Upload Idefics3blForConditionalGeneration

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 +1,170 @@
+{
+  "architectures": [
+    "Idefics3blForConditionalGeneration"
+  ],
+  "auto_map": {
+    "AutoConfig": "configuration_idefics3bl.Idefics3blConfig",
+    "AutoModelForVision2Seq": "modeling_idefics3bl.Idefics3blForConditionalGeneration"
+  },
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "image_token_id": 49190,
+  "midblock_end": -1,
+  "midblock_ratio": 1.0,
+  "midblock_start": -1,
+  "model_type": "idefics3bl",
+  "pad_token_id": 0,
+  "scale_factor": 4,
+  "text_config": {
+    "_flash_attn_2_enabled": true,
+    "_name_or_path": "None",
+    "architectures": [
+      "VLlama3ForCausalLMbl"
+    ],
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "head_dim": 64,
+    "hidden_act": "silu",
+    "hidden_size": 960,
+    "initializer_range": 0.02,
+    "intermediate_size": 2560,
+    "is_llama_config": true,
+    "max_position_embeddings": 8192,
+    "midblock_end": 31,
+    "midblock_ratio": 0.6,
+    "midblock_start": 1,
+    "mlp_bias": false,
+    "model_type": "llama",
+    "neftune_noise_alpha": 0.0,
+    "num_attention_heads": 15,
+    "num_hidden_layers": 32,
+    "num_key_value_heads": 5,
+    "pad_token_id": 2,
+    "perceiver_config": {
+      "_name_or_path": "",
+      "add_cross_attention": false,
+      "architectures": null,
+      "attention_dropout": 0.0,
+      "bad_words_ids": null,
+      "begin_suppress_tokens": null,
+      "bos_token_id": null,
+      "chunk_size_feed_forward": 0,
+      "cross_attention_hidden_size": null,
+      "decoder_start_token_id": null,
+      "diversity_penalty": 0.0,
+      "do_sample": false,
+      "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": "silu",
+      "id2label": {
+        "0": "LABEL_0",
+        "1": "LABEL_1"
+      },
+      "is_decoder": false,
+      "is_encoder_decoder": false,
+      "label2id": {
+        "LABEL_0": 0,
+        "LABEL_1": 1
+      },
+      "length_penalty": 1.0,
+      "max_length": 20,
+      "min_length": 0,
+      "model_type": "vllama3bl",
+      "no_repeat_ngram_size": 0,
+      "num_beam_groups": 1,
+      "num_beams": 1,
+      "num_key_value_heads": 1,
+      "num_return_sequences": 1,
+      "output_attentions": false,
+      "output_hidden_states": false,
+      "output_scores": false,
+      "pad_token_id": null,
+      "prefix": null,
+      "problem_type": null,
+      "pruned_heads": {},
+      "qk_layer_norms_perceiver": false,
+      "remove_invalid_values": false,
+      "repetition_penalty": 1.0,
+      "resampler_depth": 6,
+      "resampler_head_dim": 96,
+      "resampler_n_heads": 16,
+      "resampler_n_latents": 64,
+      "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": null,
+      "torchscript": false,
+      "transformers_version": "4.46.0",
+      "typical_p": 1.0,
+      "use_bfloat16": false
+    },
+    "pixel_shuffle_factor": 4,
+    "pretraining_tp": 1,
+    "qk_layer_norms": false,
+    "rms_norm_eps": 1e-05,
+    "rope_interleaved": false,
+    "rope_scaling": null,
+    "rope_theta": 100000,
+    "torch_dtype": "float32",
+    "transformers.js_config": {
+      "kv_cache_dtype": {
+        "fp16": "float16",
+        "q4f16": "float16"
+      }
+    },
+    "use_cache": true,
+    "use_resampler": false,
+    "vocab_size": 49280
+  },
+  "tie_word_embeddings": false,
+  "torch_dtype": "float32",
+  "transformers.js_config": {
+    "kv_cache_dtype": {
+      "fp16": "float16",
+      "q4f16": "float16"
+    }
+  },
+  "transformers_version": "4.53.2",
+  "use_cache": true,
+  "vision_config": {
+    "attention_dropout": 0.0,
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_size": 768,
+    "image_size": 512,
+    "initializer_range": 0.02,
+    "intermediate_size": 3072,
+    "layer_norm_eps": 1e-06,
+    "max_image_size": {
+      "longest_edge": 512
+    },
+    "midblock_end": 12,
+    "midblock_ratio": 0.5,
+    "midblock_start": 1,
+    "model_type": "idefics3_visionbl",
+    "num_attention_heads": 12,
+    "num_channels": 3,
+    "num_hidden_layers": 12,
+    "patch_size": 16,
+    "size": {
+      "longest_edge": 2048
+    },
+    "tie_word_embeddings": false,
+    "torch_dtype": "float32",
+    "use_base_siglip": false
+  },
+  "vocab_size": 49280
+}

configuration_idefics3bl.py

@@ -0,0 +1,204 @@
+# coding=utf-8
+# Copyright 2024 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.
+"""Idefics3 model configuration"""
+
+from transformers import PretrainedConfig
+from transformers.utils import logging
+from transformers import CONFIG_MAPPING, AutoConfig
+
+
+logger = logging.get_logger(__name__)
+
+
+class Idefics3blVisionConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`Idefics3VisionModel`]. It is used to instantiate a
+    Idefics3 vision encoder according to the specified arguments, defining the model architecture. Instantiating a
+    configuration with the defaults will yield a similar configuration to that of the SigLIP checkpoint
+    [google/siglip-base-patch16-224](https://huggingface.co/google/siglip-base-patch16-224) used in the Idefics3 model
+    [HuggingFaceM4/Idefics3-8B-Llama3](https://huggingface.co/HuggingFaceM4/Idefics3-8B-Llama3).
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+    Args:
+        hidden_size (`int`, *optional*, defaults to 1152):
+            Dimensionality of the encoder layers and the pooler layer.
+        intermediate_size (`int`, *optional*, defaults to 3072):
+            Dimensionality of the "intermediate" (i.e., feed-forward) layer in the Transformer encoder.
+        num_hidden_layers (`int`, *optional*, defaults to 12):
+            Number of hidden layers in the Transformer encoder.
+        num_attention_heads (`int`, *optional*, defaults to 16):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        num_channels (`int`, *optional*, defaults to 3):
+            Number of channels in the input images.
+        image_size (`int`, *optional*, defaults to 224):
+            The size (resolution) of each image.
+        patch_size (`int`, *optional*, defaults to 32):
+            The size (resolution) of each patch.
+        hidden_act (`str` or `function`, *optional*, defaults to `"gelu_pytorch_tanh"`):
+            The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,
+            `"relu"`, `"selu"` and `"gelu_new"` `"quick_gelu"` are supported.
+        layer_norm_eps (`float`, *optional*, defaults to 1e-06):
+            The epsilon used by the layer normalization layers.
+        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.
+
+    Example:
+
+    ```python
+    >>> from transformers.models.idefics3.modeling_idefics3 import Idefics3VisionTransformer
+    >>> from transformers.models.idefics3.configuration_idefics3 import Idefics3VisionConfig
+
+    >>> # Initializing a Idefics3VisionConfig with google/siglip-base-patch16-224 style configuration
+    >>> configuration = Idefics3VisionConfig()
+
+    >>> # Initializing a Idefics3VisionTransformer (with random weights) from the google/siglip-base-patch16-224 style configuration
+    >>> model = Idefics3VisionTransformer(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "idefics3_visionbl"
+    base_config_key = "vision_config"
+
+    def __init__(
+        self,
+        hidden_size=1152,
+        intermediate_size=3072,
+        num_hidden_layers=12,
+        num_attention_heads=16,
+        num_channels=3,
+        image_size=224,
+        patch_size=32,
+        hidden_act="gelu_pytorch_tanh",
+        layer_norm_eps=1e-6,
+        attention_dropout=0.0,
+        initializer_range=0.02,
+        midblock_ratio=1.0,
+        midblock_start=-1,
+        midblock_end=-1,
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        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.attention_dropout = attention_dropout
+        self.layer_norm_eps = layer_norm_eps
+        self.hidden_act = hidden_act
+        self.initializer_range = initializer_range
+        self.midblock_ratio = midblock_ratio
+        self.midblock_start = midblock_start
+        self.midblock_end = midblock_end
+        
+
+
+class Idefics3blConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`Idefics3Model`]. It is used to instantiate a
+    Idefics3 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 model of the Idefics3
+    [HuggingFaceM4/Idefics3-8B-Llama3](https://huggingface.co/HuggingFaceM4/Idefics3-8B-Llama3) architecture.
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+    Args:
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether or not the model should cache the key/value pairs of the attention mechanism. Only
+            relevant if `config.is_decoder=True`.
+        image_token_id (`int`, *optional*, defaults to 128257):
+            The id of the "image" token.
+        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
+            Whether or not to tie the word embeddings with the token embeddings.
+        vision_config (`IdeficsVisionConfig` or `dict`, *optional*, defaults to `IdeficsVisionConfig`):
+            Custom vision config or dict for the vision tower
+        text_config (`PretrainedConfig` or `dict`, *optional*, defaults to `LlamaConfig`):
+            Custom text config or dict for the text model
+        scale_factor (`int`, *optional*, defaults to 2):
+            The scale factor for the image encoder.
+        pad_token_id (`int`, *optional*, defaults to 128002):
+            The id of the padding token.
