=True)
+        input_ids = model_inputs['input_ids'].cuda()
+        attention_mask = model_inputs['attention_mask'].cuda()
+        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,
+             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)
+        image_bs = pixel_values.shape[0]
+        print(f'dynamic ViT batch size: {image_bs}')
+        if history is None:
+            history = []
+            image_tokens = IMG_START_TOKEN + IMG_CONTEXT_TOKEN * self.num_image_token * image_bs + IMG_END_TOKEN
+            question = image_tokens + '\n' + question
+        else:
+            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)
+        template.append_message(template.roles[1], None)
+        query = template.get_prompt()
+        model_inputs = tokenizer(query, return_tensors='pt')
+        input_ids = model_inputs['input_ids'].cuda()
+        attention_mask = model_inputs['attention_mask'].cuda()
+        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
+        )
+        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(image_tokens, '<image>')
+            # print(query_to_print, response)
+            return response
+        return response
+
+    @torch.no_grad()
+    def generate(
+            self,
+            pixel_values: Optional[torch.FloatTensor] = None,
+            input_ids: Optional[torch.FloatTensor] = None,
+            attention_mask: Optional[torch.LongTensor] = None,
+            visual_features: Optional[torch.FloatTensor] = None,
+            generation_config: Optional[GenerationConfig] = None,
+            output_hidden_states: Optional[bool] = None,
+            return_dict: Optional[bool] = None,
+            **generate_kwargs,
+    ) -> torch.LongTensor:
+
+        assert self.img_context_token_id is not None
+        if pixel_values is not None:
+            if visual_features is not None:
+                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)
+
+            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 = input_embeds.reshape(B, N, C)
+        else:
+            input_embeds = self.language_model.get_input_embeddings()(input_ids)
+
+        outputs = self.language_model.generate(
+            inputs_embeds=input_embeds,
+            attention_mask=attention_mask,
+            generation_config=generation_config,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+            use_cache=True,
+            **generate_kwargs,
+        )
+
+        return outputs

modeling_phi3.py

@@ -0,0 +1,1601 @@
+# Copyright 2024 Microsoft and the HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+""" PyTorch Phi-3 model."""
+
+import inspect
+import math
+import warnings
+from typing import List, Optional, Tuple, Union
+
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+from torch import nn
+from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
+from transformers.activations import ACT2FN
+from transformers.cache_utils import Cache, DynamicCache
+from transformers.modeling_attn_mask_utils import \
+    _prepare_4d_causal_attention_mask
+from transformers.modeling_outputs import (BaseModelOutputWithPast,
+                                           CausalLMOutputWithPast,
+                                           SequenceClassifierOutputWithPast,
+                                           TokenClassifierOutput)
+from transformers.modeling_utils import PreTrainedModel
+from transformers.utils import (add_code_sample_docstrings,
+                                add_start_docstrings,
+                                add_start_docstrings_to_model_forward,
+                                is_flash_attn_2_available,
+                                is_flash_attn_greater_or_equal_2_10, logging,
+                                replace_return_docstrings)
+
+from .configuration_phi3 import Phi3Config
+
+logger = logging.get_logger(__name__)
+
+# Transformers scans dependencies in the modeling file, causing issues on conditional loading. The regex only ignores try/catch blocks, but not if statements
+# if is_flash_attn_2_available():
+_flash_supports_window_size = False
+try:
+    from flash_attn import flash_attn_func, flash_attn_varlen_func
+    from flash_attn.bert_padding import (index_first_axis, pad_input,  # noqa
+                                         unpad_input)
+
+    _flash_supports_window_size = 'window_size' in list(inspect.signature(flash_attn_func).parameters)
+except ImportError as error:
+    logger.warning(
+        f'`flash-attention` package not found, consider installing for better performance: {error}.'
+    )
+    if not _flash_supports_window_size:
+        logger.warning(
+            "Current `flash-attenton` does not support `window_size`. Either upgrade or use `attn_implementation='eager'`."
+        )
+
+_CHECKPOINT_FOR_DOC = 'microsoft/Phi-3-mini-4k-instruct'
+_CONFIG_FOR_DOC = 'Phi3Config'
+
+PHI3_PRETRAINED_MODEL_ARCHIVE_LIST = [
+    'microsoft/Phi-3-mini-4k-instruct',
+    'microsoft/Phi-3-mini-128k-instruct',
+    # See all Phi-3 models at https://huggingface.co/models?filter=Phi-3
+]
+
+
+# Copied from transformers.models.llama.modeling_llama.LlamaRMSNorm with Llama->Phi3
+class Phi3RMSNorm(nn.Module):
+    def __init__(self, hidden_size, eps=1e-6):
+        """
+        Phi3RMSNorm is equivalent to T5LayerNorm
+        """
+        super().__init__()
+        self.weight = nn.Parameter(torch.ones(hidden_size))
+        self.variance_epsilon = eps
+
+    def forward(self, hidden_states):
+        input_dtype = hidden_states.dtype
+        hidden_states = hidden_states.to(torch.float32)
+        variance = hidden_states.pow(2).mean(-1, keepdim=True)
+        hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
+        return self.weight * hidden_states.to(input_dtype)
+
+
+# Copied from transformers.models.llama.modeling_llama._get_unpad_data
+def _get_unpad_data(attention_mask):
+    seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
+    indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
+    max_seqlen_in_batch = seqlens_in_batch.max().item()
+    cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
+    return (
+        indices,
+        cu_seqlens,
+        max_seqlen_in_batch,
+    )
+
+
+# Copied from transformers.models.gemma.modeling_gemma.GemmaRotaryEmbedding with gemma->phi3, Gemma->Phi3
+class Phi3RotaryEmbedding(nn.Module):
+    def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None):
+        super().__init__()
+
+        self.dim = dim
+        self.max_position_embeddings = max_position_embeddings
+        self.base = base
+        self.register_buffer('inv_freq', None, persistent=False)
+
+    @torch.no_grad()
+    def forward(self, x, position_ids, seq_len=None):
+        # x: [bs, num_attention_heads, seq_len, head_size]
+        if self.inv_freq is None:
+            self.inv_freq = 1.0 / (
+                self.base ** (torch.arange(0, self.dim, 2, dtype=torch.int64, device=x.device).float() / self.dim)
+            )
+        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1)
+        position_ids_expanded = position_ids[:, None, :].float()
+        # Force float32 since bfloat16 loses precision on long contexts
+        # See https://github.com/huggingface/transformers/pull/29285
+        device_type = x.device.type
+        device_type = device_type if isinstance(device_type, str) and device_type != 'mps' else 'cpu'
+        with torch.autocast(device_type=device_type, enabled=False):
+            freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
+            emb = torch.cat((freqs, freqs), dim=-1)
+            cos = emb.cos()
+            sin = emb.sin()
+        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
+
+
+class Phi3SuScaledRotaryEmbedding(Phi3RotaryEmbedding):
+    def __init__(self, dim, config, device=None):
+        super().__init__(dim, config.max_position_embeddings, config.rope_theta, device)
+
+        self.short_factor = config.rope_scaling['short_factor']
+        self.long_factor = config.rope_scaling['long_factor']
+        self.original_max_position_embeddings = config.original_max_position_embeddings
+
+    @torch.no_grad()
+    def forward(self, x, position_ids, seq_len=None):
+        seq_len = torch.max(position_ids) + 1
+        if seq_len > self.original_max_position_embeddings:
+            ext_factors = torch.tensor(self.long_factor, dtype=torch.float32, device=x.device)
+        else:
+            ext_factors = torch.tensor(self.short_factor, dtype=torch.float32, device=x.device)
+
+        inv_freq_shape = torch.arange(0, self.dim, 2, dtype=torch.int64, device=x.device).float() / self.dim
+        self.inv_freq = 1.0 / (ext_factors * self.base**inv_freq_shape)
+
+        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1)
+        position_ids_expanded = position_ids[:, None, :].float()
+
+        # Force float32 since bfloat16 loses precision on long contexts
+        # See https://github.com/huggingface/transformers/pull/29285
+        device_type = x.device.type
+        device_type = device_type if isinstance(device_type, str) and device_type != 'mps' else 'cpu'
+        with torch.autocast(device_type=device_type, enabled=False):
+            freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
+            emb = torch.cat((freqs, freqs), dim=-1)
+
+            scale = self.max_position_embeddings / self.original_max_position_embeddings
+            if scale <= 1.0:
+                scaling_factor = 1.0
+            else:
+                scaling_factor = math.sqrt(1 + math.log(scale) / math.log(self.original_max_position_embeddings))
+
+            cos = emb.cos() * scaling_factor
+            sin = emb.sin() * scaling_factor
+        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
+
+
+class Phi3YarnScaledRotaryEmbedding(Phi3RotaryEmbedding):
+    def __init__(self, dim, config, device=None):
+        super().__init__(dim, config.max_position_embeddings, config.rope_theta, device)
+
+        self.short_factor = config.rope_scaling['short_factor']
+        self.long_factor = config.rope_scaling['long_factor']
+        self.original_max_position_embeddings = config.original_max_position_embeddings
+
+    @torch.no_grad()
+    def forward(self, x, position_ids, seq_len=None):
+        seq_len = torch.max(position_ids) + 1
+        if seq_len > self.original_max_position_embeddings:
+            ext_factors = torch.tensor(self.long_factor, dtype=torch.float32, device=x.device)
+        else:
+            ext_factors = torch.tensor(self.short_factor, dtype=torch.float32, device=x.device)
+
+        inv_freq_shape = torch.arange(0, self.dim, 2, dtype=torch.int64, device=x.device).float() / self.dim
+        self.inv_freq = 1.0 / (ext_factors * self.base**inv_freq_shape)
+
+        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1)
+        position_ids_expanded = position_ids[:, None, :].float()
+
+        # Force float32 since bfloat16 loses precision on long contexts
+        # See https://github.com/huggingface/transformers/pull/29285
+        device_type = x.device.type
+        device_type = device_type if isinstance(device_type, str) and device_type != 'mps' else 'cpu'
+        with torch.autocast(device_type=device_type, enabled=False):
+            freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
+            emb = torch.cat((freqs, freqs), dim=-1)
+
+            scale = self.max_position_embeddings / self.original_max_position_embeddings
+            if scale <= 1.0:
+                scaling_factor = 1.0
+            else:
+                scaling_factor = 0.1 * math.log(scale) + 1.0
+
+            cos = emb.cos() * scaling_factor
+            sin = emb.sin() * scaling_factor
+        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
+
+
+# Copied from transformers.models.llama.modeling_llama.rotate_half
+def rotate_half(x):
+    """Rotates half the hidden dims of the input."""
+    x1 = x[..., : x.shape[-1] // 2]
+    x2 = x[..., x.shape[-1] // 2 :]
+    return torch.cat((-x2, x1), dim=-1)
+
+
+# Copied from transformers.models.llama.modeling_llama.apply_rotary_pos_emb
+def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_dim=1):
+    """Applies Rotary Position Embedding to the query and key tensors.
+
+    Args:
+        q (`torch.Tensor`): The query tensor.
+        k (`torch.Tensor`): The key tensor.
+        cos (`torch.Tensor`): The cosine part of the rotary embedding.
+        sin (`torch.Tensor`): The sine part of the rotary embedding.
+        position_ids (`torch.Tensor`, *optional*):
+            Deprecated and unused.
+        unsqueeze_dim (`int`, *optional*, defaults to 1):
+            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
+            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
+            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
+            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
+            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
+            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
+    Returns:
+        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
+    """
+    cos = cos.unsqueeze(unsqueeze_dim)
+    sin = sin.unsqueeze(unsqueeze_dim)
+    q_embed = (q * cos) + (rotate_half(q) * sin)
+    k_embed = (k * cos) + (rotate_half(k) * sin)
+    return q_embed, k_embed
+
+
+class Phi3MLP(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+
+        self.config = config
+        self.gate_up_proj = nn.Linear(config.hidden_size, 2 * config.intermediate_size, bias=False)
+        self.down_proj = nn.Linear(config.intermediate_size, config.hidden_size, bias=False)
+
+        self.activation_fn = ACT2FN[config.hidden_act]
+
+    def forward(self, hidden_states: torch.FloatTensor) -> torch.FloatTensor:
+        up_states = self.gate_up_proj(hidden_states)
+
+        gate, up_states = up_states.chunk(2, dim=-1)
+        up_states = up_states * self.activation_fn(gate)
+
+        return self.down_proj(up_states)
+
+
+# Copied from transformers.models.llama.modeling_llama.repeat_kv with llama->phi
+def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
+    """
+    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
+    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
+    """
+    batch, num_key_value_heads, slen, head_dim = hidden_states.shape
+    if n_rep == 1:
+        return hidden_states
+    hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)
+    return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)
+
+
+class Phi3Attention(nn.Module):
+    """Multi-headed attention from 'Attention Is All You Need' paper"""
+
+    def __init__(self, config: Phi3Config, layer_idx: Optional[int] = None):
+        super().__init__()
+        self.config = config
+        self.layer_idx = layer_idx
+        if layer_idx is None:
+            logger.warning_once(
+                f'Instantiating {self.__class__.__name__} without passing a `layer_idx` is not recommended and will '
+                'lead to errors during the forward call if caching is used. Please make sure to provide a `layer_idx` '
+                'when creating this class.'
+            )
+
+        self.attention_dropout = config.attention_dropout
+        self.hidden_size = config.hidden_size
+        self.num_heads = config.num_attention_heads
+        self.head_dim = self.hidden_size // self.num_heads
+        self.num_key_value_heads = config.num_key_value_heads
+        self.num_key_value_groups = self.num_heads // self.num_key_value_heads
+        self.max_position_embeddings = config.max_position_embeddings
+        self.original_max_position_embeddings = config.original_max_position_embeddings
+        self.rope_theta = config.rope_theta
+        self.rope_scaling = config.rope_scaling
+        self.is_causal = True
+
+        if (self.head_dim * self.num_heads) != self.hidden_size:
+            raise ValueError(
+                f'hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}'
+                f' and `num_heads`: {self.num_heads}).'
+            )
+
+        op_size = self.num_heads * self.head_dim + 2 * (self.num_key_value_heads * self.head_dim)
+        self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=False)
+        self.qkv_proj = nn.Linear(self.hidden_size, op_size, bias=False)
+        self._init_rope()
+
+    def _init_rope(self):
+        if self.rope_scaling is None:
+            self.rotary_emb = Phi3RotaryEmbedding(
+                self.head_dim,
+                max_position_embeddings=self.max_position_embeddings,
+                base=self.rope_theta,
+            )
+        else:
+            scaling_type = self.config.rope_scaling['type']
+            if scaling_type == 'su':
+                self.rotary_emb = Phi3SuScaledRotaryEmbedding(self.head_dim, self.config)
+            elif scaling_type == 'yarn':
+                self.rotary_emb = Phi3YarnScaledRotaryEmbedding(self.head_dim, self.config)
+            else:
+                raise ValueError(f'Unknown RoPE scaling type {scaling_type}')
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        logger.warning_once('You are not running the flash-attention implementation, expect numerical differences.')
+
+        bsz, q_len, _ = hidden_states.size()
+
+        qkv = self.qkv_proj(hidden_states)
+        query_pos = self.num_heads * self.head_dim
+        query_states = qkv[..., :query_pos]
+        key_states = qkv[..., query_pos : query_pos + self.num_key_value_heads * self.head_dim]
+        value_states = qkv[..., query_pos + self.num_key_value_heads * self.head_dim :]
+
+        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+
+        kv_seq_len = key_states.shape[-2]
+        if past_key_value is not None:
+            if self.layer_idx is None:
+                raise ValueError(
+                    f'The cache structure has changed since version v4.36. If you are using {self.__class__.__name__} '
+                    'for auto-regressive decoding with k/v caching, please make sure to initialize the attention class '
+                    'with a layer index.'