+
+    Example:
+    ```python
+    >>> from transformers import Idefics3Model, Idefics3Config
+    >>> # Initializing configuration
+    >>> configuration = Idefics3Config()
+    >>> # Initializing a model from the configuration
+    >>> model = Idefics3Model(configuration)
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "idefics3bl"
+    sub_configs = {"text_config": AutoConfig, "vision_config": Idefics3blVisionConfig}
+
+    def __init__(
+        self,
+        use_cache=True,
+        image_token_id=128257,
+        tie_word_embeddings=False,
+        vision_config=None,
+        text_config=None,
+        scale_factor=2,
+        pad_token_id=128_002,
+        midblock_ratio=1.0,
+        midblock_start=-1,
+        midblock_end=-1,
+        **kwargs,
+    ):
+        self.image_token_id = image_token_id
+        self.use_cache = use_cache
+        self.tie_word_embeddings = tie_word_embeddings
+
+        if vision_config is None:
+            self.vision_config = Idefics3blVisionConfig()
+            logger.info("vision_config is None, using default vision config")
+        elif isinstance(vision_config, dict):
+            self.vision_config = Idefics3blVisionConfig(**vision_config)
+        elif isinstance(vision_config, Idefics3blVisionConfig):
+            self.vision_config = vision_config
+
+        if isinstance(text_config, dict):
+            text_config["model_type"] = text_config["model_type"] if "model_type" in text_config else "llama"
+            text_config = CONFIG_MAPPING[text_config["model_type"]](**text_config)
+        elif text_config is None:
+            logger.info("text_config is None, using default text config")
+            text_config = CONFIG_MAPPING["llama"](
+                rms_norm_eps=1e-5,
+                pad_token_id=pad_token_id,
+                tie_word_embeddings=False,
+            )
+
+        self.text_config = text_config
+        self.scale_factor = scale_factor
+        
+        self.midblock_ratio = midblock_ratio
+        self.midblock_start = midblock_start
+        self.midblock_end = midblock_end
+
+        super().__init__(**kwargs, pad_token_id=pad_token_id, tie_word_embeddings=tie_word_embeddings)
+
+
+__all__ = ["Idefics3Config", "Idefics3VisionConfig"]

generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "pad_token_id": 0,
+  "transformers_version": "4.53.2"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:394df19eb2b6aa8fc83cd8e60eceb1b72f4d349214b9df66b310ae08174104e0
+size 1313825808

modeling_idefics3bl.py

@@ -0,0 +1,1165 @@
+# coding=utf-8
+# Copyright 2024 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 Idefics3 model."""
+
+from dataclasses import dataclass
+from typing import Callable, Optional, Union
+
+import torch
+import torch.utils.checkpoint
+from torch import nn
+
+from transformers.activations import ACT2FN
+from transformers.cache_utils import DynamicCache
+from transformers.generation import GenerationMixin
+from transformers.modeling_attn_mask_utils import _prepare_4d_attention_mask
+from transformers.modeling_flash_attention_utils import FlashAttentionKwargs
+from transformers.modeling_layers import GradientCheckpointingLayer
+from transformers.modeling_outputs import BaseModelOutput, ModelOutput
+from transformers.modeling_utils import ALL_ATTENTION_FUNCTIONS, PreTrainedModel
+from transformers.processing_utils import Unpack
+from transformers.utils import LossKwargs, auto_docstring, can_return_tuple, logging
+from transformers.models.auto import AutoModel
+from .configuration_idefics3bl import Idefics3blConfig, Idefics3blVisionConfig
+from .modeling_llama import *
+
+logger = logging.get_logger(__name__)
+
+
+@dataclass
+@auto_docstring(
+    custom_intro="""
+    Base class for Idefics3 model's outputs that may also contain a past key/values (to speed up sequential decoding).
+    """
+)
+class Idefics3BaseModelOutputWithPast(ModelOutput):
+    r"""
+    last_hidden_state (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`):
+        Sequence of hidden-states at the output of the last layer of the model.
+        If `past_key_values` is used only the last hidden-state of the sequences of shape `(batch_size, 1,
+        hidden_size)` is output.
+    past_key_values (`tuple(tuple(torch.FloatTensor))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`):
+        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)`) and optionally if
+        `config.is_encoder_decoder=True` 2 additional tensors of shape `(batch_size, num_heads,
+        encoder_sequence_length, embed_size_per_head)`.
+        Contains pre-computed hidden-states (key and values in the self-attention blocks and optionally if
+        `config.is_encoder_decoder=True` in the cross-attention blocks) that can be used (see `past_key_values`
+        input) to speed up sequential decoding.
+    image_hidden_states (`tuple(torch.FloatTensor)`, *optional*):
+        Tuple of `torch.FloatTensor` (one for the output of the image embeddings, `(batch_size, num_images,
+        sequence_length, hidden_size)`.
+        image_hidden_states of the model produced by the vision encoder
+    """
+
+    last_hidden_state: Optional[torch.FloatTensor] = None
+    past_key_values: Optional[tuple[tuple[torch.FloatTensor]]] = None
+    hidden_states: Optional[tuple[torch.FloatTensor]] = None
+    attentions: Optional[tuple[torch.FloatTensor]] = None
+    image_hidden_states: Optional[tuple[torch.FloatTensor]] = None
+
+
+@dataclass
+@auto_docstring(
+    custom_intro="""
+    Base class for Idefics causal language model (or autoregressive) outputs.
+    """
+)
+class Idefics3CausalLMOutputWithPast(ModelOutput):
+    r"""
+    loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided):
+        Language modeling loss (for next-token prediction).
+    logits (`torch.FloatTensor` of shape `(batch_size, sequence_length, config.vocab_size)`):
+        Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
+    past_key_values (`tuple(tuple(torch.FloatTensor))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`):
+        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)`)
+        Contains pre-computed hidden-states (key and values in the self-attention blocks) that can be used (see
+        `past_key_values` input) to speed up sequential decoding.
+    image_hidden_states (`tuple(torch.FloatTensor)`, *optional*):
+        Tuple of `torch.FloatTensor` (one for the output of the image embeddings, `(batch_size, num_images,
+        sequence_length, hidden_size)`.
+        image_hidden_states of the model produced by the vision encoder
+    """
+
+    loss: Optional[torch.FloatTensor] = None
+    logits: Optional[torch.FloatTensor] = None
+    past_key_values: Optional[list[torch.FloatTensor]] = None
+    hidden_states: Optional[tuple[torch.FloatTensor]] = None
+    attentions: Optional[tuple[torch.FloatTensor]] = None
+    image_hidden_states: Optional[tuple[torch.FloatTensor]] = None
+
+
+# Copied from transformers.models.idefics2.modeling_idefics2.Idefics2VisionEmbeddings with Idefics2->Idefics3
+class Idefics3VisionEmbeddings(nn.Module):
+    """
+    This is a modified version of `siglip.modelign_siglip.SiglipVisionEmbeddings` to enable images of variable
+    resolution.
+
+    The modifications are adapted from [Patch n' Pack: NaViT, a Vision Transformer for any Aspect Ratio and Resolution](https://huggingface.co/papers/2307.06304)
+    which allows treating images in their native aspect ratio and without the need to resize them to the same
+    fixed size. In particular, we start from the original pre-trained SigLIP model
+    (which uses images of fixed-size square images) and adapt it by training on images of variable resolutions.
+    """
+
+    def __init__(self, config: Idefics3blVisionConfig):
+        super().__init__()
+        self.embed_dim = config.hidden_size
+        self.image_size = config.image_size
+        self.patch_size = config.patch_size
+
+        self.patch_embedding = nn.Conv2d(
+            in_channels=config.num_channels,
+            out_channels=self.embed_dim,
+            kernel_size=self.patch_size,
+            stride=self.patch_size,
+            padding="valid",
+        )
+
+        self.num_patches_per_side = self.image_size // self.patch_size
+        self.num_patches = self.num_patches_per_side**2
+        self.num_positions = self.num_patches
+        self.position_embedding = nn.Embedding(self.num_positions, self.embed_dim)
+
+    def forward(self, pixel_values: torch.FloatTensor, patch_attention_mask: torch.BoolTensor) -> torch.Tensor:
+        batch_size, _, max_im_h, max_im_w = pixel_values.shape
+
+        patch_embeds = self.patch_embedding(pixel_values)
+        embeddings = patch_embeds.flatten(2).transpose(1, 2)
+
+        max_nb_patches_h, max_nb_patches_w = max_im_h // self.patch_size, max_im_w // self.patch_size
+        boundaries = torch.arange(1 / self.num_patches_per_side, 1.0, 1 / self.num_patches_per_side)
+        position_ids = torch.full(size=(batch_size, max_nb_patches_h * max_nb_patches_w), fill_value=0)
+
+        for batch_idx, p_attn_mask in enumerate(patch_attention_mask):
+            nb_patches_h = p_attn_mask[:, 0].sum()
+            nb_patches_w = p_attn_mask[0].sum()
+
+            fractional_coords_h = torch.arange(0, 1 - 1e-6, 1 / nb_patches_h)
+            fractional_coords_w = torch.arange(0, 1 - 1e-6, 1 / nb_patches_w)
+
+            bucket_coords_h = torch.bucketize(fractional_coords_h, boundaries, right=True)
+            bucket_coords_w = torch.bucketize(fractional_coords_w, boundaries, right=True)
+
+            pos_ids = (bucket_coords_h[:, None] * self.num_patches_per_side + bucket_coords_w).flatten()
+            position_ids[batch_idx][p_attn_mask.view(-1).cpu()] = pos_ids
+
+        position_ids = position_ids.to(self.position_embedding.weight.device)
+        embeddings = embeddings + self.position_embedding(position_ids)
+        return embeddings
+
+
+# Copied from transformers.models.siglip.modeling_siglip.eager_attention_forward
+def eager_attention_forward(
+    module: nn.Module,
+    query: torch.Tensor,
+    key: torch.Tensor,
+    value: torch.Tensor,
+    attention_mask: Optional[torch.Tensor],
+    scaling: float,
+    dropout: float = 0.0,
+    **kwargs,
+):
+    attn_weights = torch.matmul(query, key.transpose(-1, -2)) * scaling
+    if attention_mask is not None:
+        attn_weights = attn_weights + attention_mask
+
+    attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)
+    attn_weights = nn.functional.dropout(attn_weights, p=dropout, training=module.training)
+
+    attn_output = torch.matmul(attn_weights, value)
+    attn_output = attn_output.transpose(1, 2).contiguous()
+
+    return attn_output, attn_weights
+
+
+# Copied from transformers.models.siglip.modeling_siglip.SiglipAttention with Siglip->Idefics3Vision
+class Idefics3blVisionAttention(nn.Module):
+    """Multi-headed attention from 'Attention Is All You Need' paper"""
+
+    # Copied from transformers.models.clip.modeling_clip.CLIPAttention.__init__
+    def __init__(self, config, layer_idx: int = 0):
+        super().__init__()
+        self.config = config
+        self.embed_dim = config.hidden_size
+        
+        self.midblock_start = config.midblock_start
+        self.midblock_end = config.midblock_end
+        self.ratio = config.midblock_ratio if self.midblock_start <= layer_idx < self.midblock_end else 1.0
+        
+        self.num_heads = int(config.num_attention_heads * self.ratio)
+        self.head_dim = self.embed_dim // config.num_attention_heads
+        
+        self.q_out_size = self.num_heads * self.head_dim
+        # 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.dropout = config.attention_dropout
+
+        self.k_proj = nn.Linear(self.embed_dim, self.q_out_size)
+        self.v_proj = nn.Linear(self.embed_dim, self.q_out_size)
+        self.q_proj = nn.Linear(self.embed_dim, self.q_out_size)
+        self.out_proj = nn.Linear(self.q_out_size, self.embed_dim)
+
+        # Ignore copy
+        self.is_causal = False
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        output_attentions: Optional[bool] = False,
+    ) -> tuple[torch.Tensor, Optional[torch.Tensor]]:
+        """Input shape: Batch x Time x Channel"""
+
+        batch_size, seq_length, embed_dim = hidden_states.shape
+
+        queries = self.q_proj(hidden_states)
+        keys = self.k_proj(hidden_states)
+        values = self.v_proj(hidden_states)
+
+        queries = queries.view(batch_size, seq_length, self.num_heads, self.head_dim).transpose(1, 2)
+        keys = keys.view(batch_size, seq_length, self.num_heads, self.head_dim).transpose(1, 2)
+        values = values.view(batch_size, seq_length, self.num_heads, self.head_dim).transpose(1, 2)
+
+        attention_interface: Callable = eager_attention_forward
+        if self.config._attn_implementation != "eager":
+            if self.config._attn_implementation == "sdpa" and output_attentions:
+                logger.warning_once(
+                    "`torch.nn.functional.scaled_dot_product_attention` does not support `output_attentions=True`. Falling back to "
+                    'eager attention. This warning can be removed using the argument `attn_implementation="eager"` when loading the model.'
+                )
+            else:
+                attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]
+
+        attn_output, attn_weights = attention_interface(
+            self,
+            queries,
+            keys,
+            values,
+            attention_mask,
+            is_causal=self.is_causal,
+            scaling=self.scale,
+            dropout=0.0 if not self.training else self.dropout,
+        )
+
+        attn_output = attn_output.reshape(batch_size, seq_length, self.q_out_size).contiguous()
+        attn_output = self.out_proj(attn_output)
+
+        if not output_attentions:
+            attn_weights = None
+
+        return attn_output, attn_weights
+
+
+# Copied from transformers.models.siglip.modeling_siglip.SiglipMLP with Siglip->Idefics3Vision
+class Idefics3blVisionMLP(nn.Module):
+    def __init__(self, config, layer_idx: int = 0):
+        super().__init__()
+        self.config = config
+        self.activation_fn = ACT2FN[config.hidden_act]
+        
+        self.midblock_start = config.midblock_start
+        self.midblock_end = config.midblock_end
+        self.ratio = config.midblock_ratio if self.midblock_start <= layer_idx < self.midblock_end else 1.0
+        
+        self.intermediate_size = int(config.intermediate_size * self.ratio)
+        self.fc1 = nn.Linear(config.hidden_size, self.intermediate_size)
+        self.fc2 = nn.Linear(self.intermediate_size, config.hidden_size)
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        hidden_states = self.fc1(hidden_states)
+        hidden_states = self.activation_fn(hidden_states)
+        hidden_states = self.fc2(hidden_states)
+        return hidden_states
+
+
+class Idefics3SimpleMLP(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        input_size = config.vision_config.hidden_size * (config.scale_factor**2)
+        output_size = config.text_config.hidden_size
+        self.proj = nn.Linear(input_size, output_size, bias=False)
+
+    def forward(self, x):
+        return self.proj(x)
+
+
+# Copied from transformers.models.idefics2.modeling_idefics2.Idefics2EncoderLayer with Idefics2->Idefics3
+class Idefics3blEncoderLayer(GradientCheckpointingLayer):
+    def __init__(self, config: Idefics3blVisionConfig, layer_id: int = 0):
+        super().__init__()
+        self.embed_dim = config.hidden_size
+        self.self_attn = Idefics3blVisionAttention(config, layer_idx=layer_id)
+        self.layer_norm1 = nn.LayerNorm(self.embed_dim, eps=config.layer_norm_eps)
+        self.mlp = Idefics3blVisionMLP(config, layer_idx=layer_id)
+        self.layer_norm2 = nn.LayerNorm(self.embed_dim, eps=config.layer_norm_eps)
+
+    # Copied from transformers.models.siglip.modeling_siglip.SiglipEncoderLayer.forward
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: torch.Tensor,
+        output_attentions: Optional[bool] = False,
+    ) -> tuple[torch.FloatTensor]:
+        """
+        Args:
+            hidden_states (`torch.FloatTensor`):
+                Input to the layer of shape `(batch, seq_len, embed_dim)`.
+            attention_mask (`torch.FloatTensor`):
+                Attention mask of shape `(batch, 1, q_len, k_v_seq_len)` where padding elements are indicated by very large negative values.
+            output_attentions (`bool`, *optional*, defaults to `False`):
+                Whether or not to return the attentions tensors of all attention layers. See `attentions` under
+                returned tensors for more detail.
+        """
+        residual = hidden_states
+
+        hidden_states = self.layer_norm1(hidden_states)
+        hidden_states, attn_weights = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            output_attentions=output_attentions,
+        )
+        hidden_states = residual + hidden_states
+
+        residual = hidden_states
+        hidden_states = self.layer_norm2(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        hidden_states = residual + hidden_states
+
+        outputs = (hidden_states,)
+
+        if output_attentions:
+            outputs += (attn_weights,)
+
+        return outputs
+
+
+# Copied from transformers.models.siglip.modeling_siglip.SiglipEncoder with Siglip->Idefics3
+class Idefics3blEncoder(nn.Module):
+    """
+    Transformer encoder consisting of `config.num_hidden_layers` self attention layers. Each layer is a
+    [`Idefics3EncoderLayer`].
+
+    Args:
+        config: Idefics3Config
+    """
+
+    def __init__(self, config: Idefics3blConfig):
+        super().__init__()
+        self.config = config
+        self.layers = nn.ModuleList([Idefics3blEncoderLayer(config, layer_id) for layer_id in range(config.num_hidden_layers)])
+        self.gradient_checkpointing = False
+
+    # Ignore copy
+    def forward(
+        self,
+        inputs_embeds,
+        attention_mask: Optional[torch.Tensor] = None,
+        output_attentions: Optional[bool] = None,
+        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)`):
+                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.
+            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)
+            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.