+                )
+            kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
+        cos, sin = self.rotary_emb(value_states, position_ids, seq_len=kv_seq_len)
+
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
+
+        if past_key_value is not None:
+            cache_kwargs = {'sin': sin, 'cos': cos}  # Specific to RoPE models
+            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+        # repeat k/v heads if n_kv_heads < n_heads
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+        attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
+
+        if attn_weights.size() != (bsz, self.num_heads, q_len, kv_seq_len):
+            raise ValueError(
+                f'Attention weights should be of size {(bsz, self.num_heads, q_len, kv_seq_len)}, but is'
+                f' {attn_weights.size()}'
+            )
+
+        if attention_mask is not None:
+            if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
+                raise ValueError(
+                    f'Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.size()}'
+                )
+            attn_weights = attn_weights + attention_mask
+
+        # upcast attention to fp32
+        attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(value_states.dtype)
+        attn_weights = nn.functional.dropout(attn_weights, p=self.attention_dropout, training=self.training)
+
+        attn_output = torch.matmul(attn_weights, value_states)
+
+        if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):
+            raise ValueError(
+                f'`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is'
+                f' {attn_output.size()}'
+            )
+
+        attn_output = attn_output.transpose(1, 2).contiguous()
+        attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
+
+        attn_output = self.o_proj(attn_output)
+
+        if not output_attentions:
+            attn_weights = None
+
+        return attn_output, attn_weights, past_key_value
+
+
+class Phi3FlashAttention2(Phi3Attention):
+    """
+    Phi-3 flash attention module. This module inherits from `Phi3Attention` as the weights of the module stays
+    untouched. The only required change would be on the forward pass where it needs to correctly call the public API of
+    flash attention and deal with padding tokens in case the input contains any of them.
+    """
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaFlashAttention2.__init__
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+
+        # TODO: Should be removed once Flash Attention for RoCm is bumped to 2.1.
+        # flash_attn<2.1 generates top-left aligned causal mask, while what is needed here is bottom-right alignement, that was made default for flash_attn>=2.1. This attribute is used to handle this difference. Reference: https://github.com/Dao-AILab/flash-attention/releases/tag/v2.1.0.
+        # Beware that with flash_attn<2.1, using q_seqlen != k_seqlen (except for the case q_seqlen == 1) produces a wrong mask (top-left).
+        self._flash_attn_uses_top_left_mask = not is_flash_attn_greater_or_equal_2_10()
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.LongTensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+        **kwargs,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        # Phi3FlashAttention2 attention does not support output_attentions
+
+        if not _flash_supports_window_size:
+            logger.warning_once(
+                "The current flash attention version does not support sliding window attention. Please use `attn_implementation='eager'` or upgrade flash-attn library."
+            )
+            raise ValueError('The current flash attention version does not support sliding window attention.')
+
+        output_attentions = False
+
+        if 'padding_mask' in kwargs:
+            warnings.warn(
+                'Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`'
+            )
+
+            # overwrite attention_mask with padding_mask
+            attention_mask = kwargs.pop('padding_mask')
+
+        bsz, q_len, _ = hidden_states.size()
+
+        qkv = self.qkv_proj(hidden_states)
+        query_pos = self.num_heads * self.head_dim
+        query_states = qkv[..., :query_pos]
+        key_states = qkv[..., query_pos : query_pos + self.num_key_value_heads * self.head_dim]
+        value_states = qkv[..., query_pos + self.num_key_value_heads * self.head_dim :]
+
+        # Flash attention requires the input to have the shape
+        # batch_size x seq_length x head_dim x hidden_dim
+        # therefore we just need to keep the original shape
+        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+
+        kv_seq_len = key_states.shape[-2]
+        if past_key_value is not None:
+            if self.layer_idx is None:
+                raise ValueError(
+                    f'The cache structure has changed since version v4.36. If you are using {self.__class__.__name__} '
+                    'for auto-regressive decoding with k/v caching, please make sure to initialize the attention class '
+                    'with a layer index.'
+                )
+            kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
+
+        # Because the input can be padded, the absolute sequence length depends on the max position id.
+        rotary_seq_len = max(kv_seq_len, position_ids[:, -1].max().item()) + 1
+        cos, sin = self.rotary_emb(value_states, position_ids, seq_len=rotary_seq_len)
+
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
+
+        use_sliding_windows = (
+            _flash_supports_window_size
+            and getattr(self.config, 'sliding_window', None) is not None
+            and kv_seq_len > self.config.sliding_window
+        )
+
+        if past_key_value is not None:
+            # Activate slicing cache only if the config has a value `sliding_windows` attribute
+            cache_has_contents = past_key_value.get_seq_length(self.layer_idx) > 0
+            if (
+                getattr(self.config, 'sliding_window', None) is not None
+                and kv_seq_len > self.config.sliding_window
+                and cache_has_contents
+            ):
+                slicing_tokens = 1 - self.config.sliding_window
+
+                past_key = past_key_value[self.layer_idx][0]
+                past_value = past_key_value[self.layer_idx][1]
+
+                past_key = past_key[:, :, slicing_tokens:, :].contiguous()
+                past_value = past_value[:, :, slicing_tokens:, :].contiguous()
+
+                if past_key.shape[-2] != self.config.sliding_window - 1:
+                    raise ValueError(
+                        f'past key must have a shape of (`batch_size, num_heads, self.config.sliding_window-1, head_dim`), got'
+                        f' {past_key.shape}'
+                    )
+
+                if attention_mask is not None:
+                    attention_mask = attention_mask[:, slicing_tokens:]
+                    attention_mask = torch.cat([attention_mask, torch.ones_like(attention_mask[:, -1:])], dim=-1)
+
+            cache_kwargs = {'sin': sin, 'cos': cos}  # Specific to RoPE models
+            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+        # repeat k/v heads if n_kv_heads < n_heads
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+        attn_dropout = self.attention_dropout if self.training else 0.0
+
+        # In PEFT, usually we cast the layer norms in float32 for training stability reasons
+        # therefore the input hidden states gets silently casted in float32. Hence, we need
+        # cast them back in the correct dtype just to be sure everything works as expected.
+        # This might slowdown training & inference so it is recommended to not cast the LayerNorms
+        # in fp32.
+
+        if query_states.dtype == torch.float32:
+            if torch.is_autocast_enabled():
+                target_dtype = torch.get_autocast_gpu_dtype()
+            # Handle the case where the model is quantized
+            elif hasattr(self.config, '_pre_quantization_dtype'):
+                target_dtype = self.config._pre_quantization_dtype
+            else:
+                target_dtype = self.qkv_proj.weight.dtype
+
+            logger.warning_once(
+                f'The input hidden states seems to be silently casted in float32, this might be related to'
+                f' the fact you have upcasted embedding or layer norm layers in float32. We will cast back the input in'
+                f' {target_dtype}.'
+            )
+
+            query_states = query_states.to(target_dtype)
+            key_states = key_states.to(target_dtype)
+            value_states = value_states.to(target_dtype)
+
+        # Reashape to the expected shape for Flash Attention
+        query_states = query_states.transpose(1, 2)
+        key_states = key_states.transpose(1, 2)
+        value_states = value_states.transpose(1, 2)
+
+        attn_output = self._flash_attention_forward(
+            query_states,
+            key_states,
+            value_states,
+            attention_mask,
+            q_len,
+            dropout=attn_dropout,
+            use_sliding_windows=use_sliding_windows,
+        )
+
+        attn_output = attn_output.reshape(bsz, q_len, self.hidden_size).contiguous()
+        attn_output = self.o_proj(attn_output)
+
+        if not output_attentions:
+            attn_weights = None
+
+        return attn_output, attn_weights, past_key_value
+
+    # Copied from transformers.models.mistral.modeling_mistral.MistralFlashAttention2._flash_attention_forward
+    def _flash_attention_forward(
+        self,
+        query_states,
+        key_states,
+        value_states,
+        attention_mask,
+        query_length,
+        dropout=0.0,
+        softmax_scale=None,
+        use_sliding_windows=False,
+    ):
+        """
+        Calls the forward method of Flash Attention - if the input hidden states contain at least one padding token
+        first unpad the input, then computes the attention scores and pad the final attention scores.