+        """
+        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
+
+        encoder_states = () if output_hidden_states else None
+        all_attentions = () if output_attentions else None
+
+        hidden_states = inputs_embeds
+        for encoder_layer in self.layers:
+            if output_hidden_states:
+                encoder_states = encoder_states + (hidden_states,)
+            layer_outputs = encoder_layer(
+                hidden_states,
+                attention_mask,
+                output_attentions=output_attentions,
+            )
+
+            hidden_states = layer_outputs[0]
+
+            if output_attentions:
+                all_attentions = all_attentions + (layer_outputs[1],)
+
+        if output_hidden_states:
+            encoder_states = encoder_states + (hidden_states,)
+
+        if not return_dict:
+            return tuple(v for v in [hidden_states, encoder_states, all_attentions] if v is not None)
+        return BaseModelOutput(
+            last_hidden_state=hidden_states, hidden_states=encoder_states, attentions=all_attentions
+        )
+
+
+# Copied from transformers.models.llama.modeling_llama.repeat_kv
+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)
+
+
+# Copied from transformers.models.llama.modeling_llama.LlamaRMSNorm with Llama->Idefics3
+class Idefics3RMSNorm(nn.Module):
+    def __init__(self, hidden_size, eps=1e-6):
+        """
+        Idefics3RMSNorm 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)
+
+    def extra_repr(self):
+        return f"{tuple(self.weight.shape)}, eps={self.variance_epsilon}"
+
+
+class Idefics3Connector(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.scale_factor = config.scale_factor
+        self.modality_projection = Idefics3SimpleMLP(config)
+
+    def pixel_shuffle(self, x, scale_factor=2):
+        bsz, seq, embed_dim = x.size()
+        height = width = int(seq**0.5)
+        x = x.view(bsz, height, width, embed_dim)
+        x = x.view(bsz, height, int(width / scale_factor), embed_dim * scale_factor)
+        x = x.permute(0, 2, 1, 3)
+        x = x.reshape(bsz, int(width / scale_factor), int(height / scale_factor), embed_dim * (scale_factor**2))
+        x = x.permute(0, 2, 1, 3)
+        x = x.reshape(bsz, int(seq / (scale_factor**2)), embed_dim * (scale_factor**2))
+        return x
+
+    def forward(self, image_hidden_states):
+        image_hidden_states = self.pixel_shuffle(image_hidden_states, self.scale_factor)
+        image_hidden_states = self.modality_projection(image_hidden_states)
+        return image_hidden_states
+
+
+@auto_docstring
+class Idefics3PreTrainedModel(PreTrainedModel):
+    config_class = Idefics3blConfig
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["Idefics3blVisionAttention", "Idefics3blDecoderLayer"]
+    _skip_keys_device_placement = "past_key_values"
+    _supports_flash_attn_2 = True
+    _supports_sdpa = True
+    _supports_flex_attn = True
+    _supports_cache_class = True
+    _supports_attention_backend = True
+
+    def _init_weights(self, module):
+        std = getattr(self.config, "initializer_range", self.config.get_text_config().initializer_range)
+
+        if isinstance(module, (nn.Linear, nn.Conv2d)):
+            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_()
+        elif isinstance(module, nn.LayerNorm):
+            module.weight.data.fill_(1.0)
+            module.bias.data.zero_()
+        elif isinstance(module, Idefics3RMSNorm):
+            module.weight.data.fill_(1.0)
+
+
+@auto_docstring(
+    custom_intro="""
+    The Idefics3 Vision Transformer Model outputting raw image embedding.
+    """
+)
+class Idefics3VisionTransformer(Idefics3PreTrainedModel):
+    config_class = Idefics3blVisionConfig
+    _supports_sdpa = True
+    _supports_flash_attention_2 = True
+    _supports_flex_attn = True
+
+    def __init__(self, config: Idefics3blVisionConfig):
+        super().__init__(config)
+        embed_dim = config.hidden_size
+
+        self.embeddings = Idefics3VisionEmbeddings(config)
+        self.encoder = Idefics3blEncoder(config)
+        self.patch_size = config.patch_size
+        self.post_layernorm = nn.LayerNorm(embed_dim, eps=config.layer_norm_eps)
+        self._use_flash_attention_2 = config._attn_implementation == "flash_attention_2"
+
+    # Copied from transformers.models.idefics2.modeling_idefics2.Idefics2VisionTransformer.get_input_embeddings
+    def get_input_embeddings(self):
+        return self.embeddings
+
+    # Copied from transformers.models.idefics2.modeling_idefics2.Idefics2VisionTransformer.set_input_embeddings
+    def set_input_embeddings(self, value):
+        self.embeddings = value
+
+    def forward(
+        self,
+        pixel_values,
+        patch_attention_mask: Optional[torch.BoolTensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[tuple, BaseModelOutput]:
+        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
+
+        batch_size = pixel_values.size(0)
+        if patch_attention_mask is None:
+            patch_size = self.patch_size
+            patch_attention_mask = torch.ones(
+                (
+                    batch_size,
+                    pixel_values.size(2) // patch_size,
+                    pixel_values.size(3) // patch_size,
+                )
+            )
+            patch_attention_mask = patch_attention_mask.to(dtype=torch.bool, device=pixel_values.device)
+
+        hidden_states = self.embeddings(pixel_values=pixel_values, patch_attention_mask=patch_attention_mask)
+
+        patch_attention_mask = patch_attention_mask.view(batch_size, -1)
+        # The call to `_upad_input` in `_flash_attention_forward` is expensive
+        # So when the `patch_attention_mask` is full of 1s (i.e. attending to the whole sequence),
+        # avoiding passing the attention_mask, which is equivalent to attending to the full sequence
+        if not torch.any(~patch_attention_mask):
+            patch_attention_mask = None
+        elif not self._use_flash_attention_2:
+            patch_attention_mask = _prepare_4d_attention_mask(patch_attention_mask, hidden_states.dtype)
+
+        encoder_outputs = self.encoder(
+            inputs_embeds=hidden_states,
+            attention_mask=patch_attention_mask,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        last_hidden_state = encoder_outputs[0]
+        last_hidden_state = self.post_layernorm(last_hidden_state)
+
+        if not return_dict:
+            return (last_hidden_state,) + encoder_outputs[1:]
+
+        return BaseModelOutput(
+            last_hidden_state=last_hidden_state,
+            hidden_states=encoder_outputs.hidden_states,
+            attentions=encoder_outputs.attentions,
+        )
+
+
+class LlamablAttention(LlamaAttention):
+    """Multi-headed attention from 'Attention Is All You Need' paper"""
+
+    def __init__(self, config: LlamaConfig, layer_idx: int):
+        super().__init__(config, layer_idx=layer_idx)
+        self.config = config
+        self.layer_idx = layer_idx
+        self.head_dim = getattr(config, "head_dim", config.hidden_size // config.num_attention_heads)
+        
+        self.midblock_start = config.midblock_start
+        self.midblock_end = config.midblock_end
+        self.ratio = config.midblock_ratio if self.midblock_start <= layer_idx < self.midblock_end else 1.0
+        
+        self.num_attention_heads = int(config.num_attention_heads * self.ratio)
+        self.num_key_value_heads = int(config.num_key_value_heads * self.ratio)
+
+        self.num_key_value_groups = config.num_attention_heads // config.num_key_value_heads
+        self.scaling = self.head_dim**-0.5
+        self.attention_dropout = config.attention_dropout
+        self.is_causal = True
+
+        self.q_proj = nn.Linear(
+            config.hidden_size, self.num_attention_heads * self.head_dim, bias=config.attention_bias
+        )
+        self.k_proj = nn.Linear(
+            config.hidden_size, self.num_key_value_heads * self.head_dim, bias=config.attention_bias
+        )
+        self.v_proj = nn.Linear(
+            config.hidden_size, self.num_key_value_heads * self.head_dim, bias=config.attention_bias
+        )
+        self.o_proj = nn.Linear(
+            self.num_attention_heads * self.head_dim, config.hidden_size, bias=config.attention_bias
+        )
+
+
+class LlamablMLP(LlamaMLP):
+    def __init__(self, config, layer_idx: int):
+        super().__init__(config)
+        self.config = config
+        self.hidden_size = config.hidden_size
+        
+        self.midblock_start = config.midblock_start
+        self.midblock_end = config.midblock_end
+        self.ratio = 0.5 if self.midblock_start <= layer_idx < self.midblock_end else 1.0
+        
+        self.intermediate_size = int(config.intermediate_size * self.ratio)
+        self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=config.mlp_bias)
+        self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=config.mlp_bias)
+        self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=config.mlp_bias)
+        self.act_fn = ACT2FN[config.hidden_act]
+
+
+class LlamablDecoderLayer(LlamaDecoderLayer):
+    def __init__(self, config: LlamaConfig, layer_idx: int):
+        super().__init__(config, layer_idx=layer_idx)
+
+        self.self_attn = LlamablAttention(config=config, layer_idx=layer_idx)
+        self.mlp = LlamablMLP(config, layer_idx=layer_idx)
+
+
+class LLama3blModel(LlamaModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.midblock_ratio = self.config.midblock_ratio
+        self.midblock_start = self.config.midblock_start
+        self.midblock_end = self.config.midblock_end
+
+        self.layers = nn.ModuleList(
+            [LlamablDecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
+        )
+        # Initialize weights and apply final processing
+        self.post_init()
+        
+
+@auto_docstring(
+    custom_intro="""
+    Idefics3 model consisting of a SIGLIP vision encoder and Llama3 language decoder
+    """
+)
+class Idefics3Model(Idefics3PreTrainedModel):
+    def __init__(self, config: Idefics3blConfig):
+        super().__init__(config)
+        self.padding_idx = self.config.text_config.pad_token_id
+        self.vocab_size = self.config.text_config.vocab_size
+        
+        self.vision_model = Idefics3VisionTransformer._from_config(config.vision_config)
+        self.connector = Idefics3Connector(config)
+        self.text_model = LLama3blModel(config.text_config)
+
+        self.image_seq_len = int(
+            ((config.vision_config.image_size // config.vision_config.patch_size) ** 2) / (config.scale_factor**2)
+        )
+        self.image_token_id = self.config.image_token_id
+
+        self._use_flash_attention_2 = config.text_config._attn_implementation == "flash_attention_2"
+
+        self.post_init()
+
+    # Copied from transformers.models.idefics2.modeling_idefics2.Idefics2Model.enable_input_require_grads
+    def enable_input_require_grads(self):
+        """
+        Enables the gradients for the input embeddings.
+
+        This is useful for lora when using gradient checkpointing.
+        c.f. https://github.com/huggingface/peft/issues/1402#issuecomment-1913675032
+
+        Override to set output.requires_grad = True for both the decoder's and vision model's embeddings.
+        """
+
+        def get_lowest_module(module):
+            if len(list(module.children())) == 0:
+                # If the module has no children, it is a leaf module (e.g., Linear, Conv2d, etc.)
+                return module
+            else:
+                # Recursively call the function on each child module
+                return get_lowest_module(list(module.children())[0])
+
+        def make_inputs_require_grads(module, input, output):
+            output.requires_grad_(True)
+
+        self._text_require_grads_hook = self.get_input_embeddings().register_forward_hook(make_inputs_require_grads)
+        self._vision_require_grads_hook = get_lowest_module(self.vision_model).register_forward_hook(
+            make_inputs_require_grads
+        )
+
+    # Copied from transformers.models.idefics2.modeling_idefics2.Idefics2Model.disable_input_require_grads
+    def disable_input_require_grads(self):
+        self._text_require_grads_hook.remove()
+        self._vision_require_grads_hook.remove()
+
+    # Copied from transformers.models.idefics2.modeling_idefics2.Idefics2Model.get_input_embeddings
+    def get_input_embeddings(self):
+        return self.text_model.get_input_embeddings()
+
+    # Copied from transformers.models.idefics2.modeling_idefics2.Idefics2Model.set_input_embeddings
+    def set_input_embeddings(self, value):
+        self.text_model.set_input_embeddings(value)
+
+    def inputs_merger(
+        self,
+        input_ids: torch.LongTensor,
+        inputs_embeds: Optional[torch.Tensor],
+        image_hidden_states: Optional[torch.Tensor],
+    ):
+        """
+        This method aims at merging the token embeddings with the image hidden states into one single sequence of vectors that are fed to the transformer LM.
+        The merging happens as follows:
+        - The text token sequence is: `tok_1 tok_2 tok_3 <fake_token_around_image> <image> <image> ... <image> <fake_token_around_image> tok_4`.
+        - We get the image hidden states for the image through the vision encoder and that hidden state, after a pixel shuffle operation, is then projected into the text embedding space.
+        We thus have a sequence of image hidden states of size (1, image_seq_len, hidden_dim), where 1 is for batch_size of 1 image and hidden_dim is the hidden_dim of the LM transformer.
+        - The merging happens so that we obtain the following sequence: `vector_tok_1 vector_tok_2 vector_tok_3 vector_fake_tok_around_image {sequence of image_seq_len image hidden states} vector_fake_toke_around_image vector_tok_4`. That sequence is fed to the LM.
+        - To fit the format of that sequence, `input_ids`, `input_embeds`, `attention_mask` are all 3 adapted to insert the image hidden states.
+        """
+        special_image_token_mask = input_ids == self.image_token_id
+        # Fixes RuntimeError: a leaf Variable that requires grad is being used in an in-place operation.
+        new_inputs_embeds = inputs_embeds.clone()
+        # Flatten `image_hidden_states` if not flat yet
+        image_hidden_states = image_hidden_states.view(-1, image_hidden_states.shape[-1])
+        # cast to the dtype of the input_embeds to support quantized models
+        image_hidden_states = image_hidden_states.to(inputs_embeds.device, inputs_embeds.dtype)
+        new_inputs_embeds[special_image_token_mask] = image_hidden_states
+        return new_inputs_embeds
+
+    def get_image_features(self, pixel_values: torch.FloatTensor, pixel_attention_mask: torch.LongTensor = None):
+        """
+        Encodes images into continuous embeddings that can be forwarded to the language model.
+
+        Args:
+            pixel_values (`torch.FloatTensor` of shape `(batch_size, num_channels, image_size, image_size)`):
+                The tensors corresponding to the input images.
+            pixel_attention_mask (`torch.LongTensor`, *optional*):
+                The attention mask indicating padded regions in the image.
+        """
+        batch_size, num_images, num_channels, height, width = pixel_values.shape
+        pixel_values = pixel_values.to(dtype=self.dtype)  # fp16 compatibility
+        pixel_values = pixel_values.view(batch_size * num_images, *pixel_values.shape[2:])
+
+        # Remove padding images - padding images are full 0.
+        nb_values_per_image = pixel_values.shape[1:].numel()
+        real_images_inds = (pixel_values == 0.0).sum(dim=(-1, -2, -3)) != nb_values_per_image
+        pixel_values = pixel_values[real_images_inds].contiguous()
+
+        # Handle the vision attention mask
+        if pixel_attention_mask is None:
+            pixel_attention_mask = torch.ones(
+                size=(pixel_values.size(0), pixel_values.size(2), pixel_values.size(3)),
+                dtype=torch.bool,
+                device=pixel_values.device,
+            )
+        else:
+            # Remove padding images from the mask
+            pixel_attention_mask = pixel_attention_mask.view(batch_size * num_images, *pixel_attention_mask.shape[2:])
+            pixel_attention_mask = pixel_attention_mask[real_images_inds].contiguous()
+
+        patch_size = self.config.vision_config.patch_size
+        patches_subgrid = pixel_attention_mask.unfold(dimension=1, size=patch_size, step=patch_size)
+        patches_subgrid = patches_subgrid.unfold(dimension=2, size=patch_size, step=patch_size)
+        patch_attention_mask = (patches_subgrid.sum(dim=(-1, -2)) > 0).bool()
+
+        # Get sequence from the vision encoder
+        image_hidden_states = self.vision_model(pixel_values=pixel_values, patch_attention_mask=patch_attention_mask)
+        image_hidden_states.last_hidden_state
+
+        # Modality projection & resampling
+        image_hidden_states = self.connector(image_hidden_states.last_hidden_state)
+        return image_hidden_states
+
+    @can_return_tuple
+    @auto_docstring(
+        custom_intro="""
+        Inputs fed to the model can have an arbitrary number of images. To account for this, pixel_values fed to
+        the model have image padding -> (batch_size, max_num_images, 3, max_heights, max_widths) where
+        max_num_images is the maximum number of images among the batch_size samples in the batch.
+        Padding images are not needed beyond padding the pixel_values at the entrance of the model.
+        For efficiency, we only pass through the vision_model's forward the real images by
+        discarding the padding images i.e. pixel_values of size (image_batch_size, 3, height, width) where
+        image_batch_size would be 7 when num_images_per_sample=[1, 3, 1, 2] and max_num_images would be 3.
+        """
+    )
+    def forward(
+        self,
+        input_ids: Optional[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,
+        pixel_values: Optional[torch.FloatTensor] = None,
+        pixel_attention_mask: Optional[torch.BoolTensor] = None,
+        image_hidden_states: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        return_dict: Optional[bool] = None,
+        **kwargs: Unpack[FlashAttentionKwargs],
+    ) -> Union[tuple, Idefics3BaseModelOutputWithPast]:
+        r"""
+        pixel_attention_mask (`torch.Tensor` of shape `(batch_size, image_size, image_size)`, *optional*):
+            Mask to avoid performing attention on padding pixel indices.
+        image_hidden_states (`torch.FloatTensor` of shape `(batch_size, num_channels, image_size, image_size)`):
+            The hidden states of the image encoder after modality projection.
+        """
+        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
+
+        if self.training and self.text_model.gradient_checkpointing and use_cache:
+            logger.warning_once(
+                "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
+            )
+            use_cache = False
+
+        # retrieve input_ids and inputs_embeds
+        if input_ids is not None:
+            batch_size, seq_length = input_ids.shape
+        elif inputs_embeds is not None:
+            batch_size, seq_length, _ = inputs_embeds.shape
+        else:
+            raise ValueError("You have to specify either input_ids or inputs_embeds")
+
+        past_seen_tokens = 0
+        if use_cache:
+            if past_key_values is None:
+                past_key_values = DynamicCache()
+            past_seen_tokens = past_key_values.get_seq_length()
+
+        if inputs_embeds is None:
+            inputs_embeds = self.text_model.get_input_embeddings()(input_ids).to(self.device)
+
+        # START VISUAL INPUTS INTEGRATION
+        if pixel_values is not None and image_hidden_states is not None:
+            raise ValueError("You cannot specify both pixel_values and image_hidden_states at the same time")
+        elif pixel_values is not None:
+            image_hidden_states = self.get_image_features(pixel_values, pixel_attention_mask)
+        elif image_hidden_states is not None:
+            image_hidden_states = image_hidden_states.to(dtype=self.dtype, device=input_ids.device)
+
+        if past_seen_tokens == 0 and input_ids is not None and image_hidden_states is not None:
+            # When we generate, we don't want to replace the potential image_token_id that we generated by images
+            # that simply don't exist
+            inputs_embeds = self.inputs_merger(
+                input_ids=input_ids,
+                inputs_embeds=inputs_embeds,
+                image_hidden_states=image_hidden_states,
+            )
+
+        outputs = self.text_model(
+            inputs_embeds=inputs_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,
+            cache_position=cache_position,
+            return_dict=True,
+            **kwargs,
+        )
+
+        return Idefics3BaseModelOutputWithPast(
+            last_hidden_state=outputs.last_hidden_state,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+            image_hidden_states=image_hidden_states,
+        )
+
+
+class KwargsForCausalLM(FlashAttentionKwargs, LossKwargs): ...