+
+        Args:
+            query_states (`torch.Tensor`):
+                Input query states to be passed to Flash Attention API
+            key_states (`torch.Tensor`):
+                Input key states to be passed to Flash Attention API
+            value_states (`torch.Tensor`):
+                Input value states to be passed to Flash Attention API
+            attention_mask (`torch.Tensor`):
+                The padding mask - corresponds to a tensor of size `(batch_size, seq_len)` where 0 stands for the
+                position of padding tokens and 1 for the position of non-padding tokens.
+            dropout (`float`):
+                Attention dropout
+            softmax_scale (`float`, *optional*):
+                The scaling of QK^T before applying softmax. Default to 1 / sqrt(head_dim)
+            use_sliding_windows (`bool`, *optional*):
+                Whether to activate sliding window attention.
+        """
+        if not self._flash_attn_uses_top_left_mask:
+            causal = self.is_causal
+        else:
+            # TODO: Remove the `query_length != 1` check once Flash Attention for RoCm is bumped to 2.1. For details, please see the comment in LlamaFlashAttention2 __init__.
+            causal = self.is_causal and query_length != 1
+
+        # Contains at least one padding token in the sequence
+        if attention_mask is not None:
+            batch_size = query_states.shape[0]
+            query_states, key_states, value_states, indices_q, cu_seq_lens, max_seq_lens = self._upad_input(
+                query_states, key_states, value_states, attention_mask, query_length
+            )
+
+            cu_seqlens_q, cu_seqlens_k = cu_seq_lens
+            max_seqlen_in_batch_q, max_seqlen_in_batch_k = max_seq_lens
+
+            if not use_sliding_windows:
+                attn_output_unpad = flash_attn_varlen_func(
+                    query_states,
+                    key_states,
+                    value_states,
+                    cu_seqlens_q=cu_seqlens_q,
+                    cu_seqlens_k=cu_seqlens_k,
+                    max_seqlen_q=max_seqlen_in_batch_q,
+                    max_seqlen_k=max_seqlen_in_batch_k,
+                    dropout_p=dropout,
+                    softmax_scale=softmax_scale,
+                    causal=causal,
+                )
+            else:
+                attn_output_unpad = flash_attn_varlen_func(
+                    query_states,
+                    key_states,
+                    value_states,
+                    cu_seqlens_q=cu_seqlens_q,
+                    cu_seqlens_k=cu_seqlens_k,
+                    max_seqlen_q=max_seqlen_in_batch_q,
+                    max_seqlen_k=max_seqlen_in_batch_k,
+                    dropout_p=dropout,
+                    softmax_scale=softmax_scale,
+                    causal=causal,
+                    window_size=(self.config.sliding_window, self.config.sliding_window),
+                )
+
+            attn_output = pad_input(attn_output_unpad, indices_q, batch_size, query_length)
+        else:
+            if not use_sliding_windows:
+                attn_output = flash_attn_func(
+                    query_states,
+                    key_states,
+                    value_states,
+                    dropout,
+                    softmax_scale=softmax_scale,
+                    causal=causal,
+                )
+            else:
+                attn_output = flash_attn_func(
+                    query_states,
+                    key_states,
+                    value_states,
+                    dropout,
+                    softmax_scale=softmax_scale,
+                    causal=causal,
+                    window_size=(self.config.sliding_window, self.config.sliding_window),
+                )
+
+        return attn_output
+
+    # Copied from transformers.models.mistral.modeling_mistral.MistralFlashAttention2._upad_input
+    def _upad_input(self, query_layer, key_layer, value_layer, attention_mask, query_length):
+        batch_size, kv_seq_len, num_heads, head_dim = key_layer.shape
+
+        # On the first iteration we need to properly re-create the padding mask
+        # by slicing it on the proper place
+        if kv_seq_len != attention_mask.shape[-1]:
+            attention_mask_num_tokens = attention_mask.shape[-1]
+            attention_mask = attention_mask[:, attention_mask_num_tokens - kv_seq_len :]
+
+        indices_k, cu_seqlens_k, max_seqlen_in_batch_k = _get_unpad_data(attention_mask)
+
+        key_layer = index_first_axis(key_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k)
+        value_layer = index_first_axis(value_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k)
+
+        if query_length == kv_seq_len:
+            query_layer = index_first_axis(
+                query_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k
+            )
+            cu_seqlens_q = cu_seqlens_k
+            max_seqlen_in_batch_q = max_seqlen_in_batch_k
+            indices_q = indices_k
+        elif query_length == 1:
+            max_seqlen_in_batch_q = 1
+            cu_seqlens_q = torch.arange(
+                batch_size + 1, dtype=torch.int32, device=query_layer.device
+            )  # There is a memcpy here, that is very bad.
+            indices_q = cu_seqlens_q[:-1]
+            query_layer = query_layer.squeeze(1)
+        else:
+            # The -q_len: slice assumes left padding.
+            attention_mask = attention_mask[:, -query_length:]
+            query_layer, indices_q, cu_seqlens_q, max_seqlen_in_batch_q = unpad_input(query_layer, attention_mask)
+
+        return (
+            query_layer,
+            key_layer,
+            value_layer,
+            indices_q,
+            (cu_seqlens_q, cu_seqlens_k),
+            (max_seqlen_in_batch_q, max_seqlen_in_batch_k),
+        )
+
+
+# copied from transformers.models.llama.modeling_llama.LlamaSdpaAttention with Llama->Phi3
+# TODO @Arthur no longer copied from LLama after static cache
+class Phi3SdpaAttention(Phi3Attention):
+    """
+    Phi3 attention module using torch.nn.functional.scaled_dot_product_attention. This module inherits from
+    `Phi3Attention` as the weights of the module stays untouched. The only changes are on the forward pass to adapt to
+    SDPA API.
+    """
+
+    # Adapted from Phi3Attention.forward
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        if output_attentions:
+            # TODO: Improve this warning with e.g. `model.config.attn_implementation = "manual"` once this is implemented.
+            logger.warning_once(
+                'Phi3Model is using Phi3SdpaAttention, but `torch.nn.functional.scaled_dot_product_attention` does not support `output_attentions=True`. Falling back to the manual attention implementation, '
+                'but specifying the manual implementation will be required from Transformers version v5.0.0 onwards. This warning can be removed using the argument `attn_implementation="eager"` when loading the model.'
+            )
+            return super().forward(
+                hidden_states=hidden_states,
+                attention_mask=attention_mask,
+                position_ids=position_ids,
+                past_key_value=past_key_value,
+                output_attentions=output_attentions,
+                use_cache=use_cache,
+            )
+
+        bsz, q_len, _ = hidden_states.size()
+
+        qkv = self.qkv_proj(hidden_states)
+        query_pos = self.num_heads * self.head_dim
+        query_states = qkv[..., :query_pos]
+        key_states = qkv[..., query_pos : query_pos + self.num_key_value_heads * self.head_dim]
+        value_states = qkv[..., query_pos + self.num_key_value_heads * self.head_dim :]
+
+        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+
+        kv_seq_len = key_states.shape[-2]
+        if past_key_value is not None:
+            kv_seq_len += past_key_value.get_usable_length(kv_seq_len, self.layer_idx)
+        cos, sin = self.rotary_emb(value_states, position_ids, seq_len=kv_seq_len)
+
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
+
+        if past_key_value is not None:
+            cache_kwargs = {'sin': sin, 'cos': cos}  # Specific to RoPE models
+            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+        if attention_mask is not None:
+            if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
+                raise ValueError(
+                    f'Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.size()}'
+                )
+
+        # SDPA with memory-efficient backend is currently (torch==2.1.2) bugged with non-contiguous inputs with custom attn_mask,
+        # Reference: https://github.com/pytorch/pytorch/issues/112577.
+        if query_states.device.type == 'cuda' and attention_mask is not None:
+            query_states = query_states.contiguous()
+            key_states = key_states.contiguous()
+            value_states = value_states.contiguous()
+
+        attn_output = torch.nn.functional.scaled_dot_product_attention(
+            query_states,
+            key_states,
+            value_states,
+            attn_mask=attention_mask,
+            dropout_p=self.attention_dropout if self.training else 0.0,
+            # The q_len > 1 is necessary to match with AttentionMaskConverter.to_causal_4d that does not create a causal mask in case q_len == 1.
+            is_causal=self.is_causal and attention_mask is None and q_len > 1,
+        )
+
+        attn_output = attn_output.transpose(1, 2).contiguous()
+        attn_output = attn_output.view(bsz, q_len, self.hidden_size)
+
+        attn_output = self.o_proj(attn_output)
+
+        return attn_output, None, past_key_value
+
+
+PHI3_ATTENTION_CLASSES = {
+    'eager': Phi3Attention,
+    'flash_attention_2': Phi3FlashAttention2,
+    'sdpa': Phi3SdpaAttention,
+}
+
+
+class Phi3DecoderLayer(nn.Module):
+    def __init__(self, config: Phi3Config, layer_idx: int):
+        super().__init__()
+
+        self.config = config
+        self.self_attn = PHI3_ATTENTION_CLASSES[config._attn_implementation](config, layer_idx=layer_idx)
+
+        self.mlp = Phi3MLP(config)
+        self.input_layernorm = Phi3RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+        self.resid_attn_dropout = nn.Dropout(config.resid_pdrop)
+        self.resid_mlp_dropout = nn.Dropout(config.resid_pdrop)
+        self.post_attention_layernorm = Phi3RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Tuple[torch.Tensor]] = None,
+        output_attentions: Optional[bool] = False,
+        use_cache: Optional[bool] = False,
+        **kwargs,
+    ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
+        if 'padding_mask' in kwargs:
+            warnings.warn(
+                'Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use `attention_mask` instead.`'
+            )
+        """
+        Args:
+            hidden_states (`torch.FloatTensor`):
+                input to the layer of shape `(batch, seq_len, embed_dim)`
+            attention_mask (`torch.FloatTensor`, *optional*): attention mask of size
+                `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
+            position_ids (`torch.LongTensor` of shape `({0})`, *optional*):
+                Indices of positions of each input sequence tokens in the position embeddings. Selected in the range
+                `[0, config.n_positions - 1]`. [What are position IDs?](../glossary#position-ids)
+            output_attentions (`bool`, *optional*):
+                Whether or not to return the attentions tensors of all attention layers. See `attentions` under
+                returned tensors for more detail.
+            use_cache (`bool`, *optional*):
+                If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding
+                (see `past_key_values`).
+            past_key_value (`Tuple(torch.FloatTensor)`, *optional*): cached past key and value projection states
+        """
+
+        residual = hidden_states
+
+        hidden_states = self.input_layernorm(hidden_states)
+
+        # Self Attention
+        attn_outputs, self_attn_weights, present_key_value = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_value=past_key_value,
+            output_attentions=output_attentions,
+            use_cache=use_cache,
+        )
+
+        hidden_states = residual + self.resid_attn_dropout(attn_outputs)
+
+        residual = hidden_states
+        hidden_states = self.post_attention_layernorm(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        hidden_states = residual + self.resid_mlp_dropout(hidden_states)
+
+        outputs = (hidden_states,)
+
+        if output_attentions:
+            outputs += (self_attn_weights,)
+
+        if use_cache:
+            outputs += (present_key_value,)
+
+        return outputs
+
+
+PHI3_START_DOCSTRING = r"""
+    This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the
+    library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
+    etc.)
+
+    This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass.
+    Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage
+    and behavior.
+
+    Parameters:
+        config ([`Phi3Config`]):
+            Model configuration class with all the parameters of the model. Initializing with a config file does not
+            load the weights associated with the model, only the configuration. Check out the
+            [`~PreTrainedModel.from_pretrained`] method to load the model weights.
+"""
+
+
+@add_start_docstrings(
+    'The bare Phi-3 model outputting raw hidden-states without any specific head on top.',
+    PHI3_START_DOCSTRING,
+)
+class Phi3PreTrainedModel(PreTrainedModel):
+    config_class = Phi3Config
+    base_model_prefix = 'model'
+    supports_gradient_checkpointing = True
+    _no_split_modules = ['Phi3DecoderLayer']
+    _skip_keys_device_placement = 'past_key_values'
+    _supports_flash_attn_2 = True
+    _supports_sdpa = False
+    _supports_cache_class = True
+
+    _version = '0.0.5'
+
+    def _init_weights(self, module):
+        std = self.config.initializer_range
+        if isinstance(module, nn.Linear):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.Embedding):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.padding_idx is not None:
+                module.weight.data[module.padding_idx].zero_()
+
+
+PHI3_INPUTS_DOCSTRING = r"""
+    Args:
+        input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
+            Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
+            it.
+
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+
+            [What are input IDs?](../glossary#input-ids)
+        attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
+
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+
+            [What are attention masks?](../glossary#attention-mask)
+
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+
+            If `past_key_values` is used, optionally only the last `input_ids` have to be input (see
+            `past_key_values`).
+
+            If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`]
+            and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more
+            information on the default strategy.
+
+            - 1 indicates the head is **not masked**,
+            - 0 indicates the head is **masked**.
+        position_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
+            config.n_positions - 1]`.
+
+            [What are position IDs?](../glossary#position-ids)
+        past_key_values (`Cache` or `tuple(tuple(torch.FloatTensor))`, *optional*):
+            Pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention
+            blocks) that can be used to speed up sequential decoding. This typically consists in the `past_key_values`
+            returned by the model at a previous stage of decoding, when `use_cache=True` or `config.use_cache=True`.
+
+            Two formats are allowed:
+            - a [`~cache_utils.Cache`] instance;
+            - Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of
+            shape `(batch_size, num_heads, sequence_length, embed_size_per_head)`). This is also known as the legacy
+            cache format.
+
+            The model will output the same cache format that is fed as input. If no `past_key_values` are passed, the
+            legacy cache format will be returned.
+
+            If `past_key_values` are used, the user can optionally input only the last `input_ids` (those that don't
+            have their past key value states given to this model) of shape `(batch_size, 1)` instead of all `input_ids`
+            of shape `(batch_size, sequence_length)`.
+        inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
+            Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This
+            is useful if you want more control over how to convert `input_ids` indices into associated vectors than the
+            model's internal embedding lookup matrix.
+        use_cache (`bool`, *optional*):
+            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
+            `past_key_values`).
+        output_attentions (`bool`, *optional*):
+            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
+            tensors for more detail.
+        output_hidden_states (`bool`, *optional*):
+            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
+            more detail.
+        return_dict (`bool`, *optional*):
+            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
+"""
+
+
+@add_start_docstrings(
+    'The bare Phi-3 model outputting raw hidden-states without any specific head on top.',
+    PHI3_START_DOCSTRING,
+)
+class Phi3Model(Phi3PreTrainedModel):
+    """
+    Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`Phi3DecoderLayer`]
+
+    Args:
+        config: Phi3Config
+    """
+
+    def __init__(self, config: Phi3Config):
+        super().__init__(config)
+        self.padding_idx = config.pad_token_id
+        self.vocab_size = config.vocab_size
+
+        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
+        self.embed_dropout = nn.Dropout(config.embd_pdrop)
+        self.layers = nn.ModuleList(
+            [Phi3DecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
+        )
+        self._attn_implementation = config._attn_implementation
+        self.norm = Phi3RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+        self.gradient_checkpointing = False
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.embed_tokens = value
+
+    @add_start_docstrings_to_model_forward(PHI3_INPUTS_DOCSTRING)
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, BaseModelOutputWithPast]:
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        use_cache = use_cache if use_cache is not None else self.config.use_cache
+
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        # retrieve input_ids and inputs_embeds
+        if input_ids is not None and inputs_embeds is not None:
+            raise ValueError('You cannot specify both input_ids and inputs_embeds at the same time')
+        elif input_ids is not None:
+            batch_size, seq_length = input_ids.shape[:2]
+        elif inputs_embeds is not None:
+            batch_size, seq_length = inputs_embeds.shape[:2]
+        else:
+            raise ValueError('You have to specify either input_ids or inputs_embeds')
+
+        past_key_values_length = 0
+
+        if self.gradient_checkpointing and self.training:
+            if use_cache:
+                logger.warning_once(
+                    '`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`...'
+                )
+                use_cache = False
+
+        if use_cache:
+            use_legacy_cache = not isinstance(past_key_values, Cache)
+            if use_legacy_cache:
+                past_key_values = DynamicCache.from_legacy_cache(past_key_values)
+            past_key_values_length = past_key_values.get_usable_length(seq_length)
+
+        if position_ids is None:
+            device = input_ids.device if input_ids is not None else inputs_embeds.device
+            position_ids = torch.arange(
+                past_key_values_length, seq_length + past_key_values_length, dtype=torch.long, device=device
+            )
+            position_ids = position_ids.unsqueeze(0).view(-1, seq_length)
+        else:
+            position_ids = position_ids.view(-1, seq_length).long()
+
+        if inputs_embeds is None:
+            inputs_embeds = self.embed_tokens(input_ids)
+
+        if attention_mask is not None and self._attn_implementation == 'flash_attention_2' and use_cache:
+            is_padding_right = attention_mask[:, -1].sum().item() != batch_size
+            if is_padding_right:
+                raise ValueError(
+                    "You are attempting to perform batched generation with padding_side='right'"
+                    ' this may lead to unexpected behaviour for Flash Attention version of Phi3. Make sure to '
+                    " call `tokenizer.padding_side  = 'left'` before tokenizing the input. "
+                )
+
+        if self._attn_implementation == 'flash_attention_2':
+            # 2d mask is passed through the layers
+            attention_mask = attention_mask if (attention_mask is not None and 0 in attention_mask) else None
+        else:
+            # 4d mask is passed through the layers
+            attention_mask = _prepare_4d_causal_attention_mask(
+                attention_mask,
+                (batch_size, seq_length),
+                inputs_embeds,
+                past_key_values_length,
+                sliding_window=self.config.sliding_window,
+            )
+
+        hidden_states = inputs_embeds
+
+        # decoder layers
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attns = () if output_attentions else None
+        next_decoder_cache = None
+
+        for decoder_layer in self.layers:
+            if output_hidden_states:
+                all_hidden_states += (hidden_states,)
+
+            if self.gradient_checkpointing and self.training:
+                layer_outputs = self._gradient_checkpointing_func(
+                    decoder_layer.__call__,
+                    hidden_states,
+                    attention_mask,
+                    position_ids,
+                    past_key_values,
+                    output_attentions,
+                    use_cache,
+                )
+            else:
+                layer_outputs = decoder_layer(
+                    hidden_states,
+                    attention_mask=attention_mask,
+                    position_ids=position_ids,
+                    past_key_value=past_key_values,
+                    output_attentions=output_attentions,
+                    use_cache=use_cache,
+                )
+
+            hidden_states = layer_outputs[0]
+
+            if use_cache:
+                next_decoder_cache = layer_outputs[2 if output_attentions else 1]
+
+            if output_attentions:
+                all_self_attns += (layer_outputs[1],)
+
+        hidden_states = self.norm(hidden_states)
+
+        # add hidden states from the last decoder layer
+        if output_hidden_states:
+            all_hidden_states += (hidden_states,)
+
+        next_cache = None
+        if use_cache:
+            next_cache = next_decoder_cache.to_legacy_cache() if use_legacy_cache else next_decoder_cache
+        if not return_dict:
+            return tuple(v for v in [hidden_states, next_cache, all_hidden_states, all_self_attns] if v is not None)
+        return BaseModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=next_cache,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attns,
+        )
+
+
+class Phi3ForCausalLM(Phi3PreTrainedModel):
+    _tied_weights_keys = ['lm_head.weight']
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.__init__ with Llama->Phi3
+    def __init__(self, config):
+        super().__init__(config)
+        self.model = Phi3Model(config)
+        self.vocab_size = config.vocab_size
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.get_input_embeddings
+    def get_input_embeddings(self):
+        return self.model.embed_tokens
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.set_input_embeddings
+    def set_input_embeddings(self, value):
+        self.model.embed_tokens = value
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.get_output_embeddings
+    def get_output_embeddings(self):
+        return self.lm_head
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.set_output_embeddings
+    def set_output_embeddings(self, new_embeddings):
+        self.lm_head = new_embeddings
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.set_decoder
+    def set_decoder(self, decoder):
+        self.model = decoder
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.get_decoder
+    def get_decoder(self):
+        return self.model
+
+    # Ignore copy
+    @add_start_docstrings_to_model_forward(PHI3_INPUTS_DOCSTRING)
+    @replace_return_docstrings(output_type=CausalLMOutputWithPast, config_class=_CONFIG_FOR_DOC)
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, CausalLMOutputWithPast]:
+        r"""
+        Args:
+            labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+                Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
+                config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
+                (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.
+
+        Returns:
+
+        Example:
+
+        ```python
+        >>> from transformers import AutoTokenizer, Phi3ForCausalLM
+
+        >>> model = Phi3ForCausalLM.from_pretrained("microsoft/phi-3-mini-4k-instruct")
+        >>> tokenizer = AutoTokenizer.from_pretrained("microsoft/phi-3-mini-4k-instruct")
+
+        >>> prompt = "This is an example script ."
+        >>> inputs = tokenizer(prompt, return_tensors="pt")
+
+        >>> # Generate
+        >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
+        >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
+        'This is an example script .\n Certainly! Below is a sample script that demonstrates a simple task, such as calculating the sum'
+        ```"""
+
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
+        outputs = self.model(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        hidden_states = outputs[0]
+        logits = self.lm_head(hidden_states)
+        logits = logits.float()
+
+        loss = None
+        if labels is not None:
+            # Shift so that tokens < n predict n
+            shift_logits = logits[..., :-1, :].contiguous()
+            shift_labels = labels[..., 1:].contiguous()
+            # Flatten the tokens
+            loss_fct = CrossEntropyLoss()
+            shift_logits = shift_logits.view(-1, self.config.vocab_size)
+            shift_labels = shift_labels.view(-1)
+            # Enable model parallelism
+            shift_labels = shift_labels.to(shift_logits.device)
+            loss = loss_fct(shift_logits, shift_labels)
+
+        if not return_dict:
+            output = (logits,) + outputs[1:]
+            return (loss,) + output if loss is not None else output
+
+        return CausalLMOutputWithPast(
+            loss=loss,
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+    # Copied from transformers.models.persimmon.modeling_persimmon.PersimmonForCausalLM.prepare_inputs_for_generation
+    def prepare_inputs_for_generation(
+        self, input_ids, past_key_values=None, attention_mask=None, inputs_embeds=None, **kwargs
+    ):
+        if past_key_values is not None:
+            if isinstance(past_key_values, Cache):
+                cache_length = past_key_values.get_seq_length()
+                past_length = past_key_values.seen_tokens
+                max_cache_length = past_key_values.get_max_length()
+            else:
+                cache_length = past_length = past_key_values[0][0].shape[2]
+                max_cache_length = None
+
+            # Keep only the unprocessed tokens:
+            # 1 - If the length of the attention_mask exceeds the length of input_ids, then we are in a setting where
+            # some of the inputs are exclusively passed as part of the cache (e.g. when passing input_embeds as
+            # input)
+            if attention_mask is not None and attention_mask.shape[1] > input_ids.shape[1]:
+                input_ids = input_ids[:, -(attention_mask.shape[1] - past_length) :]
+            # 2 - If the past_length is smaller than input_ids', then input_ids holds all input tokens. We can discard
+            # input_ids based on the past_length.
+            elif past_length < input_ids.shape[1]:
+                input_ids = input_ids[:, past_length:]
+            # 3 - Otherwise (past_length >= input_ids.shape[1]), let's assume input_ids only has unprocessed tokens.
+
+            # If we are about to go beyond the maximum cache length, we need to crop the input attention mask.
+            if (
+                max_cache_length is not None
+                and attention_mask is not None
+                and cache_length + input_ids.shape[1] > max_cache_length
+            ):
+                attention_mask = attention_mask[:, -max_cache_length:]
+
+        position_ids = kwargs.get('position_ids', None)
+        if attention_mask is not None and position_ids is None:
+            # create position_ids on the fly for batch generation
+            position_ids = attention_mask.long().cumsum(-1) - 1
+            position_ids.masked_fill_(attention_mask == 0, 1)
+            if past_key_values:
+                position_ids = position_ids[:, -input_ids.shape[1] :]
+
+        # if `inputs_embeds` are passed, we only want to use them in the 1st generation step
+        if inputs_embeds is not None and past_key_values is None:
+            model_inputs = {'inputs_embeds': inputs_embeds}
+        else:
+            model_inputs = {'input_ids': input_ids}
+
+        model_inputs.update(
+            {
+                'position_ids': position_ids,
+                'past_key_values': past_key_values,
+                'use_cache': kwargs.get('use_cache'),
+                'attention_mask': attention_mask,
+            }
+        )
+        return model_inputs
+
+    @staticmethod
+    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM._reorder_cache
+    def _reorder_cache(past_key_values, beam_idx):
+        reordered_past = ()
+        for layer_past in past_key_values:
+            reordered_past += (
+                tuple(past_state.index_select(0, beam_idx.to(past_state.device)) for past_state in layer_past),
+            )
+        return reordered_past
+
+
+@add_start_docstrings(
+    """
+    The [`Phi3Model`] with a sequence classification head on top (linear layer).
+
+    [`Phi3ForSequenceClassification`] uses the last token in order to do the classification, as other causal models
+    (e.g. GPT-2) do.
+
+    Since it does classification on the last token, it requires to know the position of the last token. If a
+    `pad_token_id` is defined in the configuration, it finds the last token that is not a padding token in each row. If
+    no `pad_token_id` is defined, it simply takes the last value in each row of the batch. Since it cannot guess the
+    padding tokens when `inputs_embeds` are passed instead of `input_ids`, it does the same (take the last value in
+    each row of the batch).
+    """,
+    PHI3_START_DOCSTRING,
+)
+# Copied from transformers.models.llama.modeling_llama.LlamaForSequenceClassification with Llama->Phi3, LLAMA->PHI3, self.transformer->self.model, transformer_outputs->model_outputs
+class Phi3ForSequenceClassification(Phi3PreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.model = Phi3Model(config)
+        self.score = nn.Linear(config.hidden_size, self.num_labels, bias=False)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.model.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.model.embed_tokens = value
+
+    @add_start_docstrings_to_model_forward(PHI3_INPUTS_DOCSTRING)
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple, SequenceClassifierOutputWithPast]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
+            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
+            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        model_outputs = self.model(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+        hidden_states = model_outputs[0]
+        logits = self.score(hidden_states)
+
+        if input_ids is not None:
+            batch_size = input_ids.shape[0]
+        else:
+            batch_size = inputs_embeds.shape[0]
+
+        if self.config.pad_token_id is None and batch_size != 1:
+            raise ValueError('Cannot handle batch sizes > 1 if no padding token is defined.')
+        if self.config.pad_token_id is None:
+            sequence_lengths = -1
+        else:
+            if input_ids is not None:
+                # if no pad token found, use modulo instead of reverse indexing for ONNX compatibility
+                sequence_lengths = torch.eq(input_ids, self.config.pad_token_id).int().argmax(-1) - 1
+                sequence_lengths = sequence_lengths % input_ids.shape[-1]
+                sequence_lengths = sequence_lengths.to(logits.device)
+            else:
+                sequence_lengths = -1
+
+        pooled_logits = logits[torch.arange(batch_size, device=logits.device), sequence_lengths]
+
+        loss = None
+        if labels is not None:
+            labels = labels.to(logits.device)
+            if self.config.problem_type is None:
+                if self.num_labels == 1:
+                    self.config.problem_type = 'regression'
+                elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
+                    self.config.problem_type = 'single_label_classification'
+                else:
+                    self.config.problem_type = 'multi_label_classification'
+
+            if self.config.problem_type == 'regression':
+                loss_fct = MSELoss()
+                if self.num_labels == 1:
+                    loss = loss_fct(pooled_logits.squeeze(), labels.squeeze())
+                else:
+                    loss = loss_fct(pooled_logits, labels)
+            elif self.config.problem_type == 'single_label_classification':
+                loss_fct = CrossEntropyLoss()
+                loss = loss_fct(pooled_logits.view(-1, self.num_labels), labels.view(-1))
+            elif self.config.problem_type == 'multi_label_classification':
+                loss_fct = BCEWithLogitsLoss()
+                loss = loss_fct(pooled_logits, labels)
+        if not return_dict:
+            output = (pooled_logits,) + model_outputs[1:]
+            return ((loss,) + output) if loss is not None else output
+
+        return SequenceClassifierOutputWithPast(
+            loss=loss,
+            logits=pooled_logits,
+            past_key_values=model_outputs.past_key_values,
+            hidden_states=model_outputs.hidden_states,
+            attentions=model_outputs.attentions,
+        )
+
+
+@add_start_docstrings(
+    """
+    [`Phi3Model`] with a token classification head on top (a linear layer on top of the hidden-states output) e.g. for
+    Named-Entity-Recognition (NER) tasks.
+    """,
+    PHI3_START_DOCSTRING,
+)
+# Copied from transformers.models.mpt.modeling_mpt.MptForTokenClassification with Mpt->Phi3,MPT->PHI3,self.transformer->self.model,transformer_outputs->model_outputs
+class Phi3ForTokenClassification(Phi3PreTrainedModel):
+    def __init__(self, config: Phi3Config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+
+        self.model = Phi3Model(config)
+        if hasattr(config, 'classifier_dropout') and config.classifier_dropout is not None:
+            classifier_dropout = config.classifier_dropout
+        elif hasattr(config, 'hidden_dropout') and config.hidden_dropout is not None:
+            classifier_dropout = config.hidden_dropout
+        else:
+            classifier_dropout = 0.1
+        self.dropout = nn.Dropout(classifier_dropout)
+        self.classifier = nn.Linear(config.hidden_size, config.num_labels)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    @add_start_docstrings_to_model_forward(PHI3_INPUTS_DOCSTRING)
+    @add_code_sample_docstrings(
+        checkpoint=_CHECKPOINT_FOR_DOC,
+        output_type=TokenClassifierOutput,
+        config_class=_CONFIG_FOR_DOC,
+    )
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Tuple[Tuple[torch.Tensor, torch.Tensor], ...]] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        labels: Optional[torch.Tensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        **deprecated_arguments,
+    ) -> Union[Tuple[torch.Tensor], TokenClassifierOutput]:
+        r"""
+        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
+            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
+            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
+            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
+        """
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        model_outputs = self.model(
+            input_ids,
+            past_key_values=past_key_values,
+            attention_mask=attention_mask,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        hidden_states = model_outputs[0]
+        hidden_states = self.dropout(hidden_states)
+        logits = self.classifier(hidden_states)
+
+        loss = None
+        if labels is not None:
+            # move labels to correct device to enable model parallelism
+            labels = labels.to(logits.device)
+            batch_size, seq_length = labels.shape
+            loss_fct = CrossEntropyLoss()
+            loss = loss_fct(
+                logits.view(batch_size * seq_length, self.num_labels), labels.view(batch_size * seq_length)
+            )
+
+        if not return_dict:
+            output = (logits,) + model_outputs[2:]
+            return ((loss,) + output) if loss is not None else output
+
+        return TokenClassifierOutput(
+            loss=loss,
+            logits=logits,
+            hidden_states=model_outputs.hidden_states,
+            attentions=model_outputs.attentions,
+        )

pytorch_model-00001-of-00002.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b7cdf24050a0d5bb80762233ee658eede52c2101cad1b7fc7e5cf6f17357467
+size 4957499771

pytorch_model-00002-of-00002.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:21fdf60a0981106fe9435e15333fdbfe3d718897452dc85053fa0b28fee680e5
+size 3336407358

pytorch_model.bin.index.json

@@ -0,0 +1,548 @@
+{
+  "metadata": {
+    "total_size": 8293711872
+  },
+  "weight_map": {
+    "language_model.lm_head.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.embed_tokens.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.0.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.0.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.0.mlp.gate_up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.0.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.0.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.0.self_attn.qkv_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.1.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.1.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.1.mlp.gate_up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.1.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.1.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.1.self_attn.qkv_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.10.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.10.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.10.mlp.gate_up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.10.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.10.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.10.self_attn.qkv_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.11.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.11.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.11.mlp.gate_up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.11.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.11.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.11.self_attn.qkv_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.12.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.12.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.12.mlp.gate_up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.12.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.12.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.12.self_attn.qkv_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.13.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.13.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.13.mlp.gate_up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.13.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.13.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.13.self_attn.qkv_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.14.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.14.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.14.mlp.gate_up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.14.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.14.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.14.self_attn.qkv_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.15.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.15.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.15.mlp.gate_up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.15.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.15.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.15.self_attn.qkv_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.16.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.16.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.16.mlp.gate_up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.16.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.16.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.16.self_attn.qkv_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.17.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.17.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.17.mlp.gate_up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.17.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.17.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.17.self_attn.qkv_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.18.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.18.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.18.mlp.gate_up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.18.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.18.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.18.self_attn.qkv_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.19.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.19.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.19.mlp.gate_up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.19.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.19.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.19.self_attn.qkv_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.2.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.2.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.2.mlp.gate_up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.2.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.2.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.2.self_attn.qkv_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.20.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.20.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.20.mlp.gate_up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.20.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.20.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.20.self_attn.qkv_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.21.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.21.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.21.mlp.gate_up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.21.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.21.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.21.self_attn.qkv_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.22.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.22.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.22.mlp.gate_up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.22.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.22.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.22.self_attn.qkv_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.23.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.23.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.23.mlp.gate_up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.23.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.23.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.23.self_attn.qkv_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.24.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.24.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.24.mlp.gate_up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.24.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.24.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.24.self_attn.qkv_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.25.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.25.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.25.mlp.gate_up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.25.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.25.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.25.self_attn.qkv_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.26.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.26.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.26.mlp.gate_up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.26.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.26.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.26.self_attn.qkv_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.27.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.27.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.27.mlp.gate_up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.27.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.27.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.27.self_attn.qkv_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.28.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.28.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.28.mlp.gate_up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.28.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.28.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.28.self_attn.qkv_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.29.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.29.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.29.mlp.gate_up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.29.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.29.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.29.self_attn.qkv_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.3.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.3.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.3.mlp.gate_up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.3.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.3.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.3.self_attn.qkv_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.30.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.30.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.30.mlp.gate_up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.30.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.30.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.30.self_attn.qkv_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.31.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.31.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.31.mlp.gate_up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.31.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.31.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.31.self_attn.qkv_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "language_model.model.layers.4.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.4.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.4.mlp.gate_up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.4.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.4.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.4.self_attn.qkv_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.5.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.5.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.5.mlp.gate_up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.5.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.5.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.5.self_attn.qkv_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.6.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.6.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.6.mlp.gate_up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.6.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.6.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.6.self_attn.qkv_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.7.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.7.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.7.mlp.gate_up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.7.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.7.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.7.self_attn.qkv_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.8.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.8.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.8.mlp.gate_up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.8.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.8.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.8.self_attn.qkv_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.9.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.9.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.9.mlp.gate_up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.9.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.9.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.layers.9.self_attn.qkv_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "language_model.model.norm.weight": "pytorch_model-00002-of-00002.bin",
+    "mlp1.0.bias": "pytorch_model-00002-of-00002.bin",
+    "mlp1.0.weight": "pytorch_model-00002-of-00002.bin",
+    "mlp1.1.bias": "pytorch_model-00002-of-00002.bin",
+    "mlp1.1.weight": "pytorch_model-00002-of-00002.bin",
+    "mlp1.3.bias": "pytorch_model-00002-of-00002.bin",
+    "mlp1.3.weight": "pytorch_model-00002-of-00002.bin",
+    "vision_model.embeddings.class_embedding": "pytorch_model-00001-of-00002.bin",
+    "vision_model.embeddings.patch_embedding.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.embeddings.patch_embedding.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.embeddings.position_embedding": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.0.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.0.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.0.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.0.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.0.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.0.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.0.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.0.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.0.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.0.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.0.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.0.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.0.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.0.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.1.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.1.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.1.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.1.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.1.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.1.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.1.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.1.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.1.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.1.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.1.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.1.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.1.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.1.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.10.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.10.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.10.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.10.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.10.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.10.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.10.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.10.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.10.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.10.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.10.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.10.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.10.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.10.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.11.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.11.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.11.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.11.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.11.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.11.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.11.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.11.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.11.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.11.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.11.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.11.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.11.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.11.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.12.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.12.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.12.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.12.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.12.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.12.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.12.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.12.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.12.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.12.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.12.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.12.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.12.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.12.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.13.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.13.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.13.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.13.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.13.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.13.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.13.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.13.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.13.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.13.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.13.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.13.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.13.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.13.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.14.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.14.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.14.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.14.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.14.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.14.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.14.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.14.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.14.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.14.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.14.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.14.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.14.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.14.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.15.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.15.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.15.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.15.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.15.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.15.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.15.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.15.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.15.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.15.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.15.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.15.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.15.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.15.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.16.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.16.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.16.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.16.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.16.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.16.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.16.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.16.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.16.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.16.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.16.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.16.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.16.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.16.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.17.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.17.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.17.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.17.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.17.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.17.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.17.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.17.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.17.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.17.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.17.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.17.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.17.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.17.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.18.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.18.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.18.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.18.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.18.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.18.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.18.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.18.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.18.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.18.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.18.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.18.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.18.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.18.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.19.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.19.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.19.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.19.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.19.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.19.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.19.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.19.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.19.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.19.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.19.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.19.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.19.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.19.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.2.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.2.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.2.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.2.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.2.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.2.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.2.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.2.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.2.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.2.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.2.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.2.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.2.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.2.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.20.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.20.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.20.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.20.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.20.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.20.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.20.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.20.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.20.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.20.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.20.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.20.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.20.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.20.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.21.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.21.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.21.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.21.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.21.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.21.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.21.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.21.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.21.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.21.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.21.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.21.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.21.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.21.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.22.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.22.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.22.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.22.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.22.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.22.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.22.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.22.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.22.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.22.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.22.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.22.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.22.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.22.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.23.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.23.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.23.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.23.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.23.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.23.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.23.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.23.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.23.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.23.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.23.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.23.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.23.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.23.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.3.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.3.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.3.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.3.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.3.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.3.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.3.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.3.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.3.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.3.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.3.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.3.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.3.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.3.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.4.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.4.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.4.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.4.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.4.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.4.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.4.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.4.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.4.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.4.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.4.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.4.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.4.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.4.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.5.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.5.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.5.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.5.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.5.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.5.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.5.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.5.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.5.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.5.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.5.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.5.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.5.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.5.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.6.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.6.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.6.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.6.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.6.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.6.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.6.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.6.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.6.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.6.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.6.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.6.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.6.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.6.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.7.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.7.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.7.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.7.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.7.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.7.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.7.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.7.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.7.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.7.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.7.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.7.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.7.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.7.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.8.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.8.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.8.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.8.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.8.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.8.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.8.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.8.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.8.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.8.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.8.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.8.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.8.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.8.norm2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.9.attn.proj.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.9.attn.proj.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.9.attn.qkv.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.9.attn.qkv.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.9.ls1": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.9.ls2": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.9.mlp.fc1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.9.mlp.fc1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.9.mlp.fc2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.9.mlp.fc2.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.9.norm1.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.9.norm1.weight": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.9.norm2.bias": "pytorch_model-00001-of-00002.bin",
+    "vision_model.encoder.layers.9.norm2.weight": "pytorch_model-00001-of-00002.bin"
+  }
+}