+
+
+@auto_docstring(
+    custom_intro="""
+    The Idefics3 Model with a language modeling head. It is made up a SigLIP vision encoder, with a language modeling head on top.
+    """
+)
+class Idefics3blForConditionalGeneration(Idefics3PreTrainedModel, GenerationMixin):
+    _tied_weights_keys = ["lm_head.weight"]
+
+    # Copied from transformers.models.idefics2.modeling_idefics2.Idefics2ForConditionalGeneration.__init__ with Idefics2->Idefics3
+    def __init__(self, config):
+        super().__init__(config)
+        self.model = Idefics3Model(config)
+        self.image_token_id = self.config.image_token_id
+
+        self.lm_head = nn.Linear(config.text_config.hidden_size, config.text_config.vocab_size, bias=False)
+        self.vocab_size = config.text_config.vocab_size
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    # Copied from transformers.models.idefics2.modeling_idefics2.Idefics2ForConditionalGeneration.enable_input_require_grads
+    def enable_input_require_grads(self):
+        """
+        Enables the gradients for the input embeddings. This is useful for fine-tuning adapter weights while keeping
+        the model weights fixed.
+        """
+
+        def make_inputs_require_grads(module, input, output):
+            output.requires_grad_(True)
+
+        self._text_require_grads_hook = self.get_input_embeddings().register_forward_hook(make_inputs_require_grads)
+        self._vision_require_grads_hook = self.model.vision_model.get_input_embeddings().register_forward_hook(
+            make_inputs_require_grads
+        )
+
+    # Copied from transformers.models.idefics2.modeling_idefics2.Idefics2ForConditionalGeneration.disable_input_require_grads
+    def disable_input_require_grads(self):
+        self._text_require_grads_hook.remove()
+        self._vision_require_grads_hook.remove()
+
+    # Copied from transformers.models.idefics2.modeling_idefics2.Idefics2ForConditionalGeneration.get_input_embeddings
+    def get_input_embeddings(self):
+        return self.model.text_model.get_input_embeddings()
+
+    # Copied from transformers.models.idefics2.modeling_idefics2.Idefics2ForConditionalGeneration.set_input_embeddings
+    def set_input_embeddings(self, value):
+        self.model.text_model.set_input_embeddings(value)
+
+    # Copied from transformers.models.idefics2.modeling_idefics2.Idefics2ForConditionalGeneration.get_output_embeddings
+    def get_output_embeddings(self):
+        return self.lm_head
+
+    # Copied from transformers.models.idefics2.modeling_idefics2.Idefics2ForConditionalGeneration.set_output_embeddings
+    def set_output_embeddings(self, new_embeddings):
+        self.lm_head = new_embeddings
+
+    def get_image_features(self, pixel_values: torch.FloatTensor, pixel_attention_mask: torch.LongTensor = None):
+        return self.model.get_image_features(pixel_values=pixel_values, pixel_attention_mask=pixel_attention_mask)
+
+    @can_return_tuple
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[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,
+        pixel_values: Optional[torch.FloatTensor] = None,
+        pixel_attention_mask: Optional[torch.BoolTensor] = None,
+        image_hidden_states: 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,
+        cache_position: Optional[torch.LongTensor] = None,
+        return_dict: Optional[bool] = None,
+        logits_to_keep: Union[int, torch.Tensor] = 0,
+        **kwargs: Unpack[KwargsForCausalLM],
+    ) -> Union[tuple, Idefics3CausalLMOutputWithPast]:
+        r"""
+        pixel_attention_mask (`torch.Tensor` of shape `(batch_size, image_size, image_size)`, *optional*):
+            Mask to avoid performing attention on padding pixel indices.
+        image_hidden_states (`torch.FloatTensor` of shape `(batch_size, num_channels, image_size, image_size)`):
+            The hidden states of the image encoder after modality projection.
+        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 `model.image_token_id` (where `model` is your instance of `Idefics3ForConditionalGeneration`).
+            Tokens with indices set to `model.image_token_id` are ignored (masked), the loss is only
+            computed for the tokens with labels in `[0, ..., config.vocab_size]`.
+
+        Example:
+
+        ```python
+        >>> import requests
+        >>> import torch
+        >>> from PIL import Image
+        >>> from io import BytesIO
+
+        >>> from transformers import AutoProcessor, AutoModelForVision2Seq
+        >>> from transformers.image_utils import load_image
+
+        >>> # Note that passing the image urls (instead of the actual pil images) to the processor is also possible
+        >>> image1 = load_image("https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg")
+        >>> image2 = load_image("https://cdn.britannica.com/59/94459-050-DBA42467/Skyline-Chicago.jpg")
+        >>> image3 = load_image("https://cdn.britannica.com/68/170868-050-8DDE8263/Golden-Gate-Bridge-San-Francisco.jpg")
+
+        >>> processor = AutoProcessor.from_pretrained("HuggingFaceM4/Idefics3-8B-Llama3")
+        >>> model = AutoModelForVision2Seq.from_pretrained("HuggingFaceM4/Idefics3-8B-Llama3", torch_dtype=torch.bfloat16, device_map="auto")
+
+        >>> # Create inputs
+        >>> messages = [
+        ...     {
+        ...         "role": "user",
+        ...         "content": [
+        ...             {"type": "image"},
+        ...             {"type": "text", "text": "In this image, we can see the city of New York, and more specifically the Statue of Liberty."},
+        ...             {"type": "image"},
+        ...             {"type": "text", "text": "What can we see in this image?"},
+        ...         ]
+        ...     },
+        ...     {
+        ...         "role": "user",
+        ...         "content": [
+        ...             {"type": "image"},
+        ...             {"type": "text", "text": "In which city is that bridge located?"},
+        ...         ]
+        ...     }
+        ... ]
+
+        >>> prompts = [processor.apply_chat_template([message], add_generation_prompt=True) for message in messages]
+        >>> images = [[image1, image2], [image3]]
+        >>> inputs = processor(text=prompts, images=images, padding=True, return_tensors="pt").to(model.device)
+
+        >>> # Generate
+        >>> generated_ids = model.generate(**inputs, max_new_tokens=256)
+        >>> generated_texts = processor.batch_decode(generated_ids, skip_special_tokens=True)
+
+        >>> print(generated_texts[0])
+        Assistant: There are buildings, trees, lights, and water visible in this image.
+
+        >>> print(generated_texts[1])
+        Assistant: The bridge is in San Francisco.
+        ```"""
+        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,
+            pixel_values=pixel_values,
+            pixel_attention_mask=pixel_attention_mask,
+            image_hidden_states=image_hidden_states,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            cache_position=cache_position,
+            return_dict=True,
+            **kwargs,
+        )
+
+        hidden_states = outputs[0]
+        # Only compute necessary logits, and do not upcast them to float if we are not computing the loss
+        slice_indices = slice(-logits_to_keep, None) if isinstance(logits_to_keep, int) else logits_to_keep
+        logits = self.lm_head(hidden_states[:, slice_indices, :])
+
+        loss = None
+        if labels is not None:
+            loss = self.loss_function(
+                logits=logits, labels=labels, vocab_size=self.config.text_config.vocab_size, **kwargs
+            )
+
+        return Idefics3CausalLMOutputWithPast(
+            loss=loss,
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+            image_hidden_states=outputs.image_hidden_states,
+        )
+
+    # Copied from transformers.models.idefics2.modeling_idefics2.Idefics2ForConditionalGeneration.prepare_inputs_for_generation
+    def prepare_inputs_for_generation(
+        self,
+        input_ids,
+        past_key_values=None,
+        attention_mask=None,
+        inputs_embeds=None,
+        cache_position=None,
+        pixel_values=None,
+        pixel_attention_mask=None,
+        image_hidden_states=None,
+        logits_to_keep=None,
+        **kwargs,
+    ):
+        # Overwritten -- there are mutually exclusive inputs (if the logic to make `image_hidden_states` take
+        # precedence is moved to the model, we can remove this fn)
+
+        model_inputs = super().prepare_inputs_for_generation(
+            input_ids,
+            past_key_values=past_key_values,
+            attention_mask=attention_mask,
+            inputs_embeds=inputs_embeds,
+            cache_position=cache_position,
+            pixel_values=pixel_values,
+            pixel_attention_mask=pixel_attention_mask,
+            image_hidden_states=image_hidden_states,
+            logits_to_keep=logits_to_keep,
+            **kwargs,
+        )
+
+        # if `inputs_embeds` are passed, we only want to use them in the 1st generation step
+        # but IDEFICS requires both ids and embeds to be present
+        if inputs_embeds is not None and cache_position[0] == 0:
+            model_inputs["input_ids"] = input_ids
+
+        if image_hidden_states is not None:
+            model_inputs["pixel_values"] = None
+            model_inputs["pixel_attention_mask"] = None
+
+        return model_inputs
+
+    # Copied from transformers.models.idefics2.modeling_idefics2.Idefics2ForConditionalGeneration._update_model_kwargs_for_generation
+    def _update_model_kwargs_for_generation(self, outputs, model_kwargs, is_encoder_decoder, **kwargs):
+        model_kwargs = super()._update_model_kwargs_for_generation(
+            outputs=outputs,
+            model_kwargs=model_kwargs,
+            is_encoder_decoder=is_encoder_decoder,
+            **kwargs,
+        )
+        # Get the precomputed image_hidden_states
+        model_kwargs["image_hidden_states"] = outputs.image_hidden_states
+        return model_kwargs
+
+
+__all__ = ["Idefics3ForConditionalGeneration", "Idefics3PreTrainedModel", "Idefics3Model", "Idefics3VisionTransformer"]

modeling_llama.py

@@ -0,0 +1,816 @@
+# coding=utf-8
+# Copyright 2022 EleutherAI and the HuggingFace Inc. team. All rights reserved.
+#
+# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
+# and OPT implementations in this library. It has been modified from its
+# original forms to accommodate minor architectural differences compared
+# to GPT-NeoX and OPT used by the Meta AI team that trained the model.
+#
+# 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.
+from typing import Callable, Optional, Union
+
+import torch
+import torch.utils.checkpoint
+from torch import nn
+
+from transformers.activations import ACT2FN
+from transformers.cache_utils import Cache, DynamicCache
+from transformers.generation import GenerationMixin
+from transformers.integrations import use_kernel_forward_from_hub
+from transformers.masking_utils import create_causal_mask
+from transformers.modeling_flash_attention_utils import FlashAttentionKwargs
+from transformers.modeling_layers import GradientCheckpointingLayer
+from transformers.modeling_outputs import (
+    BaseModelOutputWithPast,
+    CausalLMOutputWithPast,
+    QuestionAnsweringModelOutput,
+    SequenceClassifierOutputWithPast,
+    TokenClassifierOutput,
+)
+from transformers.modeling_rope_utils import ROPE_INIT_FUNCTIONS, dynamic_rope_update
+from transformers.modeling_utils import ALL_ATTENTION_FUNCTIONS, PreTrainedModel
+from transformers.processing_utils import Unpack
+from transformers.utils import LossKwargs, auto_docstring, can_return_tuple, logging
+from transformers.models.llama.configuration_llama import LlamaConfig
+
+
+logger = logging.get_logger(__name__)
+
+
+@use_kernel_forward_from_hub("RMSNorm")
+class LlamaRMSNorm(nn.Module):
+    def __init__(self, hidden_size, eps=1e-6):
+        """
+        LlamaRMSNorm 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)
+
+    def extra_repr(self):
+        return f"{tuple(self.weight.shape)}, eps={self.variance_epsilon}"
+
+
+class LlamaRotaryEmbedding(nn.Module):
+    def __init__(self, config: LlamaConfig, device=None):
+        super().__init__()
+        # BC: "rope_type" was originally "type"
+        if hasattr(config, "rope_scaling") and config.rope_scaling is not None:
+            self.rope_type = config.rope_scaling.get("rope_type", config.rope_scaling.get("type"))
+        else:
+            self.rope_type = "default"
+        self.max_seq_len_cached = config.max_position_embeddings
+        self.original_max_seq_len = config.max_position_embeddings
+
+        self.config = config
+        self.rope_init_fn = ROPE_INIT_FUNCTIONS[self.rope_type]
+
+        inv_freq, self.attention_scaling = self.rope_init_fn(self.config, device)
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+        self.original_inv_freq = self.inv_freq
+
+    @torch.no_grad()
+    @dynamic_rope_update  # power user: used with advanced RoPE types (e.g. dynamic rope)
+    def forward(self, x, position_ids):
+        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1).to(x.device)
+        position_ids_expanded = position_ids[:, None, :].float()
+
+        device_type = x.device.type if isinstance(x.device.type, str) and x.device.type != "mps" else "cpu"
+        with torch.autocast(device_type=device_type, enabled=False):  # Force float32
+            freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
+            emb = torch.cat((freqs, freqs), dim=-1)
+            cos = emb.cos() * self.attention_scaling
+            sin = emb.sin() * self.attention_scaling
+
+        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
+
+
+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)
+
+
+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 LlamaMLP(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+        self.hidden_size = config.hidden_size
+        self.intermediate_size = config.intermediate_size
+        self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=config.mlp_bias)
+        self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=config.mlp_bias)
+        self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=config.mlp_bias)
+        self.act_fn = ACT2FN[config.hidden_act]
+
+    def forward(self, x):
+        down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
+        return down_proj
+
+
+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)
+
+
+def eager_attention_forward(
+    module: nn.Module,
+    query: torch.Tensor,
+    key: torch.Tensor,
+    value: torch.Tensor,
+    attention_mask: Optional[torch.Tensor],
+    scaling: float,
+    dropout: float = 0.0,
+    **kwargs,
+):
+    key_states = repeat_kv(key, module.num_key_value_groups)
+    value_states = repeat_kv(value, module.num_key_value_groups)
+
+    attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling
+    if attention_mask is not None:
+        causal_mask = attention_mask[:, :, :, : key_states.shape[-2]]
+        attn_weights = attn_weights + causal_mask
+
+    attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)
+    attn_weights = nn.functional.dropout(attn_weights, p=dropout, training=module.training)
+    attn_output = torch.matmul(attn_weights, value_states)
+    attn_output = attn_output.transpose(1, 2).contiguous()
+
+    return attn_output, attn_weights
+
+
+class LlamaAttention(nn.Module):
+    """Multi-headed attention from 'Attention Is All You Need' paper"""
+
+    def __init__(self, config: LlamaConfig, layer_idx: int):
+        super().__init__()
+        self.config = config
+        self.layer_idx = layer_idx
+        self.head_dim = getattr(config, "head_dim", config.hidden_size // config.num_attention_heads)
+        self.num_key_value_groups = config.num_attention_heads // config.num_key_value_heads
+        self.scaling = self.head_dim**-0.5
+        self.attention_dropout = config.attention_dropout
+        self.is_causal = True
+
+        self.q_proj = nn.Linear(
+            config.hidden_size, config.num_attention_heads * self.head_dim, bias=config.attention_bias
+        )
+        self.k_proj = nn.Linear(
+            config.hidden_size, config.num_key_value_heads * self.head_dim, bias=config.attention_bias
+        )
+        self.v_proj = nn.Linear(
+            config.hidden_size, config.num_key_value_heads * self.head_dim, bias=config.attention_bias
+        )
+        self.o_proj = nn.Linear(
+            config.num_attention_heads * self.head_dim, config.hidden_size, bias=config.attention_bias
+        )
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        position_embeddings: tuple[torch.Tensor, torch.Tensor],
+        attention_mask: Optional[torch.Tensor],
+        past_key_value: Optional[Cache] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs: Unpack[FlashAttentionKwargs],
+    ) -> tuple[torch.Tensor, Optional[torch.Tensor], Optional[tuple[torch.Tensor]]]:
+        input_shape = hidden_states.shape[:-1]
+        hidden_shape = (*input_shape, -1, self.head_dim)
+
+        query_states = self.q_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+        key_states = self.k_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+        value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+
+        cos, sin = position_embeddings
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
+
+        if past_key_value is not None:
+            # sin and cos are specific to RoPE models; cache_position needed for the static cache
+            cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
+            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+        attention_interface: Callable = eager_attention_forward
+        if self.config._attn_implementation != "eager":
+            attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]
+
+        attn_output, attn_weights = attention_interface(
+            self,
+            query_states,
+            key_states,
+            value_states,
+            attention_mask,
+            dropout=0.0 if not self.training else self.attention_dropout,
+            scaling=self.scaling,
+            **kwargs,
+        )
+
+        attn_output = attn_output.reshape(*input_shape, -1).contiguous()
+        attn_output = self.o_proj(attn_output)
+        return attn_output, attn_weights
+
+
+class LlamaDecoderLayer(GradientCheckpointingLayer):
+    def __init__(self, config: LlamaConfig, layer_idx: int):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+
+        self.self_attn = LlamaAttention(config=config, layer_idx=layer_idx)
+
+        self.mlp = LlamaMLP(config)
+        self.input_layernorm = LlamaRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.post_attention_layernorm = LlamaRMSNorm(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[Cache] = None,
+        output_attentions: Optional[bool] = False,
+        use_cache: Optional[bool] = False,
+        cache_position: Optional[torch.LongTensor] = None,
+        position_embeddings: Optional[tuple[torch.Tensor, torch.Tensor]] = None,  # necessary, but kept here for BC
+        **kwargs: Unpack[FlashAttentionKwargs],
+    ) -> tuple[torch.FloatTensor, Optional[tuple[torch.FloatTensor, torch.FloatTensor]]]:
+        residual = hidden_states
+        hidden_states = self.input_layernorm(hidden_states)
+
+        # Self Attention
+        hidden_states, self_attn_weights = 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,
+            cache_position=cache_position,
+            position_embeddings=position_embeddings,
+            **kwargs,
+        )
+        hidden_states = residual + hidden_states
+
+        # Fully Connected
+        residual = hidden_states
+        hidden_states = self.post_attention_layernorm(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        hidden_states = residual + hidden_states
+
+        outputs = (hidden_states,)
+        if output_attentions:
+            outputs += (self_attn_weights,)
+
+        return outputs
+
+
+@auto_docstring
+class LlamaPreTrainedModel(PreTrainedModel):
+    config_class = LlamaConfig
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["LlamaDecoderLayer"]
+    _skip_keys_device_placement = ["past_key_values"]
+    _supports_flash_attn_3 = True
+    _supports_flash_attn_2 = True
+    _supports_sdpa = True
+    _supports_flex_attn = True
+    _supports_cache_class = True
+    _supports_quantized_cache = True
+    _supports_static_cache = True
+    _supports_attention_backend = True
+
+    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_()
+        elif isinstance(module, LlamaRMSNorm):
+            module.weight.data.fill_(1.0)
+
+
+@auto_docstring
+class LlamaModel(LlamaPreTrainedModel):
+    def __init__(self, config: LlamaConfig):
+        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.layers = nn.ModuleList(
+            [LlamaDecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
+        )
+        self.norm = LlamaRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.rotary_emb = LlamaRotaryEmbedding(config=config)
+        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
+
+    @can_return_tuple
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        **flash_attn_kwargs: Unpack[FlashAttentionKwargs],
+    ) -> 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
+
+        if (input_ids is None) ^ (inputs_embeds is not None):
+            raise ValueError("You must specify exactly one of input_ids or inputs_embeds")
+
+        if self.gradient_checkpointing and self.training and use_cache:
+            logger.warning_once(
+                "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`."
+            )
+            use_cache = False
+
+        # TODO (joao): remove this exception in v4.56 -- it exists for users that try to pass a legacy cache
+        if not isinstance(past_key_values, (type(None), Cache)):
+            raise ValueError("The `past_key_values` should be either a `Cache` object or `None`.")
+
+        if inputs_embeds is None:
+            inputs_embeds = self.embed_tokens(input_ids)
+
+        if use_cache and past_key_values is None:
+            past_key_values = DynamicCache()
+
+        if cache_position is None:
+            past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
+            cache_position = torch.arange(
+                past_seen_tokens, past_seen_tokens + inputs_embeds.shape[1], device=inputs_embeds.device
+            )
+
+        if position_ids is None:
+            position_ids = cache_position.unsqueeze(0)
+
+        causal_mask = create_causal_mask(
+            config=self.config,
+            input_embeds=inputs_embeds,
+            attention_mask=attention_mask,
+            cache_position=cache_position,
+            past_key_values=past_key_values,
+            position_ids=position_ids,
+        )
+
+        hidden_states = inputs_embeds
+
+        # create position embeddings to be shared across the decoder layers
+        position_embeddings = self.rotary_emb(hidden_states, position_ids)
+
+        # decoder layers
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attns = () if output_attentions else None
+
+        for decoder_layer in self.layers[: self.config.num_hidden_layers]:
+            if output_hidden_states:
+                all_hidden_states += (hidden_states,)
+
+            layer_outputs = decoder_layer(
+                hidden_states,
+                attention_mask=causal_mask,
+                position_ids=position_ids,
+                past_key_value=past_key_values,
+                output_attentions=output_attentions,
+                use_cache=use_cache,
+                cache_position=cache_position,
+                position_embeddings=position_embeddings,
+                **flash_attn_kwargs,
+            )
+
+            hidden_states = layer_outputs[0]
+
+            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,)
+
+        return BaseModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=past_key_values if use_cache else None,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attns,
+        )
+
+
+class KwargsForCausalLM(FlashAttentionKwargs, LossKwargs): ...
+
+
+@auto_docstring
+class LlamaForCausalLM(LlamaPreTrainedModel, GenerationMixin):
+    _tied_weights_keys = ["lm_head.weight"]
+    _tp_plan = {"lm_head": "colwise_rep"}
+    _pp_plan = {"lm_head": (["hidden_states"], ["logits"])}
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.model = LlamaModel(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()
+
+    def get_input_embeddings(self):
+        return self.model.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.model.embed_tokens = value
+
+    def get_output_embeddings(self):
+        return self.lm_head
+
+    def set_output_embeddings(self, new_embeddings):
+        self.lm_head = new_embeddings
+
+    def set_decoder(self, decoder):
+        self.model = decoder
+
+    def get_decoder(self):
+        return self.model
+
+    @can_return_tuple
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Cache] = 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,
+        cache_position: Optional[torch.LongTensor] = None,
+        logits_to_keep: Union[int, torch.Tensor] = 0,
+        **kwargs: Unpack[KwargsForCausalLM],
+    ) -> CausalLMOutputWithPast:
+        r"""
+        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]`.
+
+        Example:
+
+        ```python
+        >>> from transformers import AutoTokenizer, LlamaForCausalLM
+
+        >>> model = LlamaForCausalLM.from_pretrained("meta-llama/Llama-2-7b-hf")
+        >>> tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-7b-hf")
+
+        >>> prompt = "Hey, are you conscious? Can you talk to me?"
+        >>> 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]
+        "Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you."
+        ```"""
+        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
+        )
+
+        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
+        outputs: BaseModelOutputWithPast = 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,
+            cache_position=cache_position,
+            **kwargs,
+        )
+
+        hidden_states = outputs.last_hidden_state
+        # Only compute necessary logits, and do not upcast them to float if we are not computing the loss
+        slice_indices = slice(-logits_to_keep, None) if isinstance(logits_to_keep, int) else logits_to_keep
+        logits = self.lm_head(hidden_states[:, slice_indices, :])
+
+        loss = None
+        if labels is not None:
+            loss = self.loss_function(logits=logits, labels=labels, vocab_size=self.config.vocab_size, **kwargs)
+
+        return CausalLMOutputWithPast(
+            loss=loss,
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+@auto_docstring(
+    custom_intro="""
+    The LLaMa Model transformer with a sequence classification head on top (linear layer).
+
+    [`LlamaForSequenceClassification`] 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).
+    """
+)
+class LlamaForSequenceClassification(LlamaPreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.model = LlamaModel(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
+
+    @can_return_tuple
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Cache] = 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,
+    ) -> 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).
+        """
+
+        transformer_outputs: BaseModelOutputWithPast = 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,
+        )
+        hidden_states = transformer_outputs.last_hidden_state
+        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:
+            last_non_pad_token = -1
+        elif input_ids is not None:
+            # To handle both left- and right- padding, we take the rightmost token that is not equal to pad_token_id
+            non_pad_mask = (input_ids != self.config.pad_token_id).to(logits.device, torch.int32)
+            token_indices = torch.arange(input_ids.shape[-1], device=logits.device, dtype=torch.int32)
+            last_non_pad_token = (token_indices * non_pad_mask).argmax(-1)
+        else:
+            last_non_pad_token = -1
+            logger.warning_once(
+                f"{self.__class__.__name__} will not detect padding tokens in `inputs_embeds`. Results may be "
+                "unexpected if using padding tokens in conjunction with `inputs_embeds.`"
+            )
+
+        pooled_logits = logits[torch.arange(batch_size, device=logits.device), last_non_pad_token]
+
+        loss = None
+        if labels is not None:
+            loss = self.loss_function(logits=logits, labels=labels, pooled_logits=pooled_logits, config=self.config)
+
+        return SequenceClassifierOutputWithPast(
+            loss=loss,
+            logits=pooled_logits,
+            past_key_values=transformer_outputs.past_key_values,
+            hidden_states=transformer_outputs.hidden_states,
+            attentions=transformer_outputs.attentions,
+        )
+
+
+@auto_docstring
+class LlamaForQuestionAnswering(LlamaPreTrainedModel):
+    base_model_prefix = "transformer"
+
+    # Copied from transformers.models.bloom.modeling_bloom.BloomForQuestionAnswering.__init__ with Bloom->Llama
+    def __init__(self, config):
+        super().__init__(config)
+        self.transformer = LlamaModel(config)
+        self.qa_outputs = nn.Linear(config.hidden_size, 2)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.transformer.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.transformer.embed_tokens = value
+
+    @can_return_tuple
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        start_positions: Optional[torch.LongTensor] = None,
+        end_positions: Optional[torch.LongTensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        **kwargs,
+    ) -> QuestionAnsweringModelOutput:
+        outputs: BaseModelOutputWithPast = self.transformer(
+            input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+        )
+
+        sequence_output = outputs.last_hidden_state
+
+        logits = self.qa_outputs(sequence_output)
+        start_logits, end_logits = logits.split(1, dim=-1)
+        start_logits = start_logits.squeeze(-1).contiguous()
+        end_logits = end_logits.squeeze(-1).contiguous()
+
+        loss = None
+        if start_positions is not None and end_positions is not None:
+            loss = self.loss_function(start_logits, end_logits, start_positions, end_positions, **kwargs)
+
+        return QuestionAnsweringModelOutput(
+            loss=loss,
+            start_logits=start_logits,
+            end_logits=end_logits,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+@auto_docstring
+class LlamaForTokenClassification(LlamaPreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.model = LlamaModel(config)
+        if getattr(config, "classifier_dropout", None) is not None:
+            classifier_dropout = config.classifier_dropout
+        elif getattr(config, "hidden_dropout", None) is not None:
+            classifier_dropout = config.hidden_dropout
+        else:
+            classifier_dropout = 0.1
+        self.dropout = nn.Dropout(classifier_dropout)
+        self.score = nn.Linear(config.hidden_size, config.num_labels)
+
+        # 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
+
+    @can_return_tuple
+    @auto_docstring
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Cache] = 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,
+    ) -> 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).
+        """
+
+        outputs: BaseModelOutputWithPast = 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,
+        )
+        sequence_output = outputs.last_hidden_state
+        sequence_output = self.dropout(sequence_output)
+        logits = self.score(sequence_output)
+
+        loss = None
+        if labels is not None:
+            loss = self.loss_function(logits, labels, self.config)
+
+        return TokenClassifierOutput(
+            loss=loss,
+            logits=logits,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )