Initial commit

README.md

@@ -0,0 +1,140 @@
+---
+library_name: transformers
+tags:
+- vlm
+- image-text-to-text
+- multimodal
+- pretraining
+license: apache-2.0
+language:
+- en
+pipeline_tag: image-text-to-text
+---
+
+# Dream-VL 7B
+
+Dream-VL 7B is an open diffusion vision-language model trained on 12M multimodal data from the [MAmmoTH-VL-Instruct-12M](https://huggingface.co/datasets/MAmmoTH-VL/MAmmoTH-VL-Instruct-12M) dataset. 
+The model takes language instructions and images as input and generates language outputs.
+
+All Dream-VL checkpoints, as well as our [training codebase](https://github.com/DreamLM/DreamVLX) are released under an Apache 2.0 License.
+
+For full details, please read [our blog](https://hkunlp.github.io/blog/2025/dream-vlx/) and paper (pending).
+
+## Model Summary
+
+- **Model type:** Vision-language (language, image => language)
+- **Language(s) (NLP):** en
+- **License:** apache-2.0
+- **Finetuned from:** [`Dream-7B`](https://huggingface.co/Dream-org/Dream-v0-Instruct-7B), with Qwen2ViT Vision Backbone.
+- **Pretraining Dataset:** [MAmmoTH-VL-Instruct-12M](https://huggingface.co/datasets/MAmmoTH-VL/MAmmoTH-VL-Instruct-12M).
+- **Repository:** [https://github.com/DreamLM/DreamVLX](https://github.com/DreamLM/DreamVLX)
+- **Project Page & Videos:** [https://hkunlp.github.io/blog/2025/dream-vlx](https://hkunlp.github.io/blog/2025/dream-vlx/)
+
+## Getting Started
+
+```python
+import torch
+from transformers import AutoProcessor, AutoModel
+
+model_name = "Dream-org/Dream-VL-7B"
+
+model = AutoModel.from_pretrained(
+    model_name,
+    torch_dtype=torch.bfloat16,
+    trust_remote_code=True,
+).to('cuda')
+
+processor = AutoProcessor.from_pretrained(
+    model_name,
+    trust_remote_code=True
+)
+
+####### Method 1
+from PIL import Image
+import requests
+url = "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg"
+image = Image.open(requests.get(url, stream=True).raw)
+messages = [
+    {
+        "role": "user","content": [{"type": "image"}, {"type": "text", "text": "Describe this image"}]
+    }
+]
+text = processor.apply_chat_template(
+    messages, tokenize=False, add_generation_prompt=True
+)
+print(text)
+inputs = processor(
+    text=[text], images=[image], padding=True, return_tensors="pt"
+)
+
+####### Method 2: use qwen_vl_utils
+# messages = [
+#     {
+#         "role": "user",
+#         "content": [
+#             {
+#                 "type": "image",
+#                 "image": "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen-VL/assets/demo.jpeg",
+#             },
+#             {"type": "text", "text": "Describe this image."},
+#         ],
+#     }
+# ]
+# text = processor.apply_chat_template(
+#     messages, tokenize=False, add_generation_prompt=True
+# )
+# from qwen_vl_utils import process_vision_info
+# image_inputs, video_inputs = process_vision_info(messages)
+# inputs = processor(
+#     text=[text],
+#     images=image_inputs,
+#     videos=video_inputs,
+#     padding=True,
+#     return_tensors="pt",
+# )
+
+inputs = inputs.to("cuda")
+input_ids = inputs.pop("input_ids")
+output = model.diffusion_generate(
+    input_ids,
+    max_new_tokens=128,
+    output_history=True,
+    return_dict_in_generate=True,
+    steps=128,
+    temperature=0.1,
+    top_p=1,
+    alg="maskgit_plus",
+    alg_temp=0,
+    use_cache=False,
+    **inputs
+)
+
+generations = [
+    processor.tokenizer.decode(g[len(p):].cpu().tolist())
+    for p, g in zip(input_ids, output.sequences)
+]
+
+for j in range(len(messages)):
+    print("output:", j, generations[j].split(processor.tokenizer.eos_token)[0])
+
+
+# output: The image depicts a serene beach scene featuring a young woman and a golden retriever.
+# The woman, dressed in a plaid shirt and dark pants, is seated on the sandy shore, smiling warmly at the camera.
+# The golden retriever, adorned with a colorful harness, sits attentively beside her, its gaze fixed on the woman.
+# The background reveals the vast expanse of the ocean, with waves gently kissing the shore. The sky above is a clear blue, suggesting a sunny day.
+# The overall atmosphere exudes a sense of peace and companionship between the woman and her dog.
+
+```
+
+## Citation
+
+**BibTeX:**
+
+```bibtex
+@article{ye2025dreamvla,
+  title={Dream-VL & Dream-VLA: Open Vision-Language and Vision-Language-Action Models with Diffusion Language Model Backbone},
+  author={Ye, Jiacheng and Gong, Shansan and Gao, Jiahui and Fan, Junming and Wu, Shuang and Bi, Wei and Bai, Haoli and Shang, Lifeng and Kong, Lingpeng},
+  journal={arXiv preprint},
+  year={2025}
+}
+```

added_tokens.json

@@ -0,0 +1,26 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|beginoftext|>": 151665,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|mask|>": 151666,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

chat_template.json

@@ -0,0 +1,3 @@
+{
+  "chat_template": "{% set image_count = namespace(value=0) %}{% set video_count = namespace(value=0) %}{% for message in messages %}{% if loop.first and message['role'] != 'system' %}<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n{% endif %}<|im_start|>{{ message['role'] }}\n{% if message['content'] is string %}{{ message['content'] }}<|im_end|>\n{% else %}{% for content in message['content'] %}{% if content['type'] == 'image' or 'image' in content or 'image_url' in content %}{% set image_count.value = image_count.value + 1 %}{% if add_vision_id %}Picture {{ image_count.value }}: {% endif %}<|image_pad|>{% elif content['type'] == 'video' or 'video' in content %}{% set video_count.value = video_count.value + 1 %}{% if add_vision_id %}Video {{ video_count.value }}: {% endif %}<|video_pad|>{% elif 'text' in content %}{{ content['text'] }}{% endif %}{% endfor %}<|im_end|>\n{% endif %}{% endfor %}{% if add_generation_prompt %}<|im_start|>assistant\n{% endif %}"
+}

config.json

@@ -0,0 +1,62 @@
+{
+  "architectures": [
+    "DreamVLModel"
+  ],
+  "attention_dropout": 0.0,
+  "auto_map": {
+    "AutoConfig": "configuration_dreamvl.DreamVLConfig",
+    "AutoModel": "modeling_dreamvl.DreamVLModel"
+  },
+  "bos_token_id": 151643,
+  "eos_token_id": 151643,
+  "full_attn_mask": true,
+  "hidden_act": "silu",
+  "hidden_size": 3584,
+  "image_token_id": 151655,
+  "initializer_range": 0.02,
+  "intermediate_size": 18944,
+  "mask_token_id": 151666,
+  "max_position_embeddings": 131072,
+  "max_window_layers": 28,
+  "model_type": "dream-vl",
+  "mrope_section": [
+    16,
+    24,
+    24
+  ],
+  "num_attention_heads": 28,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 4,
+  "pad_token_id": 151643,
+  "projector_hidden_act": "gelu",
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000.0,
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.51.3",
+  "use_cache": false,
+  "use_sliding_window": false,
+  "video_token_id": 151656,
+  "vision_config": {
+    "depth": 32,
+    "embed_dim": 1280,
+    "hidden_act": "quick_gelu",
+    "hidden_size": 3584,
+    "in_channels": 3,
+    "in_chans": 3,
+    "initializer_range": 0.02,
+    "mlp_ratio": 4,
+    "model_type": "dream_vl",
+    "num_heads": 16,
+    "patch_size": 14,
+    "spatial_merge_size": 2,
+    "spatial_patch_size": 14,
+    "temporal_patch_size": 2
+  },
+  "vision_end_token_id": 151653,
+  "vision_start_token_id": 151652,
+  "vision_token_id": 151654,
+  "vocab_size": 152064
+}

configuration_dreamvl.py

@@ -0,0 +1,142 @@
+# coding=utf-8
+# Copyright 2024 The DreamVL team and the HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""DreamVL model configuration"""
+
+import os
+from typing import Union
+
+from transformers.configuration_utils import PretrainedConfig
+from transformers.modeling_rope_utils import rope_config_validation
+from transformers.utils import logging
+
+
+logger = logging.get_logger("DreamVL."+__name__)
+
+class DreamVLVisionConfig(PretrainedConfig):
+    model_type = "dream_vl"
+    base_config_key = "vision_config"
+
+    def __init__(
+        self,
+        depth=32,
+        embed_dim=1280,
+        hidden_size=3584,
+        hidden_act="quick_gelu",
+        mlp_ratio=4,
+        num_heads=16,
+        in_channels=3,
+        patch_size=14,
+        spatial_merge_size=2,
+        temporal_patch_size=2,
+        initializer_range=0.02,
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+
+        self.depth = depth
+        self.embed_dim = embed_dim
+        self.hidden_size = hidden_size
+        self.hidden_act = hidden_act
+        self.mlp_ratio = mlp_ratio
+        self.num_heads = num_heads
+        self.in_channels = in_channels
+        self.patch_size = patch_size
+        self.spatial_merge_size = spatial_merge_size
+        self.temporal_patch_size = temporal_patch_size
+        self.initializer_range = initializer_range
+
+
+class DreamVLConfig(PretrainedConfig):
+    model_type = "dream-vl"
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    def __init__(
+        self,
+        vocab_size=151936,
+        hidden_size=4096,
+        intermediate_size=22016,
+        num_hidden_layers=32,
+        num_attention_heads=32,
+        num_key_value_heads=32,
+        hidden_act="silu",
+        max_position_embeddings=32768,
+        initializer_range=0.02,
+        image_token_id = 151655,
+        video_token_id = 151656,
+        vision_end_token_id = 151653,
+        vision_start_token_id = 151652,
+        vision_token_id = 151654,
+        rms_norm_eps=1e-6,
+        use_cache=False, 
+        tie_word_embeddings=False,
+        rope_theta=10000.0,
+        use_sliding_window=False,
+        sliding_window=4096,
+        max_window_layers=28,
+        attention_dropout=0.0,
+        mask_token_id=151666,
+        pad_token_id=151643,
+        vision_config=None,
+        rope_scaling=None,
+        mrope_section=[16,24,24],
+        projector_hidden_act=None,
+        **kwargs,
+    ):
+        if isinstance(vision_config, dict):
+            self.vision_config = DreamVLVisionConfig(**vision_config)
+        elif vision_config is None:
+            self.vision_config = DreamVLVisionConfig()
+
+        self.vocab_size = vocab_size
+        self.max_position_embeddings = max_position_embeddings
+        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.use_sliding_window = use_sliding_window
+        self.sliding_window = sliding_window if use_sliding_window else None
+        self.max_window_layers = max_window_layers
+        self.projector_hidden_act = projector_hidden_act
+
+        # for backward compatibility
+        if num_key_value_heads is None:
+            num_key_value_heads = num_attention_heads
+
+        self.num_key_value_heads = num_key_value_heads
+        self.hidden_act = hidden_act
+        self.initializer_range = initializer_range
+        self.rms_norm_eps = rms_norm_eps
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.rope_scaling = rope_scaling
+        self.attention_dropout = attention_dropout
+        # Validate the correctness of rotary position embeddings parameters
+        # BC: if there is a 'type' field, move it to 'rope_type'.
+        if self.rope_scaling is not None and "type" in self.rope_scaling:
+            self.rope_scaling["rope_type"] = self.rope_scaling["type"]
+        rope_config_validation(self, ignore_keys={"mrope_section"})
+        self.mrope_section = mrope_section
+        
+        super().__init__(
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )
+        self.mask_token_id = mask_token_id
+        self.pad_token_id = pad_token_id
+        self.image_token_id = image_token_id
+        self.video_token_id = video_token_id
+        self.vision_end_token_id = vision_end_token_id
+        self.vision_start_token_id = vision_start_token_id
+        self.vision_token_id = vision_token_id

generation_config.json

@@ -0,0 +1,8 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 151643,
+  "eos_token_id": 151643,
+  "pad_token_id": 151643,
+  "transformers_version": "4.51.3",
+  "use_cache": false
+}

generation_utils.py

@@ -0,0 +1,573 @@
+# coding=utf-8
+# Copyright 2024 The DreamVL team and the HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import warnings
+import copy
+from dataclasses import dataclass
+from typing import Any, Dict, Optional, Tuple, Union
+
+import torch
+import torch.distributions as dists
+from torch.nn import functional as F
+from transformers import __version__
+from transformers.generation.configuration_utils import (
+    GenerationConfig,
+)
+from transformers.utils import (
+    ModelOutput,
+    is_torchdynamo_compiling,
+    logging,
+)
+from transformers.cache_utils import (
+    Cache,
+    DynamicCache,
+)
+from transformers.generation.utils import GenerationMixin
+from transformers import TextIteratorStreamer
+
+logger = logging.get_logger("DreamVL."+__name__)
+
+class FullSequenceStreamer(TextIteratorStreamer):
+    def __init__(self, tokenizer, **kwargs):
+        super().__init__(tokenizer, **kwargs)
+
+    def put(self, value, stream_end=False):
+        # Assume full token_ids are passed in every time
+        decoded = self.tokenizer.batch_decode(value, **self.decode_kwargs)
+        self.text_queue.put(decoded)
+        if stream_end:
+            self.text_queue.put(self.stop_signal, timeout=self.timeout)
+
+    def end(self):
+        self.text_queue.put(self.stop_signal, timeout=self.timeout)
+
+def top_p_logits(logits, top_p=None):
+    sorted_logits, sorted_indices = torch.sort(logits, descending=True)
+    cumulative_probs = torch.cumsum(F.softmax(sorted_logits, dim=-1), dim=-1)
+    sorted_indices_to_remove = cumulative_probs > top_p
+    # Shift the indices to the right to keep the first token above the threshold
+    sorted_indices_to_remove[..., 1:] = sorted_indices_to_remove[..., :-1].clone()
+    sorted_indices_to_remove[..., 0] = 0
+
+    mask = torch.zeros_like(logits, dtype=torch.bool, device=logits.device)
+    mask = mask.scatter_(-1, sorted_indices, sorted_indices_to_remove)
+    logits = logits.masked_fill(mask, torch.finfo(logits.dtype).min)
+    return logits
+
+def top_k_logits(logits, top_k=None):
+    top_k = min(top_k, logits.size(-1))  # Safety check
+    # Remove all tokens with a probability less than the last token of the top-k
+    indices_to_remove = logits < torch.topk(logits, top_k)[0][..., -1, None]
+    logits = logits.masked_fill(indices_to_remove, torch.finfo(logits.dtype).min)
+    return logits
+
+
+def sample_tokens(logits, temperature=0.0, top_p=None, top_k=None, margin_confidence=False, neg_entropy=False):
+
+    if temperature > 0:
+        logits = logits / temperature
+    if top_p is not None and top_p < 1:
+        logits = top_p_logits(logits, top_p)
+    if top_k is not None:
+        logits = top_k_logits(logits, top_k)
+    probs = torch.softmax(logits, dim=-1)
+
+    if temperature > 0:
+        try:
+            x0 = dists.Categorical(probs=probs).sample()
+            confidence = torch.gather(probs, -1, x0.unsqueeze(-1)).squeeze(-1)
+        except:
+            confidence, x0 = probs.max(dim=-1)
+    else:
+        confidence, x0 = probs.max(dim=-1)
+    
+    if margin_confidence:
+        sorted_probs, _ = torch.sort(probs, dim=-1, descending=True)
+        # Extract top1 and top2 probabilities
+        top1_probs = sorted_probs[:, 0] 
+        top2_probs = sorted_probs[:, 1] 
+        # Calculate confidence as top1 - top2
+        confidence = top1_probs - top2_probs 
+    
+    if neg_entropy:
+        epsilon = 1e-10
+        log_probs = torch.log(probs + epsilon)
+        confidence = torch.sum(probs * log_probs, dim=-1)
+    
+    return confidence, x0
+
+
+@dataclass
+class DreamVLModelOutput(ModelOutput):
+    sequences: torch.LongTensor = None
+    history: Optional[Tuple[torch.FloatTensor]] = None
+
+
+class DreamVLGenerationConfig(GenerationConfig):
+    def __init__(self, **kwargs):
+        # cache parameter
+        self.use_cache: bool = kwargs.pop("use_cache", False)
+        # general generation parameter
+        self.temperature: float = kwargs.pop("temperature", 0.0)
+        self.top_p: Optional[float] = kwargs.pop("top_p", None)
+        self.top_k: Optional[int] = kwargs.pop("top_k", None)
+        self.max_length = kwargs.pop("max_length", 20)
+        self.max_new_tokens = kwargs.pop("max_new_tokens", None)
+        # diffusion specific params
+        self.eps: float = kwargs.pop("eps", 1e-3)
+        self.steps: int = kwargs.pop("steps", 512)
+        self.alg: str = kwargs.pop("alg", 'origin')
+        self.alg_temp: Optional[float] = kwargs.pop("alg_temp", None)
+        self.eos_penalty: Optional[float] = kwargs.pop("eos_penalty", 0)
+
+        # Parameters that define the output variables of `generate`
+        self.num_return_sequences: int = kwargs.pop("num_return_sequences", 1)
+        self.return_dict_in_generate: bool = kwargs.pop("return_dict_in_generate", False)
+        self.output_history: bool = kwargs.pop("output_history", False)
+
+        # Special tokens that can be used at generation time
+        self.mask_token_id = kwargs.pop("mask_token_id", None)
+        self.pad_token_id = kwargs.pop("pad_token_id", None)
+        self.bos_token_id = kwargs.pop("bos_token_id", None)
+        self.eos_token_id = kwargs.pop("eos_token_id", None)
+
+        # Wild card
+        self.generation_kwargs = kwargs.pop("generation_kwargs", {})
+
+        # The remaining attributes do not parametrize `.generate()`, but are informative and/or used by the hub
+        # interface.
+        self._from_model_config = kwargs.pop("_from_model_config", False)
+        self._commit_hash = kwargs.pop("_commit_hash", None)
+        self.transformers_version = kwargs.pop("transformers_version", __version__)
+
+        # Additional attributes without default values
+        if not self._from_model_config:
+            # we don't want to copy values from the model config if we're initializing a `GenerationConfig` from a
+            # model's default configuration file
+            for key, value in kwargs.items():
+                try:
+                    setattr(self, key, value)
+                except AttributeError as err:
+                    logger.error(f"Can't set {key} with value {value} for {self}")
+                    raise err
+
+        # Validate the values of the attributes
+        self.validate(is_init=True)
+
+    def validate(self, is_init=False):
+        pass
+
+class DreamVLGenerationMixin:
+    @staticmethod
+    def _expand_inputs_for_generation(
+        expand_size: int = 1,
+        input_ids: Optional[torch.LongTensor] = None,
+        **model_kwargs
+    ) -> Tuple[torch.LongTensor, Dict[str, Any]]:
+        """Expands tensors from [batch_size, ...] to [batch_size * expand_size, ...]"""
+        pixel_values = model_kwargs.get("pixel_values", None)
+        image_grid_thw = model_kwargs.get("image_grid_thw", None)
+        if expand_size == 1:
+            return GenerationMixin._expand_inputs_for_generation(
+                expand_size=expand_size,
+                input_ids=input_ids,
+                **model_kwargs
+            )
+        elif pixel_values is None and image_grid_thw is None:
+            return GenerationMixin._expand_inputs_for_generation(
+                expand_size=expand_size,
+                input_ids=input_ids,
+                **model_kwargs
+            )
+        else:
+            raise ValueError(
+                "Does not support expansion for image inputs. "
+            )
+
+    def _validate_generated_length(self, generation_config, input_ids_length, has_default_max_length):
+        """Performs validation related to the resulting generated length"""
+
+        # Can't throw warnings/exceptions during compilation
+        if is_torchdynamo_compiling():
+            return
+
+        # 1. Max length warnings related to poor parameterization
+        if has_default_max_length and generation_config.max_new_tokens is None and generation_config.max_length == 20:
+            # 20 is the default max_length of the generation config
+            logger.warning_once(
+                f"Using the model-agnostic default `max_length` (={generation_config.max_length}) to control the "
+                "generation length. We recommend setting `max_new_tokens` to control the maximum length of the "
+                "generation."
+            )
+        if input_ids_length >= generation_config.max_length:
+            input_ids_string = "input_ids"
+            raise ValueError(
+                f"Input length of {input_ids_string} is {input_ids_length}, but `max_length` is set to"
+                f" {generation_config.max_length}. This can lead to unexpected behavior. You should consider"
+                " increasing `max_length` or, better yet, setting `max_new_tokens`."
+            )
+
+    def _prepare_generated_length(
+        self,
+        generation_config,
+        has_default_max_length,
+        input_ids_length,
+    ):
+        """Prepared max and min length in generation configs to avoid clashes between similar attributes"""
+
+        if generation_config.max_new_tokens is not None:
+            if not has_default_max_length and generation_config.max_length is not None:
+                logger.warning_once(
+                    f"Both `max_new_tokens` (={generation_config.max_new_tokens}) and `max_length`(="
+                    f"{generation_config.max_length}) seem to have been set. `max_new_tokens` will take precedence. "
+                    "Please refer to the documentation for more information. "
+                    "(https://huggingface.co/docs/transformers/main/en/main_classes/text_generation)"
+                )
+            generation_config.max_length = generation_config.max_new_tokens + input_ids_length
+
+        elif has_default_max_length:
+            if generation_config.max_length == DreamVLGenerationConfig().max_length:
+                generation_config.max_length = generation_config.max_length + input_ids_length
+                max_position_embeddings = getattr(self.config, "max_position_embeddings", None)
+                if max_position_embeddings is not None:
+                    generation_config.max_length = min(generation_config.max_length, max_position_embeddings)
+
+        return generation_config
+
+    def _prepare_generation_config(
+        self, generation_config: Optional[DreamVLGenerationConfig], **kwargs: Dict
+    ) -> DreamVLGenerationConfig:
+        """
+        Prepares the base generation config, then applies any generation configuration options from kwargs. This
+        function handles retrocompatibility with respect to configuration files.
+        """
+        # priority: `generation_config` argument > `model.generation_config` (the default generation config)
+        using_model_generation_config = False
+        if generation_config is None:
+            generation_config = DreamVLGenerationConfig.from_model_config(self.config)
+            using_model_generation_config = True
+
+        # `torch.compile` can't compile `copy.deepcopy`, arguments in `kwargs` that are part of `generation_config`
+        # will mutate the object with `.update`. As such, passing these arguments through `kwargs` is disabled -- an
+        # exception will be raised in `_validate_model_kwargs`
+        if not is_torchdynamo_compiling():
+            generation_config = copy.deepcopy(generation_config)
+            model_kwargs = generation_config.update(**kwargs)
+            # If `generation_config` is provided, let's fallback ALL special tokens to the default values for the model
+            if not using_model_generation_config:
+                if generation_config.bos_token_id is None:
+                    generation_config.bos_token_id = self.generation_config.bos_token_id
+                if generation_config.eos_token_id is None:
+                    generation_config.eos_token_id = self.generation_config.eos_token_id
+                if generation_config.pad_token_id is None:
+                    generation_config.pad_token_id = self.generation_config.pad_token_id
+                if generation_config.mask_token_id is None:
+                    generation_config.mask_token_id = self.generation_config.mask_token_id
+
+        return generation_config, model_kwargs
+
+    def _prepare_special_tokens(
+        self,
+        generation_config: DreamVLGenerationConfig,
+        device: Optional[Union[torch.device, str]] = None,
+    ):
+        """
+        Prepares the special tokens for generation, overwriting the generation config with their processed versions
+        converted to tensor.
+        Note that `generation_config` is changed in place and stops being serializable after this method is called.
+        That is no problem if called within `generate` (`generation_config` is a local copy that doesn't leave the
+        function). However, if called outside `generate`, consider creating a copy of `generation_config` first.
+        """
+
+        # Convert special tokens to tensors
+        def _tensor_or_none(token, device=None):
+            if token is None:
+                return token
+
+            device = device if device is not None else self.device
+            if isinstance(token, torch.Tensor):
+                return token.to(device)
+            return torch.tensor(token, device=device, dtype=torch.long)
+
+        bos_token_tensor = _tensor_or_none(generation_config.bos_token_id, device=device)
+        eos_token_tensor = _tensor_or_none(generation_config.eos_token_id, device=device)
+        pad_token_tensor = _tensor_or_none(generation_config.pad_token_id, device=device)
+        mask_token_tensor = _tensor_or_none(generation_config.mask_token_id, device=device)
+
+        # We can have more than one eos token. Always treat it as a 1D tensor (when it exists).
+        if eos_token_tensor is not None and eos_token_tensor.ndim == 0:
+            eos_token_tensor = eos_token_tensor.unsqueeze(0)
+
+        # Set pad token if unset (and there are conditions to do so)
+        if pad_token_tensor is None and eos_token_tensor is not None:
+            pad_token_tensor = eos_token_tensor[0]
+            logger.warning_once(f"Setting `pad_token_id` to `eos_token_id`:{pad_token_tensor} for open-end generation.")
+
+        # Update generation config with the updated special tokens tensors
+        # NOTE: this must be written into a different attribute name than the one holding the original special tokens
+        # (in their non-tensor form), in order to enable end-to-end compilation. See
+        # https://pytorch.org/docs/stable/torch.compiler_cudagraph_trees.html#limitations
+        generation_config._bos_token_tensor = bos_token_tensor
+        generation_config._eos_token_tensor = eos_token_tensor
+        generation_config._pad_token_tensor = pad_token_tensor
+        generation_config._mask_token_tensor = mask_token_tensor
+
+    def _mask_pad_inputs_for_generation(
+        self,
+        input_ids: torch.LongTensor,
+        generation_config: DreamVLGenerationConfig,
+        **model_kwargs,
+    ) -> Tuple[torch.LongTensor, Dict[str, Any]]:
+        """
+        pad tokens in the input ids and attentions for generation. This is used to insert mask tokens into the input_ids
+        """
+        max_length = generation_config.max_length
+        mask_token_id = generation_config.mask_token_id
+        attention_mask = model_kwargs.get("attention_mask", None)
+
+        # pad input_ids to max_length
+        input_ids = F.pad(input_ids, (0, max_length - input_ids.shape[1]), value=mask_token_id)
+        if attention_mask is not None:
+            attention_mask = F.pad(attention_mask, (0, max_length - attention_mask.shape[1]), value=1.0)
+            model_kwargs["attention_mask"] = attention_mask
+        else:
+            raise ValueError(
+                "attention_mask should be provided. "
+            )
+
+        return input_ids, model_kwargs
+
+    def _update_model_kwargs_for_generation(
+        self,
+        outputs: ModelOutput,
+        model_kwargs: Dict[str, Any]
+    ) -> Dict[str, Any]:
+        # update past_key_values keeping its naming used in model code
+        if model_kwargs["use_cache"]:
+            assert outputs.past_key_values is not None, "Cache should not be None if use_cache is True"
+            assert outputs.past_key_values.get_seq_length() == model_kwargs["total_sequence_length"], \
+                f"Cache length {outputs.past_key_values.get_seq_length()} should be equal to the total sequence length {model_kwargs['total_sequence_length']}"
+            # The crop operation requires "left padding for batch processing"
+            outputs.past_key_values.crop(max_length = model_kwargs["prompt_length"])
+            # if model_kwargs["past_key_values"].get_seq_length() > 0:
+            #     assert self.compare_past_key_values(model_kwargs["past_key_values"], outputs.past_key_values), \
+            #         f"Cache {model_kwargs['past_key_values']} should be equal to the new cache {outputs.past_key_values}"
+        else:
+            assert outputs.past_key_values is None, "Cache should be None if use_cache is False"
+        model_kwargs["past_key_values"] = outputs.past_key_values
+
+        # update cache position
+        if model_kwargs["use_cache"]:
+            model_kwargs["cache_position"] = model_kwargs["cache_position"][-(model_kwargs["total_sequence_length"] - model_kwargs["prompt_length"]):]
+        else:
+            assert model_kwargs["cache_position"] is None, "Cache position should be None if use_cache is False"
+
+        if model_kwargs.get("rope_deltas", None) is not None:
+            assert torch.equal(
+                model_kwargs["rope_deltas"], outputs.rope_deltas), \
+                f"Rope deltas {model_kwargs['rope_deltas']} should be equal to the new rope deltas {outputs.rope_deltas}"
+        model_kwargs["rope_deltas"] = outputs.rope_deltas
+        return model_kwargs
+
+    @torch.no_grad()
+    def diffusion_generate(
+        self,
+        inputs: Optional[torch.Tensor] = None,
+        generation_config: Optional[DreamVLGenerationConfig] = None,
+        streamer: Optional[FullSequenceStreamer]=None,
+        **kwargs,
+    ) -> Union[DreamVLModelOutput, torch.LongTensor]:
+        # 1. Handle `generation_config` and kwargs that might update it, and validate the `.generate()` call
+        generation_config, model_kwargs = self._prepare_generation_config(generation_config, **kwargs)
+        generation_tokens_hook_func = model_kwargs.pop("generation_tokens_hook_func", lambda step, x, logits: x)
+        generation_logits_hook_func = model_kwargs.pop("generation_logits_hook_func", lambda step, x, logits: logits)
+        attention_mask = kwargs.pop("attention_mask", None)
+
+        # 2. Define model inputs
+        assert inputs is not None
+        input_ids = inputs
+        device = input_ids.device
+        self._prepare_special_tokens(generation_config, device=device)
+
+        # 3. Prepare `max_length`.
+        input_ids_length = input_ids.shape[-1]
+        has_default_max_length = kwargs.get("max_length") is None and generation_config.max_length is not None
+        generation_config = self._prepare_generated_length(
+            generation_config=generation_config,
+            has_default_max_length=has_default_max_length,
+            input_ids_length=input_ids_length,
+        )
+
+        self._validate_generated_length(generation_config, input_ids_length, has_default_max_length)
+        
+        # 4. Check input_ids
+        if not is_torchdynamo_compiling() and self.device.type != input_ids.device.type:
+            logger.warning_once(
+                "You are calling .generate() with the `input_ids` being on a device type different"
+                f" than your model's device. `input_ids` is on {input_ids.device.type}, whereas the model"
+                f" is on {self.device.type}. You may experience unexpected behaviors or slower generation."
+                " Please make sure that you have put `input_ids` to the"
+                f" correct device by calling for example input_ids = input_ids.to('{self.device.type}') before"
+                " running `.generate()`."
+            )
+        if (
+            hasattr(generation_config, "pad_token_id") and
+            torch.any(input_ids == generation_config.pad_token_id) and 
+            attention_mask is None
+        ):
+            logger.warning_once(
+                "Padding was detected but no attention mask is passed here. For correct "
+                "generation results, please set `attention_mask` when batch-padding inputs."
+            )
+
+        # 5. initialize kv cache
+        model_kwargs["use_cache"] = generation_config.use_cache
+        if model_kwargs["use_cache"]:
+            model_kwargs["past_key_values"] = DynamicCache()
+            model_kwargs["prompt_length"] = input_ids.shape[1] - 1
+        else:
+            model_kwargs["past_key_values"] = None
+            model_kwargs["prompt_length"] = input_ids.shape[1]  - 1
+
+        # 6. Expand inputs for generation
+        input_ids, model_kwargs = self._expand_inputs_for_generation(
+            input_ids=input_ids,
+            expand_size=generation_config.num_return_sequences,
+            **model_kwargs,
+        )
+
+        # 7. pad mask for generation
+        input_ids, model_kwargs = self._mask_pad_inputs_for_generation(
+            input_ids=input_ids,
+            generation_config=generation_config,
+            **model_kwargs,
+        )
+        model_kwargs["total_sequence_length"] = input_ids.shape[1]               
+
+        # 8. initialize cache position
+        if model_kwargs["use_cache"]:
+            model_kwargs["cache_position"] = torch.ones_like(input_ids[0, :], dtype=torch.int64).cumsum(0) - 1
+        else:
+            model_kwargs["cache_position"] = None
+        # 9. Generate
+        result = self._sample(
+            input_ids,
+            generation_config=generation_config,
+            generation_tokens_hook_func=generation_tokens_hook_func,
+            generation_logits_hook_func=generation_logits_hook_func,
+            streamer = streamer,
+            **model_kwargs,
+        )
+        return result
+
+    def _sample(
+        self,
+        input_ids: torch.LongTensor,
+        generation_config: DreamVLGenerationConfig,
+        generation_tokens_hook_func,
+        generation_logits_hook_func,
+        streamer: Optional[FullSequenceStreamer] = None,
+        **model_kwargs,
+    ) -> Union[DreamVLModelOutput, torch.LongTensor]:
+        # init values
+        output_history = generation_config.output_history
+        return_dict_in_generate = generation_config.return_dict_in_generate
+        max_length = generation_config.max_length
+        mask_token_id = generation_config.mask_token_id
+        pad_token_id = generation_config.pad_token_id
+        steps = generation_config.steps
+        eps = generation_config.eps
+        alg = generation_config.alg
+        alg_temp = generation_config.alg_temp
+        temperature = generation_config.temperature
+        eos_penalty = generation_config.eos_penalty
+        top_p = generation_config.top_p
+        top_k = generation_config.top_k
+
+        histories = [] if (return_dict_in_generate and output_history) else None
+
+        timesteps = torch.linspace(1, eps, steps + 1, device=input_ids.device)
+
+        x = generation_tokens_hook_func(None, input_ids, None)
+
+        # this allows user-defined token control of the intermediate steps
+        for i in range(steps):
+            model_inputs = self.prepare_inputs_for_generation(x, **model_kwargs)
+            x = model_inputs.pop("input_ids").clone()
+            mask_index = (x == mask_token_id)
+            outputs = self(x, **model_inputs)
+            
+            if 'inputs_embeds' not in model_kwargs:
+                # initialize the inputs_embeds for caching
+                model_kwargs['inputs_embeds'] = outputs.inputs_embeds
+
+            model_kwargs = self._update_model_kwargs_for_generation(outputs, model_kwargs)
+            
+            logits = outputs.logits
+            assert torch.all(x[:,0] != mask_token_id), "The first token should not be a mask token"
+            logits = torch.cat([logits[:,:1], logits[:, :-1]], dim=1)
+
+            # this allows user-defined logits control of the intermediate steps
+            logits = generation_logits_hook_func(i, x, logits)
+
+            mask_logits = logits[mask_index]
+            t = timesteps[i]
+            s = timesteps[i + 1]
+            mask_logits[:,pad_token_id] += eos_penalty * torch.log(1-t+eps)
+
+            if alg == 'origin':
+                p_transfer = 1 - s / t if i < steps - 1 else 1
+                x0 = torch.zeros_like(x[mask_index], device=self.device, dtype=torch.long) + mask_token_id
+                transfer_index_t_s = torch.rand(*x0.shape, device=self.device) < p_transfer
+                _, x0[transfer_index_t_s]= sample_tokens(mask_logits[transfer_index_t_s], temperature=temperature, top_p=top_p, top_k=top_k)
+                x[mask_index] = x0.clone()
+            else:
+                if alg == 'maskgit_plus':
+                    confidence, x0 = sample_tokens(mask_logits, temperature=temperature, top_p=top_p, top_k=top_k)
+                elif alg == 'topk_margin':
+                    confidence, x0 = sample_tokens(mask_logits, temperature=temperature, top_p=top_p, top_k=top_k, margin_confidence=True)
+                elif alg == 'entropy':
+                    confidence, x0 = sample_tokens(mask_logits, temperature, top_p=top_p, top_k=top_k, neg_entropy=True)
+                else:
+                    raise RuntimeError(f"Unknown alg: {alg}")
+                num_mask_token = mask_index.sum()
+                number_transfer_tokens = int(num_mask_token * (1 - s / t)) if i < steps - 1 else num_mask_token
+                if number_transfer_tokens > 0:
+                    if alg_temp is None or alg_temp == 0:
+                        _, transfer_index = torch.topk(confidence, number_transfer_tokens)
+                    else:
+                        confidence = confidence / alg_temp
+                        confidence = F.softmax(confidence, dim=-1)
+                        transfer_index = torch.multinomial(confidence, num_samples=number_transfer_tokens)
+                    x0_ = torch.zeros_like(x0, device=self.device, dtype=torch.long) + mask_token_id
+                    x0_[transfer_index] = x0[transfer_index].clone()
+                    x[mask_index] = x0_
+
+            # this allows user-defined token control of the intermediate steps
+            x = generation_tokens_hook_func(i, x, logits)
+
+            if histories is not None:
+                histories.append(x.clone())
+
+            ## update inputs_embeds of all the mask tokens where some are just unmasked
+            model_kwargs['inputs_embeds'][mask_index] = self.get_input_embeddings()(x[mask_index])
+
+        if return_dict_in_generate:
+            return DreamVLModelOutput(
+                sequences=x,
+                history=histories,
+            )
+        else:
+            return x

image_processing_dreamvl.py

@@ -0,0 +1,469 @@
+# coding=utf-8
+# Copyright 2024 The Qwen team, Alibaba Group 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.
+"""Image processor class for Dream-VL."""
+
+import math
+from typing import Dict, List, Optional, Union
+
+import numpy as np
+
+from transformers.image_processing_utils import BaseImageProcessor, BatchFeature
+from transformers.image_transforms import (
+    convert_to_rgb,
+    resize,
+    to_channel_dimension_format,
+)
+from transformers.image_utils import (
+    OPENAI_CLIP_MEAN,
+    OPENAI_CLIP_STD,
+    ChannelDimension,
+    ImageInput,
+    PILImageResampling,
+    VideoInput,
+    get_image_size,
+    infer_channel_dimension_format,
+    is_scaled_image,
+    is_valid_image,
+    make_list_of_images,
+    to_numpy_array,
+    valid_images,
+    validate_preprocess_arguments,
+)
+from transformers.utils import TensorType, is_vision_available, logging
+
+logger = logging.get_logger(__name__)
+
+if is_vision_available():
+    from PIL import Image
+
+
+def make_batched_images(images) -> List[List[ImageInput]]:
+    """
+    Accepts images in list or nested list format, and makes a list of images for preprocessing.
+
+    Args:
+        images (`Union[List[List[ImageInput]], List[ImageInput], ImageInput]`):
+            The input image.
+
+    Returns:
+        list: A list of images.
+    """
+    if isinstance(images, (list, tuple)) and isinstance(images[0], (list, tuple)) and is_valid_image(images[0][0]):
+        return [img for img_list in images for img in img_list]
+
+    elif isinstance(images, (list, tuple)) and is_valid_image(images[0]):
+        return images
+
+    elif is_valid_image(images):
+        return [images]
+
+    raise ValueError(f"Could not make batched images from {images}")
+
+
+# Copied from transformers.models.emova_next_video.image_processing_emova_next_video.make_batched_videos
+def make_batched_videos(videos) -> List[VideoInput]:
+    if isinstance(videos, (list, tuple)) and isinstance(videos[0], (list, tuple)) and is_valid_image(videos[0][0]):
+        return videos
+
+    elif isinstance(videos, (list, tuple)) and is_valid_image(videos[0]):
+        if isinstance(videos[0], Image.Image):
+            return [videos]
+        elif len(videos[0].shape) == 4:
+            return [list(video) for video in videos]
+
+    elif is_valid_image(videos) and len(videos.shape) == 4:
+        return [list(videos)]
+
+    raise ValueError(f"Could not make batched video from {videos}")
+
+
+def smart_resize(
+        height: int, width: int, factor: int = 28, min_pixels: int = 56 * 56, max_pixels: int = 14 * 14 * 4 * 1280
+):
+    """Rescales the image so that the following conditions are met:
+
+    1. Both dimensions (height and width) are divisible by 'factor'.
+
+    2. The total number of pixels is within the range ['min_pixels', 'max_pixels'].
+
+    3. The aspect ratio of the image is maintained as closely as possible.
+
+    """
+    if height < factor or width < factor:
+        # print("height, width", height, width)
+        if height < width:
+            h_bar = factor
+            w_bar = round(width / height * factor)
+        else:
+            h_bar = round(height / width * factor)
+            w_bar = factor
+        # print("h_bar, w_bar", h_bar, w_bar)
+        height, width = h_bar, w_bar
+        # raise ValueError(f"height:{height} or width:{width} must be larger than factor:{factor}")
+    elif max(height, width) / min(height, width) > 200:
+        raise ValueError(
+            f"absolute aspect ratio must be smaller than 200, got {max(height, width) / min(height, width)}"
+        )
+    h_bar = round(height / factor) * factor
+    w_bar = round(width / factor) * factor
+    if h_bar * w_bar > max_pixels:
+        beta = math.sqrt((height * width) / max_pixels)
+        h_bar = math.floor(height / beta / factor) * factor
+        w_bar = math.floor(width / beta / factor) * factor
+    elif h_bar * w_bar < min_pixels:
+        beta = math.sqrt(min_pixels / (height * width))
+        h_bar = math.ceil(height * beta / factor) * factor
+        w_bar = math.ceil(width * beta / factor) * factor
+    return h_bar, w_bar
+
+
+class DreamVLImageProcessor(BaseImageProcessor):
+    r"""
+    Constructs a Dream-VL image processor that dynamically resizes images based on the original images.
+
+    Args:
+        do_resize (`bool`, *optional*, defaults to `True`):
+            Whether to resize the image's (height, width) dimensions.
+        resample (`PILImageResampling`, *optional*, defaults to `Resampling.BICUBIC`):
+            Resampling filter to use when resizing the image.
+        do_rescale (`bool`, *optional*, defaults to `True`):
+            Whether to rescale the image by the specified scale `rescale_factor`.
+        rescale_factor (`int` or `float`, *optional*, defaults to `1/255`):
+            Scale factor to use if rescaling the image.
+        do_normalize (`bool`, *optional*, defaults to `True`):
+            Whether to normalize the image.
+        image_mean (`float` or `List[float]`, *optional*, defaults to `[0.48145466, 0.4578275, 0.40821073]`):
+            Mean to use if normalizing the image. This is a float or list of floats for each channel in the image.
+        image_std (`float` or `List[float]`, *optional*, defaults to `[0.26862954, 0.26130258, 0.27577711]`):
+            Standard deviation to use if normalizing the image. This is a float or list of floats for each channel in the image.
+        do_convert_rgb (`bool`, *optional*, defaults to `True`):
+            Whether to convert the image to RGB.
+        min_pixels (`int`, *optional*, defaults to `56 * 56`):
+            The min pixels of the image to resize the image.
+        max_pixels (`int`, *optional*, defaults to `28 * 28 * 1280`):
+            The max pixels of the image to resize the image.
+        patch_size (`int`, *optional*, defaults to 14):
+            The spacial patch size of the vision encoder.
+        temporal_patch_size (`int`, *optional*, defaults to 2):
+            The temporal patch size of the vision encoder.
+        merge_size (`int`, *optional*, defaults to 2):
+            The merge size of the vision encoder to llm encoder.
+    """
+
+    model_input_names = ["pixel_values", "image_grid_thw", "pixel_values_videos", "video_grid_thw"]
+
+    def __init__(
+            self,
+            do_resize: bool = True,
+            resample: PILImageResampling = PILImageResampling.BICUBIC,
+            do_rescale: bool = True,
+            rescale_factor: Union[int, float] = 1 / 255,
+            do_normalize: bool = True,
+            image_mean: Optional[Union[float, List[float]]] = None,
+            image_std: Optional[Union[float, List[float]]] = None,
+            do_convert_rgb: bool = True,
+            min_pixels: int = 56 * 56,
+            max_pixels: int = 28 * 28 * 1280,
+            patch_size: int = 14,
+            temporal_patch_size: int = 2,
+            merge_size: int = 2,
+            **kwargs,
+    ) -> None:
+        super().__init__(**kwargs)
+        self.do_resize = do_resize
+        self.resample = resample
+        self.do_rescale = do_rescale
+        self.rescale_factor = rescale_factor
+        self.do_normalize = do_normalize
+        self.image_mean = image_mean if image_mean is not None else OPENAI_CLIP_MEAN
+        self.image_std = image_std if image_std is not None else OPENAI_CLIP_STD
+        self.min_pixels = min_pixels
+        self.max_pixels = max_pixels
+        self.patch_size = patch_size
+        self.temporal_patch_size = temporal_patch_size
+        self.merge_size = merge_size
+        self.size = {"min_pixels": min_pixels, "max_pixels": max_pixels}
+        self.do_convert_rgb = do_convert_rgb
+
+    def _preprocess(
+            self,
+            images: Union[ImageInput, VideoInput],
+            do_resize: bool = None,
+            resample: PILImageResampling = None,
+            do_rescale: bool = None,
+            rescale_factor: float = None,
+            do_normalize: bool = None,
+            image_mean: Optional[Union[float, List[float]]] = None,
+            image_std: Optional[Union[float, List[float]]] = None,
+            do_convert_rgb: bool = None,
+            data_format: Optional[ChannelDimension] = ChannelDimension.FIRST,
+            input_data_format: Optional[Union[str, ChannelDimension]] = None,
+    ):
+        """
+        Preprocess an image or batch of images. Copy of the `preprocess` method from `CLIPImageProcessor`.
+
+        Args:
+            images (`ImageInput`):
+                Image or batch of images to preprocess. Expects pixel values ranging from 0 to 255. If pixel values range from 0 to 1, set `do_rescale=False`.
+            vision_info (`List[Dict]`, *optional*):
+                Optional list of dictionaries containing additional information about vision inputs.
+            do_resize (`bool`, *optional*, defaults to `self.do_resize`):
+                Whether to resize the image.
+            resample (`PILImageResampling`, *optional*, defaults to `self.resample`):
+                Resampling filter to use if resizing the image. This can be one of the `PILImageResampling` enums.
+            do_rescale (`bool`, *optional*, defaults to `self.do_rescale`):
+                Whether to rescale the image.
+            rescale_factor (`float`, *optional*, defaults to `self.rescale_factor`):
+                Scale factor to use if rescaling the image.
+            do_normalize (`bool`, *optional*, defaults to `self.do_normalize`):
+                Whether to normalize the image.
+            image_mean (`float` or `List[float]`, *optional*, defaults to `self.image_mean`):
+                Mean to use if normalizing the image. Can be a float or a list of floats corresponding to the number of channels in the image.
+            image_std (`float` or `List[float]`, *optional*, defaults to `self.image_std`):
+                Standard deviation to use if normalizing the image. Can be a float or a list of floats corresponding to the number of channels in the image.
+            do_convert_rgb (`bool`, *optional*, defaults to `self.do_convert_rgb`):
+                Whether to convert the image to RGB.
+            data_format (`ChannelDimension`, *optional*, defaults to `ChannelDimension.FIRST`):
+                The channel dimension format for the output image. Can be one of:
+                - `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format.
+                - `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format.
+                - Unset: Use the channel dimension format of the input image.
+            input_data_format (`ChannelDimension` or `str`, *optional*):
+                The channel dimension format for the input image. Can be one of:
+                - `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format.
+                - `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format.
+                - `"none"` or `ChannelDimension.NONE`: image in (height, width) format.   - `"none"` or `ChannelDimension.NONE`: image in (height, width) format.
+        """
+        # import pdb; pdb.set_trace()
+        # print("images", images)
+        # for image in images:
+        #     print("image", image.size)
+        images = make_list_of_images(images)
+
+        if do_convert_rgb:
+            images = [convert_to_rgb(image) for image in images]
+
+        # All transformations expect numpy arrays.
+        images = [to_numpy_array(image) for image in images]
+
+        if is_scaled_image(images[0]) and do_rescale:
+            logger.warning_once(
+                "It looks like you are trying to rescale already rescaled images. If the input"
+                " images have pixel values between 0 and 1, set `do_rescale=False` to avoid rescaling them again."
+            )
+        if input_data_format is None:
+            # We assume that all images have the same channel dimension format.
+            input_data_format = infer_channel_dimension_format(images[0])
+
+        height, width = get_image_size(images[0], channel_dim=input_data_format)
+        resized_height, resized_width = height, width
+        processed_images = []
+        for image in images:
+            if do_resize:
+                resized_height, resized_width = smart_resize(
+                    height,
+                    width,
+                    factor=self.patch_size * self.merge_size,
+                    min_pixels=self.min_pixels,
+                    max_pixels=self.max_pixels,
+                )
+                image = resize(
+                    image, size=(resized_height, resized_width), resample=resample, input_data_format=input_data_format
+                )
+
+            if do_rescale:
+                image = self.rescale(image, scale=rescale_factor, input_data_format=input_data_format)
+
+            if do_normalize:
+                image = self.normalize(
+                    image=image, mean=image_mean, std=image_std, input_data_format=input_data_format
+                )
+
+            image = to_channel_dimension_format(image, data_format, input_channel_dim=input_data_format)
+            processed_images.append(image)
+
+        patches = np.array(processed_images)
+        if data_format == ChannelDimension.LAST:
+            patches = patches.transpose(0, 3, 1, 2)
+        if patches.shape[0] == 1:
+            patches = np.tile(patches, (self.temporal_patch_size, 1, 1, 1))
+        channel = patches.shape[1]
+        grid_t = patches.shape[0] // self.temporal_patch_size
+        grid_h, grid_w = resized_height // self.patch_size, resized_width // self.patch_size
+        patches = patches.reshape(
+            grid_t,
+            self.temporal_patch_size,
+            channel,
+            grid_h // self.merge_size,
+            self.merge_size,
+            self.patch_size,
+            grid_w // self.merge_size,
+            self.merge_size,
+            self.patch_size,
+        )
+        patches = patches.transpose(0, 3, 6, 4, 7, 2, 1, 5, 8)
+        flatten_patches = patches.reshape(
+            grid_t * grid_h * grid_w, channel * self.temporal_patch_size * self.patch_size * self.patch_size
+        )
+
+        return flatten_patches, (grid_t, grid_h, grid_w)
+
+    def preprocess(
+            self,
+            images: ImageInput,
+            videos: VideoInput = None,
+            do_resize: bool = None,
+            size: Dict[str, int] = None,
+            resample: PILImageResampling = None,
+            do_rescale: bool = None,
+            rescale_factor: float = None,
+            do_normalize: bool = None,
+            image_mean: Optional[Union[float, List[float]]] = None,
+            image_std: Optional[Union[float, List[float]]] = None,
+            do_convert_rgb: bool = None,
+            return_tensors: Optional[Union[str, TensorType]] = None,
+            data_format: Optional[ChannelDimension] = ChannelDimension.FIRST,
+            input_data_format: Optional[Union[str, ChannelDimension]] = None,
+    ):
+        """
+        Args:
+            images (`ImageInput`):
+                Image to preprocess. Expects a single or batch of images with pixel values ranging from 0 to 255. If
+                passing in images with pixel values between 0 and 1, set `do_rescale=False`.
+            videos (`VideoInput`):
+                Video to preprocess. Expects a single or batch of videos with pixel values ranging from 0 to 255. If
+                passing in videos with pixel values between 0 and 1, set `do_rescale=False`.
+            do_resize (`bool`, *optional*, defaults to `self.do_resize`):
+                Whether to resize the image.
+            size (`Dict[str, int]`, *optional*, defaults to `self.size`):
+                Size of the image after resizing. Shortest edge of the image is resized to size["shortest_edge"], with
+                the longest edge resized to keep the input aspect ratio.
+            resample (`int`, *optional*, defaults to `self.resample`):
+                Resampling filter to use if resizing the image. This can be one of the enum `PILImageResampling`. Only
+                has an effect if `do_resize` is set to `True`.
+            do_rescale (`bool`, *optional*, defaults to `self.do_rescale`):
+                Whether to rescale the image.
+            rescale_factor (`float`, *optional*, defaults to `self.rescale_factor`):
+                Rescale factor to rescale the image by if `do_rescale` is set to `True`.
+            do_normalize (`bool`, *optional*, defaults to `self.do_normalize`):
+                Whether to normalize the image.
+            image_mean (`float` or `List[float]`, *optional*, defaults to `self.image_mean`):
+                Image mean to use for normalization. Only has an effect if `do_normalize` is set to `True`.
+            image_std (`float` or `List[float]`, *optional*, defaults to `self.image_std`):
+                Image standard deviation to use for normalization. Only has an effect if `do_normalize` is set to
+                `True`.
+            do_convert_rgb (`bool`, *optional*, defaults to `self.do_convert_rgb`):
+                Whether to convert the image to RGB.
+            return_tensors (`str` or `TensorType`, *optional*):
+                The type of tensors to return. Can be one of:
+                - Unset: Return a list of `np.ndarray`.
+                - `TensorType.TENSORFLOW` or `'tf'`: Return a batch of type `tf.Tensor`.
+                - `TensorType.PYTORCH` or `'pt'`: Return a batch of type `torch.Tensor`.
+                - `TensorType.NUMPY` or `'np'`: Return a batch of type `np.ndarray`.
+                - `TensorType.JAX` or `'jax'`: Return a batch of type `jax.numpy.ndarray`.
+            data_format (`ChannelDimension` or `str`, *optional*, defaults to `ChannelDimension.FIRST`):
+                The channel dimension format for the output image. Can be one of:
+                - `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format.
+                - `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format.
+                - Unset: Use the channel dimension format of the input image.
+            input_data_format (`ChannelDimension` or `str`, *optional*):
+                The channel dimension format for the input image. If unset, the channel dimension format is inferred
+                from the input image. Can be one of:
+                - `"channels_first"` or `ChannelDimension.FIRST`: image in (num_channels, height, width) format.
+                - `"channels_last"` or `ChannelDimension.LAST`: image in (height, width, num_channels) format.
+                - `"none"` or `ChannelDimension.NONE`: image in (height, width) format.
+
+        """
+        do_resize = do_resize if do_resize is not None else self.do_resize
+        size = size if size is not None else self.size
+        resample = resample if resample is not None else self.resample
+        do_rescale = do_rescale if do_rescale is not None else self.do_rescale
+        rescale_factor = rescale_factor if rescale_factor is not None else self.rescale_factor
+        do_normalize = do_normalize if do_normalize is not None else self.do_normalize
+        image_mean = image_mean if image_mean is not None else self.image_mean
+        image_std = image_std if image_std is not None else self.image_std
+        do_convert_rgb = do_convert_rgb if do_convert_rgb is not None else self.do_convert_rgb
+
+        if images is not None:
+            images = make_batched_images(images)
+        if videos is not None:
+            videos = make_batched_videos(videos)
+
+        if images is not None and not valid_images(images):
+            raise ValueError(
+                "Invalid image type. Must be of type PIL.Image.Image, numpy.ndarray, "
+                "torch.Tensor, tf.Tensor or jax.ndarray."
+            )
+
+        validate_preprocess_arguments(
+            rescale_factor=rescale_factor,
+            do_normalize=do_normalize,
+            image_mean=image_mean,
+            image_std=image_std,
+            do_resize=do_resize,
+            size=size,
+            resample=resample,
+        )
+
+        if images is not None:
+            pixel_values, vision_grid_thws = [], []
+            for image in images:
+                patches, image_grid_thw = self._preprocess(
+                    image,
+                    do_resize=do_resize,
+                    resample=resample,
+                    do_rescale=do_rescale,
+                    rescale_factor=rescale_factor,
+                    do_normalize=do_normalize,
+                    image_mean=image_mean,
+                    image_std=image_std,
+                    data_format=data_format,
+                    do_convert_rgb=do_convert_rgb,
+                    input_data_format=input_data_format,
+                )
+                pixel_values.extend(patches)
+                vision_grid_thws.append(image_grid_thw)
+            pixel_values = np.array(pixel_values)
+            vision_grid_thws = np.array(vision_grid_thws)
+            data = {"pixel_values": pixel_values, "image_grid_thw": vision_grid_thws}
+
+        if videos is not None:
+            pixel_values, vision_grid_thws = [], []
+            for images in videos:
+                patches, video_grid_thw = self._preprocess(
+                    images,
+                    do_resize=do_resize,
+                    resample=resample,
+                    do_rescale=do_rescale,
+                    rescale_factor=rescale_factor,
+                    do_normalize=do_normalize,
+                    image_mean=image_mean,
+                    image_std=image_std,
+                    data_format=data_format,
+                    do_convert_rgb=do_convert_rgb,
+                    input_data_format=input_data_format,
+                )
+                pixel_values.extend(patches)
+                vision_grid_thws.append(video_grid_thw)
+            pixel_values = np.array(pixel_values)
+            vision_grid_thws = np.array(vision_grid_thws)
+            data = {"pixel_values_videos": pixel_values, "video_grid_thw": vision_grid_thws}
+
+        return BatchFeature(data=data, tensor_type=return_tensors)

model-00001-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e02923172aea9f20f153dd757d6130b6b13b6127e9cfd5a21ab729876f353916
+size 4966659944

model-00002-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bfb5526cece03cfc9eaae5165ab6e2051135fa94ade9162a58c65ef4d8fc5a51
+size 4991495816

model-00003-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8de04723034801fb8c47bc623a5148a20031fbdfae8448dc41b162f25262824e
+size 4932751040

model-00004-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1988f4e1766b0948424ed1e6dfb8d97f2cbfe98bccc0172d96f8b30b1bcde2c4
+size 1743319344

model.safetensors.index.json

@@ -0,0 +1,741 @@
+{
+  "metadata": {
+    "total_size": 16634145792
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00004-of-00004.safetensors",
+    "model.embed_tokens.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.10.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.16.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.17.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.17.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.17.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.17.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.18.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.18.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.18.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.2.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.20.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.input_layernorm.weight": "model-00004-of-00004.safetensors",
+    "model.layers.26.mlp.down_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.26.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.post_attention_layernorm.weight": "model-00004-of-00004.safetensors",
+    "model.layers.26.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.input_layernorm.weight": "model-00004-of-00004.safetensors",
+    "model.layers.27.mlp.down_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.27.mlp.gate_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.27.mlp.up_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.27.post_attention_layernorm.weight": "model-00004-of-00004.safetensors",
+    "model.layers.27.self_attn.k_proj.bias": "model-00004-of-00004.safetensors",
+    "model.layers.27.self_attn.k_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.27.self_attn.o_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.27.self_attn.q_proj.bias": "model-00004-of-00004.safetensors",
+    "model.layers.27.self_attn.q_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.27.self_attn.v_proj.bias": "model-00004-of-00004.safetensors",
+    "model.layers.27.self_attn.v_proj.weight": "model-00004-of-00004.safetensors",
+    "model.layers.3.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.5.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.5.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.5.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.6.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.6.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.6.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.6.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.7.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.7.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.7.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.7.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.8.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.8.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.8.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.norm.weight": "model-00004-of-00004.safetensors",
+    "projector.linear_1.bias": "model-00004-of-00004.safetensors",
+    "projector.linear_1.weight": "model-00004-of-00004.safetensors",
+    "projector.linear_2.bias": "model-00004-of-00004.safetensors",
+    "projector.linear_2.weight": "model-00004-of-00004.safetensors",
+    "visual.blocks.0.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.0.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.0.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.0.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.0.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.0.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.0.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.0.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.0.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.0.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.0.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.0.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.1.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.1.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.1.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.1.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.1.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.1.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.1.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.1.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.1.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.1.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.1.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.1.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.10.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.10.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.10.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.10.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.10.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.10.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.10.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.10.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.10.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.10.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.10.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.10.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.11.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.11.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.11.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.11.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.11.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.11.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.11.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.11.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.11.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.11.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.11.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.11.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.12.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.12.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.12.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.12.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.12.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.12.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.12.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.12.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.12.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.12.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.12.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.12.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.13.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.13.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.13.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.13.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.13.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.13.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.13.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.13.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.13.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.13.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.13.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.13.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.14.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.14.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.14.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.14.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.14.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.14.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.14.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.14.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.14.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.14.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.14.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.14.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.15.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.15.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.15.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.15.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.15.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.15.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.15.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.15.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.15.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.15.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.15.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.15.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.16.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.16.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.16.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.16.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.16.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.16.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.16.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.16.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.16.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.16.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.16.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.16.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.17.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.17.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.17.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.17.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.17.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.17.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.17.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.17.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.17.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.17.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.17.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.17.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.18.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.18.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.18.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.18.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.18.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.18.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.18.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.18.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.18.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.18.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.18.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.18.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.19.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.19.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.19.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.19.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.19.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.19.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.19.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.19.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.19.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.19.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.19.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.19.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.2.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.2.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.2.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.2.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.2.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.2.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.2.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.2.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.2.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.2.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.2.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.2.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.20.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.20.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.20.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.20.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.20.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.20.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.20.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.20.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.20.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.20.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.20.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.20.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.21.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.21.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.21.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.21.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.21.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.21.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.21.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.21.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.21.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.21.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.21.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.21.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.22.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.22.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.22.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.22.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.22.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.22.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.22.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.22.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.22.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.22.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.22.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.22.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.23.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.23.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.23.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.23.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.23.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.23.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.23.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.23.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.23.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.23.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.23.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.23.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.24.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.24.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.24.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.24.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.24.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.24.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.24.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.24.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.24.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.24.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.24.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.24.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.25.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.25.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.25.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.25.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.25.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.25.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.25.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.25.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.25.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.25.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.25.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.25.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.26.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.26.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.26.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.26.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.26.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.26.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.26.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.26.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.26.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.26.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.26.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.26.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.27.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.27.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.27.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.27.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.27.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.27.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.27.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.27.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.27.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.27.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.27.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.27.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.28.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.28.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.28.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.28.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.28.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.28.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.28.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.28.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.28.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.28.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.28.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.28.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.29.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.29.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.29.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.29.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.29.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.29.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.29.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.29.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.29.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.29.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.29.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.29.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.3.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.3.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.3.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.3.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.3.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.3.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.3.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.3.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.3.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.3.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.3.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.3.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.30.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.30.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.30.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.30.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.30.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.30.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.30.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.30.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.30.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.30.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.30.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.30.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.31.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.31.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.31.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.31.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.31.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.31.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.31.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.31.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.31.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.31.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.31.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.31.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.4.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.4.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.4.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.4.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.4.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.4.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.4.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.4.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.4.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.4.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.4.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.4.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.5.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.5.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.5.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.5.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.5.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.5.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.5.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.5.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.5.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.5.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.5.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.5.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.6.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.6.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.6.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.6.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.6.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.6.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.6.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.6.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.6.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.6.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.6.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.6.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.7.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.7.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.7.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.7.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.7.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.7.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.7.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.7.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.7.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.7.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.7.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.7.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.8.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.8.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.8.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.8.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.8.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.8.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.8.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.8.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.8.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.8.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.8.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.8.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.9.attn.proj.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.9.attn.proj.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.9.attn.qkv.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.9.attn.qkv.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.9.mlp.fc1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.9.mlp.fc1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.9.mlp.fc2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.9.mlp.fc2.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.9.norm1.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.9.norm1.weight": "model-00001-of-00004.safetensors",
+    "visual.blocks.9.norm2.bias": "model-00001-of-00004.safetensors",
+    "visual.blocks.9.norm2.weight": "model-00001-of-00004.safetensors",
+    "visual.merger.ln_q.bias": "model-00001-of-00004.safetensors",
+    "visual.merger.ln_q.weight": "model-00001-of-00004.safetensors",
+    "visual.merger.mlp.0.bias": "model-00001-of-00004.safetensors",
+    "visual.merger.mlp.0.weight": "model-00001-of-00004.safetensors",
+    "visual.merger.mlp.2.bias": "model-00001-of-00004.safetensors",
+    "visual.merger.mlp.2.weight": "model-00001-of-00004.safetensors",
+    "visual.patch_embed.proj.weight": "model-00001-of-00004.safetensors"
+  }
+}

modeling_dreamvl.py

@@ -0,0 +1,1824 @@
+# coding=utf-8
+# Copyright 2024 The DreamVL team 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.
+"""PyTorch DreamVL model."""
+
+import math, os
+from dataclasses import dataclass
+from typing import Any, Dict, List, Optional, Tuple, Union
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import torch.utils.checkpoint
+from torch.nn import CrossEntropyLoss, LayerNorm
+
+from transformers.activations import ACT2FN
+from transformers.cache_utils import Cache, SlidingWindowCache, StaticCache, DynamicCache
+from transformers.modeling_outputs import (
+    BaseModelOutputWithPast,
+    ModelOutput,
+    BaseModelOutput,
+    MaskedLMOutput,
+)
+from transformers.modeling_rope_utils import ROPE_INIT_FUNCTIONS
+from transformers.modeling_utils import PreTrainedModel
+from transformers.utils import (
+    add_start_docstrings,
+    add_start_docstrings_to_model_forward,
+    is_flash_attn_2_available,
+    is_flash_attn_greater_or_equal_2_10,
+    logging,
+    replace_return_docstrings,
+    is_torchdynamo_compiling
+)
+from transformers import PretrainedConfig
+
+from transformers.modeling_attn_mask_utils import (
+    AttentionMaskConverter,
+)
+
+from .configuration_dreamvl import DreamVLConfig, DreamVLVisionConfig
+from .generation_utils import DreamVLGenerationMixin, DreamVLGenerationConfig
+
+
+if is_flash_attn_2_available():
+    from flash_attn import flash_attn_varlen_func
+
+    from transformers.modeling_flash_attention_utils import _flash_attention_forward
+else:
+    flash_attn_varlen_func = None
+
+
+logger = logging.get_logger("DreamVL."+__name__)
+
+_CHECKPOINT_FOR_DOC = "DreamVL-7B"
+_CONFIG_FOR_DOC = "DreamVLConfig"
+
+@dataclass
+class DreamVLModelOutput(ModelOutput):
+    """
+    Base class for DreamVL outputs.
+    Args:
+        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.
+        hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`):
+            Tuple of `torch.FloatTensor` (one for the output of the embeddings, if the model has an embedding layer, +
+            one for the output of each layer) of shape `(batch_size, sequence_length, hidden_size)`.
+            Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.
+        attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`):
+            Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length,
+            sequence_length)`.
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
+        rope_deltas (`torch.LongTensor` of shape `(batch_size, )`, *optional*):
+            The rope index difference between sequence length and multimodal rope.
+        inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
+            The input embeddings, used for caching image feature during inference when use_cache=False.
+    """
+
+    logits: torch.FloatTensor = None
+    loss: 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
+    rope_deltas: Optional[torch.LongTensor] = None
+    inputs_embeds: Optional[torch.FloatTensor] = None
+
+
+class DreamVLRotaryEmbedding(nn.Module):
+    def __init__(
+        self,
+        dim=None,
+        max_position_embeddings=2048,
+        base=10000,
+        device=None,
+        scaling_factor=1.0,
+        rope_type="default",
+        config: Optional[DreamVLConfig] = None,
+    ):
+        super().__init__()
+        # TODO (joao): remove the `if` below, only used for BC
+        self.rope_kwargs = {}
+        if config is None:
+            logger.warning_once(
+                "`DreamVLRotaryEmbedding` can now be fully parameterized by passing the model config through the "
+                "`config` argument. All other arguments will be removed in v4.46"
+            )
+            self.rope_kwargs = {
+                "rope_type": rope_type,
+                "factor": scaling_factor,
+                "dim": dim,
+                "base": base,
+                "max_position_embeddings": max_position_embeddings,
+            }
+            self.rope_type = rope_type
+            self.max_seq_len_cached = max_position_embeddings
+            self.original_max_seq_len = max_position_embeddings
+        else:
+            # BC: "rope_type" was originally "type"
+            if 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.rope_kwargs)
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+        self.original_inv_freq = self.inv_freq
+
+    def reset_parameters(self):
+        inv_freq, self.attention_scaling = self.rope_init_fn(self.config, self.inv_freq.device, **self.rope_kwargs)
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+        self.original_inv_freq = self.inv_freq
+
+
+    def _dynamic_frequency_update(self, position_ids, device):
+        """
+        dynamic RoPE layers should recompute `inv_freq` in the following situations:
+        1 - growing beyond the cached sequence length (allow scaling)
+        2 - the current sequence length is in the original scale (avoid losing precision with small sequences)
+        """
+        seq_len = torch.max(position_ids) + 1
+        if seq_len > self.max_seq_len_cached:  # growth
+            inv_freq, self.attention_scaling = self.rope_init_fn(
+                self.config, device, seq_len=seq_len, **self.rope_kwargs
+            )
+            self.register_buffer("inv_freq", inv_freq, persistent=False)  # TODO joao: may break with compilation
+            self.max_seq_len_cached = seq_len
+
+        if seq_len < self.original_max_seq_len and self.max_seq_len_cached > self.original_max_seq_len:  # reset
+            self.register_buffer("inv_freq", self.original_inv_freq, persistent=False)
+            self.max_seq_len_cached = self.original_max_seq_len
+
+    @torch.no_grad()
+    def forward(self, x, position_ids):
+        if "dynamic" in self.rope_type:
+            self._dynamic_frequency_update(position_ids, device=x.device)
+
+        # Core RoPE block. In contrast to other models, DreamVL has different position ids for thw grids
+        # So we expand the inv_freq to shape (3, ...)
+        inv_freq_expanded = self.inv_freq[None, None, :, None].float().expand(3, position_ids.shape[1], -1, 1)
+        position_ids_expanded = position_ids[:, :, None, :].float()  # shape (3, bs, 1, positions)
+        # Force float32 (see https://github.com/huggingface/transformers/pull/29285)
+        device_type = x.device.type
+        device_type = device_type if isinstance(device_type, str) and device_type != "mps" else "cpu"
+        with torch.autocast(device_type=device_type, enabled=False):
+            freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(2, 3)
+            emb = torch.cat((freqs, freqs), dim=-1)
+            cos = emb.cos()
+            sin = emb.sin()
+
+        # Advanced RoPE types (e.g. yarn) apply a post-processing scaling factor, equivalent to scaling attention
+        cos = cos * self.attention_scaling
+        sin = sin * self.attention_scaling
+
+        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
+
+# Copied from transformers.models.qwen2.modeling_qwen2.Qwen2RMSNorm
+class DreamVLRMSNorm(nn.Module):
+    def __init__(self, hidden_size, eps=1e-6):
+        """
+        DreamVLRMSNorm 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}"
+
+# Copied from transformers.models.llama.modeling_llama.rotate_half
+def rotate_half(x):
+    """Rotates half the hidden dims of the input."""
+    x1 = x[..., : x.shape[-1] // 2]
+    x2 = x[..., x.shape[-1] // 2 :]
+    return torch.cat((-x2, x1), dim=-1)
+
+
+def apply_multimodal_rotary_pos_emb(q, k, cos, sin, mrope_section, unsqueeze_dim=1):
+    """Applies Rotary Position Embedding with Multimodal Sections to the query and key tensors (https://qwenlm.github.io/blog/qwen2-vl/).
+    Explanation:
+        Multimodal 3D rotary position embedding is an extension to 1D rotary position embedding. The input embedding
+        sequence contains vision (images / videos) embedding and text embedding or just contains text embedding. For
+        vision embedding part, we apply rotary position embedding on temporal, height and width dimension seperately.
+        Here we split the channel dimension to 3 chunks for the temporal, height and width rotary position embedding.
+        For text embedding part, we just apply 1D rotary position embedding. The three rotary position index (temporal,
+        height and width) of text embedding is always the same, so the text embedding rotary position embedding has no
+        difference with modern LLMs.
+    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`):
+            The position indices of the tokens corresponding to the query and key tensors. For example, this can be
+            used to pass offsetted position ids when working with a KV-cache.
+        mrope_section(`List(int)`):
+            Multimodal rope section is for channel dimension of temporal, height and width in rope calculation.
+        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.
+    """
+    mrope_section = mrope_section * 2
+    cos = torch.cat([m[i % 3] for i, m in enumerate(cos.split(mrope_section, dim=-1))], dim=-1).unsqueeze(
+        unsqueeze_dim
+    )
+    sin = torch.cat([m[i % 3] for i, m in enumerate(sin.split(mrope_section, dim=-1))], dim=-1).unsqueeze(
+        unsqueeze_dim
+    )
+
+    q_embed = (q * cos) + (rotate_half(q) * sin)
+    k_embed = (k * cos) + (rotate_half(k) * sin)
+    return q_embed, k_embed
+
+
+def apply_rotary_pos_emb_vision(
+    q: torch.Tensor, k: torch.Tensor, cos: torch.Tensor, sin: torch.Tensor
+) -> Tuple[torch.Tensor, torch.Tensor]:
+    orig_q_dtype = q.dtype
+    orig_k_dtype = k.dtype
+    q, k = q.float(), k.float()
+    cos, sin = cos.unsqueeze(-2).float(), sin.unsqueeze(-2).float()
+    q_embed = (q * cos) + (rotate_half(q) * sin)
+    k_embed = (k * cos) + (rotate_half(k) * sin)
+    q_embed = q_embed.to(orig_q_dtype)
+    k_embed = k_embed.to(orig_k_dtype)
+    return q_embed, k_embed
+
+# Copied from transformers.models.qwen2.modeling_qwen2.Qwen2MLP
+class DreamVLMLP(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+        self.intermediate_size = config.intermediate_size
+        self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
+        self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
+        self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=False)
+        self.act_fn = ACT2FN[config.hidden_act]
+
+    def forward(self, hidden_state):
+        return self.down_proj(self.act_fn(self.gate_proj(hidden_state)) * self.up_proj(hidden_state))
+
+
+# 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)
+
+
+class DreamVLAttention(nn.Module):
+    """
+    Multi-headed attention from 'Attention Is All You Need' paper. Modified to use sliding window attention: Longformer
+    and "Generating Long Sequences with Sparse Transformers".
+    """
+
+    def __init__(self, config: DreamVLConfig, layer_idx: Optional[int] = None):
+        super().__init__()
+        self.config = config
+        self.layer_idx = layer_idx
+        if layer_idx is None:
+            logger.warning_once(
+                f"Instantiating {self.__class__.__name__} without passing `layer_idx` is not recommended and will "
+                "to errors during the forward call, if caching is used. Please make sure to provide a `layer_idx` "
+                "when creating this class."
+            )
+
+        self.hidden_size = config.hidden_size
+        self.num_heads = config.num_attention_heads
+        self.head_dim = self.hidden_size // self.num_heads
+        self.num_key_value_heads = config.num_key_value_heads
+        self.num_key_value_groups = self.num_heads // self.num_key_value_heads
+        self.max_position_embeddings = config.max_position_embeddings
+        self.rope_theta = config.rope_theta
+        self.is_causal = False # not used in Dream
+        self.attention_dropout = config.attention_dropout
+        self.rope_scaling = config.rope_scaling # in Dream rope scaling is None
+        self.mrope_section = config.mrope_section
+
+        if (self.head_dim * self.num_heads) != self.hidden_size:
+            raise ValueError(
+                f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}"
+                f" and `num_heads`: {self.num_heads})."
+            )
+        self.q_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=True)
+        self.k_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=True)
+        self.v_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=True)
+        self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=False)
+        
+        self.rotary_emb = DreamVLRotaryEmbedding(config=self.config)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+        cache_position: Optional[torch.LongTensor] = None,
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # will become mandatory in v4.46
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        bsz, q_len, _ = hidden_states.size()
+
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+
+        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+
+        if position_embeddings is None:
+            logger.warning_once(
+                "The attention layers in this model are transitioning from computing the RoPE embeddings internally "
+                "through `position_ids` (2D tensor with the indexes of the tokens), to using externally computed "
+                "`position_embeddings` (Tuple of tensors, containing cos and sin). In v4.46 `position_ids` will be "
+                "removed and `position_embeddings` will be mandatory."
+            )
+            cos, sin = self.rotary_emb(value_states, position_ids)
+        else:
+            cos, sin = position_embeddings
+        query_states, key_states = apply_multimodal_rotary_pos_emb(
+            query_states, key_states, cos, sin, self.mrope_section
+        )
+
+        if past_key_value is not None:
+            cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}  # Specific to RoPE models
+            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+        # repeat k/v heads if n_kv_heads < n_heads
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+        attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim)
+        if attention_mask is not None:  # no matter the length, we just slice it
+            causal_mask = attention_mask[:, :, :, : key_states.shape[-2]]
+            attn_weights = attn_weights + causal_mask
+
+        # Fix precision issues in DreamVL float16 inference
+        # Replace inf values with zeros in attention weights to prevent NaN propagation
+        if query_states.dtype == torch.float16:
+            attn_weights = torch.where(torch.isinf(attn_weights), torch.zeros_like(attn_weights), attn_weights)
+
+        # upcast attention to fp32
+        attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype)
+        attn_weights = nn.functional.dropout(attn_weights, p=self.attention_dropout, training=self.training)
+        attn_output = torch.matmul(attn_weights, value_states)
+
+        if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):
+            raise ValueError(
+                f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is"
+                f" {attn_output.size()}"
+            )
+
+        attn_output = attn_output.transpose(1, 2).contiguous()
+        attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)
+
+        attn_output = self.o_proj(attn_output)
+
+        if not output_attentions:
+            attn_weights = None
+
+        return attn_output, attn_weights, past_key_value
+
+
+
+class DreamVLFlashAttention2(DreamVLAttention):
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+        cache_position: Optional[torch.LongTensor] = None,
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # necessary, but kept here for BC
+    ):
+        bsz, q_len, _ = hidden_states.size()
+
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+
+        query_states = query_states.view(bsz, q_len, -1, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, -1, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, -1, self.head_dim).transpose(1, 2)
+
+        # Because the input can be padded, the absolute sequence length depends on the max position id.
+        if position_embeddings is None:
+            logger.warning_once(
+                "The attention layers in this model are transitioning from computing the RoPE embeddings internally "
+                "through `position_ids` (2D tensor with the indexes of the tokens), to using externally computed "
+                "`position_embeddings` (Tuple of tensors, containing cos and sin). In v4.46 `position_ids` will be "
+                "removed and `position_embeddings` will be mandatory."
+            )
+            cos, sin = self.rotary_emb(value_states, position_ids)
+        else:
+            cos, sin = position_embeddings
+        query_states, key_states = apply_multimodal_rotary_pos_emb(
+            query_states, key_states, cos, sin, self.mrope_section
+        )
+
+        # repeat k/v heads if n_kv_heads < n_heads
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+        dropout_rate = 0.0 if not self.training else self.attention_dropout
+
+        # In PEFT, usually we cast the layer norms in float32 for training stability reasons
+        # therefore the input hidden states gets silently casted in float32. Hence, we need
+        # cast them back in float16 just to be sure everything works as expected.
+        input_dtype = query_states.dtype
+        if input_dtype == torch.float32:
+            if torch.is_autocast_enabled():
+                target_dtype = torch.get_autocast_gpu_dtype()
+            # Handle the case where the model is quantized
+            elif hasattr(self.config, "_pre_quantization_dtype"):
+                target_dtype = self.config._pre_quantization_dtype
+            else:
+                target_dtype = self.q_proj.weight.dtype
+
+            logger.warning_once(
+                f"The input hidden states seems to be silently casted in float32, this might be related to"
+                f" the fact you have upcasted embedding or layer norm layers in float32. We will cast back the input in"
+                f" {target_dtype}."
+            )
+
+            query_states = query_states.to(target_dtype)
+            key_states = key_states.to(target_dtype)
+            value_states = value_states.to(target_dtype)
+
+        # Reashape to the expected shape for Flash Attention
+        query_states = query_states.transpose(1, 2)
+        key_states = key_states.transpose(1, 2)
+        value_states = value_states.transpose(1, 2)
+
+        if (
+            self.config.use_sliding_window
+            and getattr(self.config, "sliding_window", None) is not None
+            and self.layer_idx >= self.config.max_window_layers
+        ):
+            sliding_window = self.config.sliding_window
+        else:
+            sliding_window = None
+
+        attn_output = _flash_attention_forward(
+            query_states,
+            key_states,
+            value_states,
+            attention_mask,
+            q_len,
+            dropout=dropout_rate,
+            sliding_window=sliding_window,
+            is_causal=False
+        )
+
+        attn_output = attn_output.reshape(bsz, q_len, self.hidden_size).contiguous()
+        attn_output = self.o_proj(attn_output)
+
+        if not output_attentions:
+            attn_weights = None
+
+        return attn_output, attn_weights, past_key_value
+
+
+class DreamVLSdpaAttention(DreamVLAttention):
+    """
+    DreamVL attention module using torch.nn.functional.scaled_dot_product_attention. This module inherits from
+    `DreamVLAttention` as the weights of the module stays untouched. The only changes are on the forward pass to adapt to
+    SDPA API.
+    """
+
+    # Adapted from DreamVLAttention.forward
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: bool = False,
+        use_cache: bool = False,
+        cache_position: Optional[torch.LongTensor] = None,
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # will become mandatory in v4.46
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        if output_attentions:
+            # TODO: Improve this warning with e.g. `model.config.attn_implementation = "manual"` once this is implemented.
+            logger.warning_once(
+                "DreamVLModel is using DreamVLSdpaAttention, but `torch.nn.functional.scaled_dot_product_attention` does not support `output_attentions=True`. Falling back to the manual attention implementation, "
+                'but specifying the manual implementation will be required from Transformers version v5.0.0 onwards. This warning can be removed using the argument `attn_implementation="eager"` when loading the model.'
+            )
+            return super().forward(
+                hidden_states=hidden_states,
+                attention_mask=attention_mask,
+                position_ids=position_ids,
+                past_key_value=past_key_value,
+                output_attentions=output_attentions,
+                use_cache=use_cache,
+                # cache_position=cache_position, # not used in Dream
+            )
+
+        bsz, q_len, _ = hidden_states.size()
+
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+
+        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
+
+        if position_embeddings is None:
+            logger.warning_once(
+                "The attention layers in this model are transitioning from computing the RoPE embeddings internally "
+                "through `position_ids` (2D tensor with the indexes of the tokens), to using externally computed "
+                "`position_embeddings` (Tuple of tensors, containing cos and sin). In v4.46 `position_ids` will be "
+                "removed and `position_embeddings` will be mandatory."
+            )
+            cos, sin = self.rotary_emb(value_states, position_ids)
+        else:
+            cos, sin = position_embeddings
+        query_states, key_states = apply_multimodal_rotary_pos_emb(
+            query_states, key_states, cos, sin, self.mrope_section
+        )
+
+        if past_key_value is not None:
+            logger.warning_once(
+                f"In {self.__class__}, cache is used."
+            )
+            # print("cache is used")
+            cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}  # Specific to RoPE models
+            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+        # SDPA with memory-efficient backend is currently (torch==2.1.2) bugged with non-contiguous inputs with custom attn_mask,
+        # Reference: https://github.com/pytorch/pytorch/issues/112577.
+        if query_states.device.type == "cuda" and attention_mask is not None:
+            query_states = query_states.contiguous()
+            key_states = key_states.contiguous()
+            value_states = value_states.contiguous()
+            
+            if isinstance(attention_mask, torch.Tensor) and len(attention_mask.shape) == 2:
+                # attention_mask is of shape [B, N], here broadcast to [B, 1, N, N]
+                attention_mask = torch.logical_and(
+                    attention_mask.unsqueeze(1).unsqueeze(-2),
+                    attention_mask.unsqueeze(1).unsqueeze(-1),
+                )
+        # We dispatch to SDPA's Flash Attention or Efficient kernels via this `is_causal` if statement instead of an inline conditional assignment
+        # in SDPA to support both torch.compile's dynamic shapes and full graph options. An inline conditional prevents dynamic shapes from compiling.
+        # The q_len > 1 is necessary to match with AttentionMaskConverter.to_causal_4d that does not create a causal mask in case q_len == 1.
+        # is_causal = True if causal_mask is None and q_len > 1 else False # not used in Dream
+
+        attn_output = torch.nn.functional.scaled_dot_product_attention(
+            query_states,
+            key_states,
+            value_states,
+            attn_mask=attention_mask if isinstance(attention_mask, torch.Tensor) else None,
+            dropout_p=self.attention_dropout if self.training else 0.0,
+            is_causal=False, # hard coded
+        )
+        # if torch.__version__ < "2.5":
+        #     attn_output = torch.nan_to_num(attn_output, nan=0.0)
+
+        attn_output = attn_output.transpose(1, 2).contiguous()
+        attn_output = attn_output.view(bsz, q_len, self.hidden_size)
+
+        attn_output = self.o_proj(attn_output)
+
+        return attn_output, None, past_key_value
+
+
+DreamVL_ATTENTION_CLASSES = {
+    "eager": DreamVLAttention,
+    "flash_attention_2": DreamVLFlashAttention2,
+    "sdpa": DreamVLSdpaAttention,
+}
+
+
+class DreamVLDecoderLayer(nn.Module):
+    def __init__(self, config: DreamVLConfig, layer_idx: int):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+
+        if config.sliding_window and config._attn_implementation != "flash_attention_2":
+            logger.warning_once(
+                f"Sliding Window Attention is enabled but not implemented for `{config._attn_implementation}`; "
+                "unexpected results may be encountered."
+            )
+        # self.self_attn = DreamVL_ATTENTION_CLASSES[config._attn_implementation](config, layer_idx)
+        self.self_attn = DreamVLSdpaAttention(config, layer_idx)
+        # self.self_attn = DreamVLFlashAttention2(config, layer_idx)
+
+        self.mlp = DreamVLMLP(config)
+        self.input_layernorm = DreamVLRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.post_attention_layernorm = DreamVLRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Tuple[torch.Tensor]] = None,
+        output_attentions: Optional[bool] = False,
+        use_cache: Optional[bool] = False,
+        cache_position: Optional[torch.LongTensor] = None,
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # will become mandatory in v4.46
+        **kwargs,
+    ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
+        """
+        Args:
+            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
+            attention_mask (`torch.FloatTensor`, *optional*): attention mask of size
+                `(batch, sequence_length)` where padding elements are indicated by 0.
+            output_attentions (`bool`, *optional*):
+                Whether or not to return the attentions tensors of all attention layers. See `attentions` under
+                returned tensors for more detail.
+            use_cache (`bool`, *optional*):
+                If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding
+                (see `past_key_values`).
+            past_key_value (`Tuple(torch.FloatTensor)`, *optional*): cached past key and value projection states
+            cache_position (`torch.LongTensor` of shape `(sequence_length)`, *optional*):
+                Indices depicting the position of the input sequence tokens in the sequence.
+            position_embeddings (`Tuple[torch.FloatTensor, torch.FloatTensor]`, *optional*):
+                Tuple containing the cosine and sine positional embeddings of shape `(batch_size, seq_len, head_dim)`,
+                with `head_dim` being the embedding dimension of each attention head.
+            kwargs (`dict`, *optional*):
+                Arbitrary kwargs to be ignored, used for FSDP and other methods that injects code
+                into the model
+        """
+
+        residual = hidden_states
+
+        hidden_states = self.input_layernorm(hidden_states)
+
+        # Self Attention
+        hidden_states, self_attn_weights, present_key_value = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_value=past_key_value,
+            output_attentions=output_attentions,
+            use_cache=use_cache,
+            cache_position=cache_position,
+            position_embeddings=position_embeddings,
+        )
+        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,)
+
+        if use_cache:
+            outputs += (present_key_value,)
+
+        return outputs
+
+######## START VISION ########
+class VisionRotaryEmbedding(nn.Module):
+    def __init__(self, dim: int, theta: float = 10000.0) -> None:
+        super().__init__()
+        inv_freq = 1.0 / (theta ** (torch.arange(0, dim, 2, dtype=torch.float) / dim))
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+
+    def forward(self, seqlen: int) -> torch.Tensor:
+        seq = torch.arange(seqlen, device=self.inv_freq.device, dtype=self.inv_freq.dtype)
+        freqs = torch.outer(seq, self.inv_freq)
+        return freqs
+
+
+class PatchEmbed(nn.Module):
+    def __init__(
+        self,
+        patch_size: int = 14,
+        temporal_patch_size: int = 2,
+        in_channels: int = 3,
+        embed_dim: int = 1152,
+    ) -> None:
+        super().__init__()
+        self.patch_size = patch_size
+        self.temporal_patch_size = temporal_patch_size
+        self.in_channels = in_channels
+        self.embed_dim = embed_dim
+
+        kernel_size = [temporal_patch_size, patch_size, patch_size]
+        self.proj = nn.Conv3d(in_channels, embed_dim, kernel_size=kernel_size, stride=kernel_size, bias=False)
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        target_dtype = self.proj.weight.dtype
+        hidden_states = hidden_states.view(
+            -1, self.in_channels, self.temporal_patch_size, self.patch_size, self.patch_size
+        )
+        hidden_states = self.proj(hidden_states.to(dtype=target_dtype)).view(-1, self.embed_dim)
+        return hidden_states
+
+
+class PatchMerger(nn.Module):
+    def __init__(self, dim: int, context_dim: int, spatial_merge_size: int = 2) -> None:
+        super().__init__()
+        self.hidden_size = context_dim * (spatial_merge_size**2)
+        self.ln_q = LayerNorm(context_dim, eps=1e-6)
+        self.mlp = nn.Sequential(
+            nn.Linear(self.hidden_size, self.hidden_size),
+            nn.GELU(),
+            nn.Linear(self.hidden_size, dim),
+        )
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        x = self.mlp(self.ln_q(x).view(-1, self.hidden_size))
+        return x
+
+class VisionMlp(nn.Module):
+    def __init__(self, dim: int, hidden_dim: int, hidden_act: str) -> None:
+        super().__init__()
+        self.fc1 = nn.Linear(dim, hidden_dim)
+        self.act = ACT2FN[hidden_act]
+        self.fc2 = nn.Linear(hidden_dim, dim)
+
+    def forward(self, x) -> torch.Tensor:
+        return self.fc2(self.act(self.fc1(x)))
+
+class VisionAttention(nn.Module):
+    def __init__(self, dim: int, num_heads: int = 16) -> None:
+        super().__init__()
+        self.num_heads = num_heads
+        self.head_dim = dim // num_heads
+        self.qkv = nn.Linear(dim, dim * 3, bias=True)
+        self.proj = nn.Linear(dim, dim)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        cu_seqlens: torch.Tensor,
+        rotary_pos_emb: Optional[torch.Tensor] = None,
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
+    ) -> torch.Tensor:
+        seq_length = hidden_states.shape[0]
+        q, k, v = self.qkv(hidden_states).reshape(seq_length, 3, self.num_heads, -1).permute(1, 0, 2, 3).unbind(0)
+        if position_embeddings is None:
+            logger.warning_once(
+                "The attention layers in this model are transitioning from computing the RoPE embeddings internally "
+                "through `rotary_pos_emb` (2D tensor of RoPE theta values), to using externally computed "
+                "`position_embeddings` (Tuple of tensors, containing cos and sin). In v4.54 `rotary_pos_emb` will be "
+                "removed and `position_embeddings` will be mandatory."
+            )
+            emb = torch.cat((rotary_pos_emb, rotary_pos_emb), dim=-1)
+            cos = emb.cos()
+            sin = emb.sin()
+        else:
+            cos, sin = position_embeddings
+        q, k = apply_rotary_pos_emb_vision(q, k, cos, sin)
+
+        attention_mask = torch.full(
+            [1, seq_length, seq_length], torch.finfo(q.dtype).min, device=q.device, dtype=q.dtype
+        )
+        for i in range(1, len(cu_seqlens)):
+            attention_mask[..., cu_seqlens[i - 1] : cu_seqlens[i], cu_seqlens[i - 1] : cu_seqlens[i]] = 0
+
+        q = q.transpose(0, 1)
+        k = k.transpose(0, 1)
+        v = v.transpose(0, 1)
+        attn_weights = torch.matmul(q, k.transpose(1, 2)) / math.sqrt(self.head_dim)
+        attn_weights = attn_weights + attention_mask
+        attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(q.dtype)
+        attn_output = torch.matmul(attn_weights, v)
+        attn_output = attn_output.transpose(0, 1)
+        attn_output = attn_output.reshape(seq_length, -1)
+        attn_output = self.proj(attn_output)
+        return attn_output
+
+class VisionFlashAttention2(nn.Module):
+    def __init__(self, dim: int, num_heads: int = 16) -> None:
+        super().__init__()
+        self.num_heads = num_heads
+        self.qkv = nn.Linear(dim, dim * 3, bias=True)
+        self.proj = nn.Linear(dim, dim)
+
+    def forward(
+            self, 
+            hidden_states: torch.Tensor, 
+            cu_seqlens: torch.Tensor, 
+            rotary_pos_emb: Optional[torch.Tensor] = None,
+            position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
+    ) -> torch.Tensor:
+        seq_length = hidden_states.shape[0]
+        q, k, v = self.qkv(hidden_states).reshape(seq_length, 3, self.num_heads, -1).permute(1, 0, 2, 3).unbind(0)
+        if position_embeddings is None:
+            logger.warning_once(
+                "The attention layers in this model are transitioning from computing the RoPE embeddings internally "
+                "through `rotary_pos_emb` (2D tensor of RoPE theta values), to using externally computed "
+                "`position_embeddings` (Tuple of tensors, containing cos and sin). In v4.54 `rotary_pos_emb` will be "
+                "removed and `position_embeddings` will be mandatory."
+            )
+            emb = torch.cat((rotary_pos_emb, rotary_pos_emb), dim=-1)
+            cos = emb.cos()
+            sin = emb.sin()
+        else:
+            cos, sin = position_embeddings
+        q, k = apply_rotary_pos_emb_vision(q, k, cos, sin)
+        q = apply_rotary_pos_emb_vision(q.unsqueeze(0), rotary_pos_emb).squeeze(0)
+        k = apply_rotary_pos_emb_vision(k.unsqueeze(0), rotary_pos_emb).squeeze(0)
+
+        max_seqlen = (cu_seqlens[1:] - cu_seqlens[:-1]).max().item()
+        attn_output = flash_attn_varlen_func(q, k, v, cu_seqlens, cu_seqlens, max_seqlen, max_seqlen).reshape(
+            seq_length, -1
+        )
+        attn_output = self.proj(attn_output)
+        return attn_output
+
+class VisionSdpaAttention(nn.Module):
+    def __init__(self, dim: int, num_heads: int = 16) -> None:
+        super().__init__()
+        self.num_heads = num_heads
+        self.qkv = nn.Linear(dim, dim * 3, bias=True)
+        self.proj = nn.Linear(dim, dim)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        cu_seqlens: torch.Tensor,
+        rotary_pos_emb: Optional[torch.Tensor] = None,
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
+    ) -> torch.Tensor:
+        seq_length = hidden_states.shape[0]
+        q, k, v = self.qkv(hidden_states).reshape(seq_length, 3, self.num_heads, -1).permute(1, 0, 2, 3).unbind(0)
+        if position_embeddings is None:
+            logger.warning_once(
+                "The attention layers in this model are transitioning from computing the RoPE embeddings internally "
+                "through `rotary_pos_emb` (2D tensor of RoPE theta values), to using externally computed "
+                "`position_embeddings` (Tuple of tensors, containing cos and sin). In v4.54 `rotary_pos_emb` will be "
+                "removed and `position_embeddings` will be mandatory."
+            )
+            emb = torch.cat((rotary_pos_emb, rotary_pos_emb), dim=-1)
+            cos = emb.cos()
+            sin = emb.sin()
+        else:
+            cos, sin = position_embeddings
+        q, k = apply_rotary_pos_emb_vision(q, k, cos, sin)
+
+        attention_mask = torch.zeros([1, seq_length, seq_length], device=q.device, dtype=torch.bool)
+        for i in range(1, len(cu_seqlens)):
+            attention_mask[..., cu_seqlens[i - 1] : cu_seqlens[i], cu_seqlens[i - 1] : cu_seqlens[i]] = True
+        q = q.transpose(0, 1)
+        k = k.transpose(0, 1)
+        v = v.transpose(0, 1)
+        attn_output = F.scaled_dot_product_attention(
+            q.unsqueeze(0), k.unsqueeze(0), v.unsqueeze(0), attention_mask, dropout_p=0.0
+        )
+        attn_output = attn_output.squeeze(0).transpose(0, 1)
+        attn_output = attn_output.reshape(seq_length, -1)
+        attn_output = self.proj(attn_output)
+        return attn_output
+
+
+VISION_ATTENTION_CLASSES = {
+    "eager": VisionAttention,
+    "flash_attention_2": VisionFlashAttention2,
+    "sdpa": VisionSdpaAttention,
+}
+
+class VisionBlock(nn.Module):
+    def __init__(self, config, attn_implementation: str = "sdpa") -> None:
+        super().__init__()
+        self.norm1 = LayerNorm(config.embed_dim, eps=1e-6)
+        self.norm2 = LayerNorm(config.embed_dim, eps=1e-6)
+        mlp_hidden_dim = int(config.embed_dim * config.mlp_ratio)
+
+        self.attn = VISION_ATTENTION_CLASSES[attn_implementation](
+            config.embed_dim, num_heads=config.num_heads
+        )
+        self.mlp = VisionMlp(dim=config.embed_dim, hidden_dim=mlp_hidden_dim, hidden_act=config.hidden_act)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        cu_seqlens: torch.Tensor,
+        rotary_pos_emb: Optional[torch.Tensor] = None,
+        position_embeddings: Optional[tuple[torch.Tensor, torch.Tensor]] = None,
+    ) -> torch.Tensor:
+        hidden_states = hidden_states + self.attn(
+            self.norm1(hidden_states),
+            cu_seqlens=cu_seqlens,
+            rotary_pos_emb=rotary_pos_emb,
+            position_embeddings=position_embeddings,
+        )
+        hidden_states = hidden_states + self.mlp(self.norm2(hidden_states))
+        return hidden_states
+    
+
+######## END VISION ########
+
+DreamVL_START_DOCSTRING = r"""
+    This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the
+    library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
+    etc.)
+    This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass.
+    Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage
+    and behavior.
+    Parameters:
+        config ([`DreamVLConfig`]):
+            Model configuration class with all the parameters of the model. Initializing with a config file does not
+            load the weights associated with the model, only the configuration. Check out the
+            [`~PreTrainedModel.from_pretrained`] method to load the model weights.
+"""
+
+
+@add_start_docstrings(
+    "The bare DreamVL Model outputting raw hidden-states without any specific head on top.",
+    DreamVL_START_DOCSTRING,
+)
+class DreamVLPreTrainedModel(PreTrainedModel):
+    config_class = DreamVLConfig
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["DreamVLDecoderLayer", "DreamVLVisionBlock"]
+    _skip_keys_device_placement = "past_key_values"
+    _supports_flash_attn_2 = True
+    _supports_sdpa = True
+    _supports_cache_class = True
+    _supports_quantized_cache = True
+    _supports_static_cache = True
+
+    def _init_weights(self, module):
+        std = self.config.initializer_range
+        if isinstance(module, (nn.Linear, nn.Conv3d)):
+            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, DreamVLRMSNorm):
+            module.weight.data.fill_(1.0)
+
+    @classmethod
+    def from_pretrained(
+        cls,
+        pretrained_model_name_or_path: Optional[Union[str, os.PathLike]],
+        *model_args,
+        config: Optional[Union[PretrainedConfig, str, os.PathLike]] = None,
+        cache_dir: Optional[Union[str, os.PathLike]] = None,
+        ignore_mismatched_sizes: bool = False,
+        force_download: bool = False,
+        local_files_only: bool = False,
+        token: Optional[Union[str, bool]] = None,
+        revision: str = "main",
+        use_safetensors: Optional[bool] = None,
+        weights_only: bool = True,
+        **kwargs,
+    ):
+        _model = super().from_pretrained(
+            pretrained_model_name_or_path,
+            *model_args,
+            config=config,
+            cache_dir=cache_dir,
+            ignore_mismatched_sizes=ignore_mismatched_sizes,
+            force_download=force_download,
+            local_files_only=local_files_only,
+            token=token,
+            revision=revision,
+            use_safetensors=use_safetensors,
+            weights_only=weights_only,
+            **kwargs,
+        )
+        # NOTE(Lin): we need to override the generation config
+        # because the generation config loaded in `from_pretrained` 
+        # does not include all the attributes of DreamVLGenerationConfig
+        resume_download = kwargs.get("resume_download", None)
+        proxies = kwargs.get("proxies", None)
+        subfolder = kwargs.get("subfolder", "")
+        from_auto_class = kwargs.get("_from_auto", False)
+        from_pipeline = kwargs.get("_from_pipeline", None)
+        _model.generation_config = DreamVLGenerationConfig.from_pretrained(
+            pretrained_model_name_or_path,
+            cache_dir=cache_dir,
+            force_download=force_download,
+            resume_download=resume_download,
+            proxies=proxies,
+            local_files_only=local_files_only,
+            token=token,
+            revision=revision,
+            subfolder=subfolder,
+            _from_auto=from_auto_class,
+            _from_pipeline=from_pipeline,
+        )
+        return _model
+
+
+class DreamVLVisionTransformerPretrainedModel(DreamVLPreTrainedModel):
+    config_class = DreamVLVisionConfig
+    _no_split_modules = ["DreamVLVisionBlock"] 
+
+    def __init__(self, config) -> None:
+        super().__init__(config)
+        self.spatial_merge_size = config.spatial_merge_size
+
+        self.patch_embed = PatchEmbed(
+            patch_size=config.patch_size,
+            temporal_patch_size=config.temporal_patch_size,
+            in_channels=config.in_channels,
+            embed_dim=config.embed_dim,
+        )
+
+        head_dim = config.embed_dim // config.num_heads
+        self.rotary_pos_emb = VisionRotaryEmbedding(head_dim // 2)
+
+        self.blocks = nn.ModuleList(
+            [VisionBlock(config, config._attn_implementation) for _ in range(config.depth)]
+        )
+        self.merger = PatchMerger(
+            dim=config.hidden_size, context_dim=config.embed_dim, spatial_merge_size=config.spatial_merge_size
+        )
+        self.gradient_checkpointing = False
+
+    def rot_pos_emb(self, grid_thw):
+        pos_ids = []
+        for t, h, w in grid_thw:
+            hpos_ids = torch.arange(h).unsqueeze(1).expand(-1, w)
+            hpos_ids = hpos_ids.reshape(
+                h // self.spatial_merge_size,
+                self.spatial_merge_size,
+                w // self.spatial_merge_size,
+                self.spatial_merge_size,
+            )
+            hpos_ids = hpos_ids.permute(0, 2, 1, 3)
+            hpos_ids = hpos_ids.flatten()
+
+            wpos_ids = torch.arange(w).unsqueeze(0).expand(h, -1)
+            wpos_ids = wpos_ids.reshape(
+                h // self.spatial_merge_size,
+                self.spatial_merge_size,
+                w // self.spatial_merge_size,
+                self.spatial_merge_size,
+            )
+            wpos_ids = wpos_ids.permute(0, 2, 1, 3)
+            wpos_ids = wpos_ids.flatten()
+            pos_ids.append(torch.stack([hpos_ids, wpos_ids], dim=-1).repeat(t, 1))
+        pos_ids = torch.cat(pos_ids, dim=0)
+        max_grid_size = grid_thw[:, 1:].max()
+        rotary_pos_emb_full = self.rotary_pos_emb(max_grid_size)
+        rotary_pos_emb = rotary_pos_emb_full[pos_ids].flatten(1)
+        return rotary_pos_emb
+    
+    def forward(self, hidden_states: torch.Tensor, grid_thw: torch.Tensor) -> torch.Tensor:
+        r"""
+        grid_thw (`torch.LongTensor` of shape `(num_images, 3)`):
+            The temporal, height and width dimensions of feature shape for each image. Each row contains [t, h, w] values.
+        """
+        hidden_states = self.patch_embed(hidden_states)
+        rotary_pos_emb = self.rot_pos_emb(grid_thw)
+        emb = torch.cat((rotary_pos_emb, rotary_pos_emb), dim=-1)
+        position_embeddings = (emb.cos(), emb.sin())
+
+        cu_seqlens = torch.repeat_interleave(grid_thw[:, 1] * grid_thw[:, 2], grid_thw[:, 0]).cumsum(
+            dim=0,
+            # Select dtype based on the following factors:
+            #  - FA2 requires that cu_seqlens_q must have dtype int32
+            #  - torch.onnx.export requires that cu_seqlens_q must have same dtype as grid_thw
+            # See https://github.com/huggingface/transformers/pull/34852 for more information
+            dtype=grid_thw.dtype if torch.jit.is_tracing() else torch.int32,
+        )
+        cu_seqlens = F.pad(cu_seqlens, (1, 0), value=0)
+
+        for blk in self.blocks:
+            if self.gradient_checkpointing and self.training:
+                hidden_states = self._gradient_checkpointing_func(
+                    blk.__call__, hidden_states, cu_seqlens, None, position_embeddings
+                )
+            else:
+                hidden_states = blk(hidden_states, cu_seqlens=cu_seqlens, position_embeddings=position_embeddings)
+
+        return self.merger(hidden_states)
+
+# Copied from transformers.models.llava.modeling_llava.LlavaMultiModalProjector with Llava->DreamVL
+class DreamVLMultiModalProjector(nn.Module):
+    def __init__(self, config: DreamVLConfig):
+        super().__init__()
+
+        self.linear_1 = nn.Linear(config.vision_config.hidden_size, config.hidden_size, bias=True)
+        self.act = ACT2FN[config.projector_hidden_act]
+        self.linear_2 = nn.Linear(config.hidden_size, config.hidden_size, bias=True)
+
+    def forward(self, image_features):
+        hidden_states = self.linear_1(image_features)
+        hidden_states = self.act(hidden_states)
+        hidden_states = self.linear_2(hidden_states)
+        return hidden_states
+
+@add_start_docstrings(
+    "The bare DreamVL Model outputting raw hidden-states without any specific head on top.",
+    DreamVL_START_DOCSTRING,
+)
+class DreamVLBaseModel(DreamVLPreTrainedModel):
+    def __init__(self, config: DreamVLConfig):
+        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(
+            [DreamVLDecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
+        )
+        self._attn_implementation = config._attn_implementation
+        self.norm = DreamVLRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.rotary_emb = DreamVLRotaryEmbedding(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
+
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = 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
+        )
+        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 (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:
+            if use_cache:
+                use_cache = False
+
+        if inputs_embeds is None:
+            inputs_embeds = self.embed_tokens(input_ids)
+
+        if use_cache and past_key_values is None:
+            logger.warning_once(
+                    "This should not be triggered, in either training or inference, but if it is, please report it to us."
+                )
+            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
+            )
+
+        # the hard coded `3` is for temporal, height and width.
+        if position_ids is None:
+            logger.warning_once(
+                    "This should not be triggered, in either training or inference, but if it is, please report it to us."
+                )
+            position_ids = cache_position.view(1, 1, -1).expand(3, inputs_embeds.shape[0], -1)
+        elif position_ids.dim() == 2:
+            logger.warning_once(
+                    "This should not be triggered, in either training or inference, but if it is, please report it to us."
+                )
+            position_ids = position_ids[None, ...].expand(3, position_ids.shape[0], -1)
+
+        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:
+            if output_hidden_states:
+                all_hidden_states += (hidden_states,)
+
+            if self.gradient_checkpointing and self.training:
+                layer_outputs = self._gradient_checkpointing_func(
+                    decoder_layer.__call__,
+                    hidden_states,
+                    attention_mask,
+                    position_ids,
+                    past_key_values,
+                    output_attentions,
+                    use_cache,
+                    cache_position,
+                    position_embeddings,
+                )
+            else:
+                layer_outputs = decoder_layer(
+                    hidden_states,
+                    attention_mask=attention_mask,
+                    position_ids=position_ids,
+                    past_key_value=past_key_values,
+                    output_attentions=output_attentions,
+                    use_cache=use_cache,
+                    cache_position=cache_position,
+                    position_embeddings=position_embeddings,
+                )
+
+            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,)
+
+        if not return_dict:
+            return tuple(v for v in [hidden_states, all_hidden_states, all_self_attns] if v is not None)
+        return BaseModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=past_key_values,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attns,
+        )
+
+DreamVL_INPUTS_DOCSTRING = r"""
+    Args:
+        input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
+            Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
+            it.
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+            [What are input IDs?](../glossary#input-ids)
+        attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+            [What are attention masks?](../glossary#attention-mask)
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+            If `past_key_values` is used, optionally only the last `decoder_input_ids` have to be input (see
+            `past_key_values`).
+            If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`]
+            and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more
+            information on the default strategy.
+            - 1 indicates the head is **not masked**,
+            - 0 indicates the head is **masked**.
+        position_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
+            config.n_positions - 1]`. [What are position IDs?](../glossary#position-ids)
+        past_key_values (`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 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 in the cross-attention
+            blocks) that can be used (see `past_key_values` input) to speed up sequential decoding.
+            If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those that
+            don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of all
+            `decoder_input_ids` of shape `(batch_size, sequence_length)`.
+        inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
+            Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This
+            is useful if you want more control over how to convert `input_ids` indices into associated vectors than the
+            model's internal embedding lookup matrix.
+        use_cache (`bool`, *optional*):
+            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
+            `past_key_values`).
+        output_attentions (`bool`, *optional*):
+            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
+            tensors for more detail.
+        output_hidden_states (`bool`, *optional*):
+            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
+            more detail.
+        return_dict (`bool`, *optional*):
+            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
+        pixel_values (`torch.FloatTensor` of shape `(seq_length, num_channels * image_size * image_size)):
+            The tensors corresponding to the input images. Pixel values can be obtained using
+            [`AutoImageProcessor`]. See [`DreamVLImageProcessor.__call__`] for details. [`DreamVLProcessor`] uses
+            [`DreamVLImageProcessor`] for processing images.
+        pixel_values_videos (`torch.FloatTensor` of shape `(seq_length, num_channels * temporal_size * image_size * image_size)):
+            The tensors corresponding to the input videos. Pixel values can be obtained using
+            [`AutoImageProcessor`]. See [`DreamVLImageProcessor.__call__`] for details. [`DreamVLProcessor`] uses
+            [`DreamVLImageProcessor`] for processing videos.
+        image_grid_thw (`torch.LongTensor` of shape `(num_images, 3)`, *optional*):
+            The temporal, height and width of feature shape of each image in LLM.
+        video_grid_thw (`torch.LongTensor` of shape `(num_videos, 3)`, *optional*):
+            The temporal, height and width of feature shape of each video in LLM.
+        rope_deltas (`torch.LongTensor` of shape `(batch_size, )`, *optional*):
+            The rope index difference between sequence length and multimodal rope.
+"""
+
+
+class DreamVLModel(DreamVLGenerationMixin, DreamVLPreTrainedModel):
+    _tied_weights_keys = ["lm_head.weight"]
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.visual = DreamVLVisionTransformerPretrainedModel._from_config(config.vision_config)
+        self.model = DreamVLBaseModel(config)
+        self.projector = DreamVLMultiModalProjector(config)
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def reset_rope_parameters(self):
+        self.model.rotary_emb.reset_parameters()
+        for layer in self.model.layers:
+            layer.self_attn.rotary_emb.reset_parameters()
+
+    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
+
+    def get_rope_index(
+        self,
+        input_ids: torch.LongTensor,
+        image_grid_thw: Optional[torch.LongTensor] = None,
+        video_grid_thw: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        """
+        Calculate the 3D rope index based on image and video's temporal, height and width in LLM.
+        Explanation:
+            Each embedding sequence contains vision embedding and text embedding or just contains text embedding.
+            For pure text embedding sequence, the rotary position embedding has no difference with mordern LLMs.
+            Examples:
+                input_ids: [T T T T T], here T is for text.
+                temporal position_ids: [0, 1, 2, 3, 4]
+                height position_ids: [0, 1, 2, 3, 4]
+                width position_ids: [0, 1, 2, 3, 4]
+            For vision and text embedding sequence, we calculate 3D rotary position embedding for vision part
+            and 1D rotary position embeddin for text part.
+            Examples:
+                Assume we have a video input with 3 temporal patches, 2 height patches and 2 width patches.
+                input_ids: [V V V V V V V V V V V V T T T T T], here V is for vision.
+                vision temporal position_ids: [0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2]
+                vision height position_ids: [0, 0, 1, 1, 0, 0, 1, 1, 0, 0, 1, 1]
+                vision width position_ids: [0, 1, 0, 1, 0, 1, 0, 1, 0, 1, 0, 1]
+                text temporal position_ids: [3, 4, 5, 6, 7]
+                text height position_ids: [3, 4, 5, 6, 7]
+                text width position_ids: [3, 4, 5, 6, 7]
+                Here we calculate the text start position_ids as the max vision position_ids plus 1.
+        Args:
+            input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
+                Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
+                it.
+            image_grid_thw (`torch.LongTensor` of shape `(num_images, 3)`, *optional*):
+                The temporal, height and width of feature shape of each image in LLM.
+            video_grid_thw (`torch.LongTensor` of shape `(num_videos, 3)`, *optional*):
+                The temporal, height and width of feature shape of each video in LLM.
+            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**.
+        Returns:
+            position_ids (`torch.LongTensor` of shape `(3, batch_size, sequence_length)`)
+            mrope_position_deltas (`torch.Tensor` of shape `(batch_size)`)
+        """
+        spatial_merge_size = self.config.vision_config.spatial_merge_size
+        image_token_id = self.config.image_token_id
+        video_token_id = self.config.video_token_id
+        vision_start_token_id = self.config.vision_start_token_id
+        mrope_position_deltas = []
+        if image_grid_thw is not None or video_grid_thw is not None:
+            total_input_ids = input_ids
+            if attention_mask is None:
+                attention_mask = torch.ones_like(total_input_ids)
+            position_ids = torch.ones(
+                3, input_ids.shape[0], input_ids.shape[1], dtype=input_ids.dtype, device=input_ids.device
+            )
+            image_index, video_index = 0, 0
+            for i, input_ids in enumerate(total_input_ids):
+                input_ids = input_ids[attention_mask[i] == 1]
+                image_nums, video_nums = 0, 0
+                vision_start_indices = torch.argwhere(input_ids == vision_start_token_id).squeeze(1)
+                vision_tokens = input_ids[vision_start_indices + 1]
+                image_nums = (vision_tokens == image_token_id).sum()
+                video_nums = (vision_tokens == video_token_id).sum()
+                input_tokens = input_ids.tolist()
+                llm_pos_ids_list: list = []
+                st = 0
+                remain_images, remain_videos = image_nums, video_nums
+                for _ in range(image_nums + video_nums):
+                    if image_token_id in input_tokens and remain_images > 0:
+                        ed_image = input_tokens.index(image_token_id, st)
+                    else:
+                        ed_image = len(input_tokens) + 1
+                    if video_token_id in input_tokens and remain_videos > 0:
+                        ed_video = input_tokens.index(video_token_id, st)
+                    else:
+                        ed_video = len(input_tokens) + 1
+                    if ed_image < ed_video:
+                        t, h, w = (
+                            image_grid_thw[image_index][0],
+                            image_grid_thw[image_index][1],
+                            image_grid_thw[image_index][2],
+                        )
+                        image_index += 1
+                        remain_images -= 1
+                        ed = ed_image
+                    else:
+                        t, h, w = (
+                            video_grid_thw[video_index][0],
+                            video_grid_thw[video_index][1],
+                            video_grid_thw[video_index][2],
+                        )
+                        video_index += 1
+                        remain_videos -= 1
+                        ed = ed_video
+                    llm_grid_t, llm_grid_h, llm_grid_w = (
+                        t.item(),
+                        h.item() // spatial_merge_size,
+                        w.item() // spatial_merge_size,
+                    )
+                    text_len = ed - st
+
+                    st_idx = llm_pos_ids_list[-1].max() + 1 if len(llm_pos_ids_list) > 0 else 0
+                    llm_pos_ids_list.append(torch.arange(text_len).view(1, -1).expand(3, -1) + st_idx)
+
+                    t_index = torch.arange(llm_grid_t).view(-1, 1).expand(-1, llm_grid_h * llm_grid_w).flatten()
+                    h_index = torch.arange(llm_grid_h).view(1, -1, 1).expand(llm_grid_t, -1, llm_grid_w).flatten()
+                    w_index = torch.arange(llm_grid_w).view(1, 1, -1).expand(llm_grid_t, llm_grid_h, -1).flatten()
+                    llm_pos_ids_list.append(torch.stack([t_index, h_index, w_index]) + text_len + st_idx)
+                    st = ed + llm_grid_t * llm_grid_h * llm_grid_w
+
+                if st < len(input_tokens):
+                    st_idx = llm_pos_ids_list[-1].max() + 1 if len(llm_pos_ids_list) > 0 else 0
+                    text_len = len(input_tokens) - st
+                    llm_pos_ids_list.append(torch.arange(text_len).view(1, -1).expand(3, -1) + st_idx)
+
+                llm_positions = torch.cat(llm_pos_ids_list, dim=1).reshape(3, -1)
+                position_ids[..., i, attention_mask[i] == 1] = llm_positions.to(position_ids.device)
+                mrope_position_deltas.append(llm_positions.max() + 1 - len(total_input_ids[i]))
+            mrope_position_deltas = torch.tensor(mrope_position_deltas, device=input_ids.device).unsqueeze(1)
+            return position_ids, mrope_position_deltas
+        else:
+            if attention_mask is not None:
+                position_ids = attention_mask.long().cumsum(-1) - 1
+                position_ids.masked_fill_(attention_mask == 0, 1)
+                position_ids = position_ids.unsqueeze(0).expand(3, -1, -1).to(input_ids.device)
+                max_position_ids = position_ids.max(0, keepdim=False)[0].max(-1, keepdim=True)[0]
+                mrope_position_deltas = max_position_ids + 1 - attention_mask.shape[-1]
+            else:
+                position_ids = (
+                    torch.arange(input_ids.shape[1], device=input_ids.device)
+                    .view(1, 1, -1)
+                    .expand(3, input_ids.shape[0], -1)
+                )
+                mrope_position_deltas = torch.zeros(
+                    [input_ids.shape[0], 1],
+                    device=input_ids.device,
+                    dtype=input_ids.dtype,
+                )
+
+            return position_ids, mrope_position_deltas
+
+    def get_video_features(
+        self, pixel_values_videos: torch.FloatTensor, video_grid_thw: Optional[torch.LongTensor] = None
+    ):
+        """
+        Encodes videos into continuous embeddings that can be forwarded to the language model.
+
+        Args:
+            pixel_values_videos (`torch.FloatTensor` of shape `(batch_size, num_channels, image_size, image_size)`):
+                The tensors corresponding to the input videos.
+            video_grid_thw (`torch.LongTensor` of shape `(num_videos, 3)`, *optional*):
+                The temporal, height and width of feature shape of each video in LLM.
+        """
+        pixel_values_videos = pixel_values_videos.type(self.visual.dtype)
+        video_embeds = self.visual(pixel_values_videos, grid_thw=video_grid_thw)
+        split_sizes = (video_grid_thw.prod(-1) // self.visual.spatial_merge_size**2).tolist()
+        video_embeds = torch.split(video_embeds, split_sizes)
+        return video_embeds
+
+    def get_image_features(self, pixel_values: torch.FloatTensor, image_grid_thw: Optional[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.
+            image_grid_thw (`torch.LongTensor` of shape `(num_images, 3)`, *optional*):
+                The temporal, height and width of feature shape of each image in LLM.
+        """
+        pixel_values = pixel_values.type(self.visual.dtype)
+        image_embeds = self.visual(pixel_values, grid_thw=image_grid_thw)
+        split_sizes = (image_grid_thw.prod(-1) // self.visual.spatial_merge_size**2).tolist()
+        image_embeds = torch.split(image_embeds, split_sizes)
+        return image_embeds
+
+    @add_start_docstrings_to_model_forward(DreamVL_INPUTS_DOCSTRING)
+    @replace_return_docstrings(output_type=DreamVLModelOutput, config_class=_CONFIG_FOR_DOC)
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[List[torch.FloatTensor]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        pixel_values: Optional[torch.Tensor] = None,
+        pixel_values_videos: Optional[torch.FloatTensor] = None,
+        image_grid_thw: Optional[torch.LongTensor] = None,
+        video_grid_thw: Optional[torch.LongTensor] = None,
+        rope_deltas: Optional[torch.LongTensor] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        num_logits_to_keep: int = 0,
+        **loss_kwargs,
+    ) -> Union[Tuple, DreamVLModelOutput]:
+        r"""
+        Args:
+            labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+                Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
+                config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
+                (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.
+        Returns:
+        Example:
+        ```python
+        >>> from PIL import Image
+        >>> import requests
+        >>> from transformers import AutoProcessor, DreamVLForConditionalGeneration
+        >>> model = DreamVLForConditionalGeneration.from_pretrained(" ")
+        >>> processor = AutoProcessor.from_pretrained(" ")
+        >>> messages = [
+            {
+                "role": "user",
+                "content": [
+                    {"type": "image"},
+                    {"type": "text", "text": "What is shown in this image?"},
+                ],
+            },
+        ]
+        >>> url = "https://www.ilankelman.org/stopsigns/australia.jpg"
+        >>> image = Image.open(requests.get(url, stream=True).raw)
+        >>> text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+        >>> inputs = processor(text=[text], images=[image], vision_infos=[vision_infos])
+        >>> # 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]
+        "The image shows a street scene with a red stop sign in the foreground. In the background, there is a large red gate with Chinese characters ..."
+        ```"""
+
+        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
+
+        if inputs_embeds is None:
+            inputs_embeds = self.get_input_embeddings()(input_ids)
+            if pixel_values is not None:
+                image_embeds = self.get_image_features(pixel_values, image_grid_thw)
+                image_embeds = torch.cat(image_embeds, dim=0)
+                n_image_tokens = (input_ids == self.config.image_token_id).sum()
+                n_image_features = image_embeds.shape[0]
+                if not is_torchdynamo_compiling() and n_image_tokens != n_image_features:
+                    raise ValueError(
+                        f"Image features and image tokens do not match: tokens: {n_image_tokens}, features {n_image_features}"
+                    )
+
+                mask = input_ids == self.config.image_token_id
+                mask_unsqueezed = mask.unsqueeze(-1)
+                mask_expanded = mask_unsqueezed.expand_as(inputs_embeds)
+
+                image_mask = mask_expanded.to(inputs_embeds.device)
+                image_embeds = image_embeds.to(inputs_embeds.device, inputs_embeds.dtype)
+                image_embeds_projected = self.projector(image_embeds)
+
+                inputs_embeds = inputs_embeds.masked_scatter(image_mask, image_embeds_projected)
+
+            if pixel_values_videos is not None:
+                video_embeds = self.get_video_features(pixel_values_videos, video_grid_thw)
+                video_embeds = torch.cat(video_embeds, dim=0)
+                n_video_tokens = (input_ids == self.config.video_token_id).sum()
+                n_video_features = video_embeds.shape[0]
+                if not is_torchdynamo_compiling() and n_video_tokens != n_video_features:
+                    raise ValueError(
+                        f"Video features and video tokens do not match: tokens: {n_video_tokens}, features {n_video_features}"
+                    )
+
+                mask = input_ids == self.config.video_token_id
+                mask_unsqueezed = mask.unsqueeze(-1)
+                mask_expanded = mask_unsqueezed.expand_as(inputs_embeds)
+
+                video_mask = mask_expanded.to(inputs_embeds.device)
+                video_embeds = video_embeds.to(inputs_embeds.device, inputs_embeds.dtype)
+                video_embeds_projected = self.projector(video_embeds)
+                
+                inputs_embeds = inputs_embeds.masked_scatter(video_mask, video_embeds_projected)
+
+        outputs = self.model(
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+            cache_position=cache_position,
+        )
+
+        hidden_states = outputs[0]
+        # Only compute necessary logits, and do not upcast them to float if we are not computing the loss
+        logits = self.lm_head(hidden_states[:, -num_logits_to_keep:, :])
+
+        loss = None
+        if labels is not None:
+            loss = self.loss_function(logits, labels, self.vocab_size, **loss_kwargs)
+
+        if not return_dict:
+            output = (logits,) + outputs[1:]
+            return (loss,) + output if loss is not None else output
+
+        return DreamVLModelOutput(
+            logits=logits,
+            loss=loss,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+            rope_deltas=rope_deltas,
+            inputs_embeds=inputs_embeds
+        )
+
+    def prepare_inputs_for_generation(
+        self,
+        input_ids,
+        past_key_values=None,
+        attention_mask=None,
+        inputs_embeds=None,
+        cache_position=None,
+        position_ids=None,
+        use_cache=True,
+        pixel_values=None,
+        pixel_values_videos=None,
+        image_grid_thw=None,
+        video_grid_thw=None,
+        rope_deltas = None,
+        **kwargs,
+    ):
+        # never remove input ids
+
+        if use_cache:
+            if past_key_values is None:
+                raise ValueError(
+                    "If `use_cache=True`, `past_key_values` must be provided. Please make sure to pass `past_key_values` to the model."
+                )
+            else:
+                pass
+        else:
+            past_key_values = None
+
+        if use_cache:
+            if cache_position is None:
+                raise ValueError(
+                    "If `use_cache=True`, `cache_position` must be provided. Please make sure to pass `cache_position` to the model."
+                )
+            else:
+                pass
+        else:
+            cache_position = None
+
+        if use_cache:
+            if input_ids.shape[1] != cache_position.shape[0]:
+                input_ids = input_ids[:, cache_position]
+            else:
+                pass
+        else:
+            pass
+
+        if position_ids is None:
+            if not use_cache:
+                position_ids, rope_deltas = self.get_rope_index(
+                    input_ids, image_grid_thw, video_grid_thw, attention_mask
+                )
+            else:
+                if cache_position[0] == 0:
+                    position_ids, rope_deltas = self.get_rope_index(
+                        input_ids, image_grid_thw, video_grid_thw, attention_mask
+                    )
+                else:
+                    batch_size, seq_length = input_ids.shape
+                    delta = (
+                        cache_position[0] + rope_deltas if cache_position is not None and rope_deltas is not None else 0
+                    )
+                    position_ids = torch.arange(seq_length, device=input_ids.device)
+                    position_ids = position_ids.view(1, -1).expand(batch_size, -1)
+                    position_ids = position_ids.add(delta)
+                    position_ids = position_ids.unsqueeze(0).expand(3, -1, -1)
+    
+        else:
+            raise NotImplementedError(
+                "position_ids is not None, please check the code in prepare_inputs_for_generation"
+            )
+
+        if use_cache:
+            if cache_position[0] != 0:
+                pixel_values = None
+                pixel_values_videos = None
+                logger.debug(f"after prefill, the pixel_values and pixel_values_videos are None.")
+        else:
+            pass
+
+        # if `inputs_embeds` are passed, we only want to use them in the 1st generation step
+        # if inputs_embeds is not None:
+        #     raise NotImplementedError(
+        #         "inputs_embeds is not None, please check the code in prepare_inputs_for_generation"
+        #     )
+        # else:
+        #     model_inputs = {"input_ids": input_ids, "inputs_embeds": None}
+
+        model_inputs = {
+                "input_ids": input_ids, 
+                "inputs_embeds": inputs_embeds,
+                "position_ids": position_ids,
+                "past_key_values": past_key_values,
+                "use_cache": use_cache,
+                "attention_mask": attention_mask,
+                "pixel_values": pixel_values,
+                "pixel_values_videos": pixel_values_videos,
+                "image_grid_thw": image_grid_thw,
+                "video_grid_thw": video_grid_thw,
+                "cache_position": cache_position,
+                "rope_deltas": rope_deltas,
+            }
+        
+        return model_inputs

preprocessor_config.json

@@ -0,0 +1,33 @@
+{
+  "auto_map": {
+    "AutoImageProcessor": "image_processing_dreamvl.DreamVLImageProcessor",
+    "AutoProcessor": "processing_dreamvl.DreamVLProcessor"
+  },
+  "do_convert_rgb": true,
+  "do_normalize": true,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.48145466,
+    0.4578275,
+    0.40821073
+  ],
+  "image_processor_type": "DreamVLImageProcessor",
+  "image_std": [
+    0.26862954,
+    0.26130258,
+    0.27577711
+  ],
+  "max_pixels": 3211264,
+  "merge_size": 2,
+  "min_pixels": 3136,
+  "patch_size": 14,
+  "processor_class": "DreamVLProcessor",
+  "resample": 3,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "max_pixels": 3211264,
+    "min_pixels": 3136
+  },
+  "temporal_patch_size": 2
+}

processing_dreamvl.py

@@ -0,0 +1,183 @@
+# coding=utf-8
+# Copyright 2024 The Qwen team, Alibaba Group 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.
+"""
+Processor class for Dream-VL.
+"""
+
+from typing import List, Union
+
+try:
+    from typing import Unpack
+except ImportError:
+    from typing_extensions import Unpack
+
+from transformers.feature_extraction_utils import BatchFeature
+from transformers.image_utils import ImageInput, VideoInput
+from transformers.processing_utils import (
+    ProcessingKwargs,
+    ProcessorMixin,
+)
+from transformers.tokenization_utils_base import PreTokenizedInput, TextInput
+from transformers.utils import logging
+
+logger = logging.get_logger(__name__)
+
+
+class DreamVLProcessorKwargs(ProcessingKwargs, total=False):
+    _defaults = {
+        "text_kwargs": {
+            "padding": False,
+        },
+    }
+
+
+class DreamVLProcessor(ProcessorMixin):
+    r"""
+    Constructs a Dream-VL processor which wraps a Dream-VL image processor and a Dream tokenizer into a single processor.
+    [`DreamVLProcessor`] offers all the functionalities of [`DreamVLImageProcessor`] and [`DreamTokenizer`]. See the
+    [`~DreamVLProcessor.__call__`] and [`~DreamVLProcessor.decode`] for more information.
+    Args:
+        image_processor ([`DreamVLImageProcessor`], *optional*):
+            The image processor is a required input.
+        tokenizer ([`DreamTokenizer`], *optional*):
+            The tokenizer is a required input.
+        chat_template (`str`, *optional*): A Jinja template which will be used to convert lists of messages
+            in a chat into a tokenizable string.
+    """
+
+    attributes = ["image_processor", "tokenizer"]
+    valid_kwargs = ["chat_template"]
+    image_processor_class = "AutoImageProcessor"
+    tokenizer_class = ("AutoTokenizer")
+
+    def __init__(self, image_processor=None, tokenizer=None, chat_template=None, **kwargs):
+        super().__init__(image_processor, tokenizer, chat_template=chat_template)
+
+    def __call__(
+            self,
+            images: ImageInput = None,
+            text: Union[TextInput, PreTokenizedInput, List[TextInput], List[PreTokenizedInput]] = None,
+            videos: VideoInput = None,
+            **kwargs: Unpack[DreamVLProcessorKwargs],
+    ) -> BatchFeature:
+        """
+        Main method to prepare for the model one or several sequences(s) and image(s). This method forwards the `text`
+        and `kwargs` arguments to DreamTokenizer's [`~DreamTokenizer.__call__`] if `text` is not `None` to encode
+        the text. To prepare the vision inputs, this method forwards the `vision_infos` and `kwrags` arguments to
+        DreamVLImageProcessor's [`~DreamVLImageProcessor.__call__`] if `vision_infos` is not `None`.
+
+        Args:
+            images (`PIL.Image.Image`, `np.ndarray`, `torch.Tensor`, `List[PIL.Image.Image]`, `List[np.ndarray]`, `List[torch.Tensor]`):
+                The image or batch of images to be prepared. Each image can be a PIL image, NumPy array or PyTorch
+                tensor. Both channels-first and channels-last formats are supported.
+            text (`str`, `List[str]`, `List[List[str]]`):
+                The sequence or batch of sequences to be encoded. Each sequence can be a string or a list of strings
+                (pretokenized string). If the sequences are provided as list of strings (pretokenized), you must set
+                `is_split_into_words=True` (to lift the ambiguity with a batch of sequences).
+            videos (`np.ndarray`, `torch.Tensor`, `List[np.ndarray]`, `List[torch.Tensor]`):
+                The image or batch of videos to be prepared. Each video can be a 4D NumPy array or PyTorch
+                tensor, or a nested list of 3D frames. Both channels-first and channels-last formats are supported.
+            return_tensors (`str` or [`~utils.TensorType`], *optional*):
+                If set, will return tensors of a particular framework. Acceptable values are:
+                - `'tf'`: Return TensorFlow `tf.constant` objects.
+                - `'pt'`: Return PyTorch `torch.Tensor` objects.
+                - `'np'`: Return NumPy `np.ndarray` objects.
+                - `'jax'`: Return JAX `jnp.ndarray` objects.
+
+        Returns:
+            [`BatchFeature`]: A [`BatchFeature`] with the following fields:
+
+            - **input_ids** -- List of token ids to be fed to a model. Returned when `text` is not `None`.
+            - **attention_mask** -- List of indices specifying which tokens should be attended to by the model (when
+              `return_attention_mask=True` or if *"attention_mask"* is in `self.model_input_names` and if `text` is not
+              `None`).
+            - **pixel_values** -- Pixel values to be fed to a model. Returned when `images` is not `None`.
+            - **pixel_values_videos** -- Pixel values of videos to be fed to a model. Returned when `videos` is not `None`.
+            - **image_grid_thw** -- List of image 3D grid in LLM. Returned when `images` is not `None`.
+            - **video_grid_thw** -- List of video 3D grid in LLM. Returned when `videos` is not `None`.
+        """
+        output_kwargs = self._merge_kwargs(
+            DreamVLProcessorKwargs,
+            tokenizer_init_kwargs=self.tokenizer.init_kwargs,
+            **kwargs,
+        )
+        if images is not None:
+            image_inputs = self.image_processor(images=images, videos=None, **output_kwargs["images_kwargs"])
+            image_grid_thw = image_inputs["image_grid_thw"]
+        else:
+            image_inputs = {}
+            image_grid_thw = None
+
+        if videos is not None:
+            videos_inputs = self.image_processor(images=None, videos=videos, **output_kwargs["videos_kwargs"])
+            video_grid_thw = videos_inputs["video_grid_thw"]
+        else:
+            videos_inputs = {}
+            video_grid_thw = None
+
+        if not isinstance(text, list):
+            text = [text]
+
+        if image_grid_thw is not None:
+            merge_length = self.image_processor.merge_size ** 2
+            index = 0
+            for i in range(len(text)):
+                while "<|image_pad|>" in text[i]:
+                    text[i] = text[i].replace(
+                        "<|image_pad|>", "<|placeholder|>" * (image_grid_thw[index].prod() // merge_length), 1
+                    )
+                    index += 1
+                text[i] = text[i].replace("<|placeholder|>", "<|image_pad|>")
+
+        if video_grid_thw is not None:
+            merge_length = self.image_processor.merge_size ** 2
+            index = 0
+            for i in range(len(text)):
+                while "<|video_pad|>" in text[i]:
+                    text[i] = text[i].replace(
+                        "<|video_pad|>", "<|placeholder|>" * (video_grid_thw[index].prod() // merge_length), 1
+                    )
+                    index += 1
+                text[i] = text[i].replace("<|placeholder|>", "<|video_pad|>")
+
+        _ = output_kwargs["text_kwargs"].pop("padding_side", None)
+        text_inputs = self.tokenizer(text, **output_kwargs["text_kwargs"])
+
+        return BatchFeature(data={**text_inputs, **image_inputs, **videos_inputs})
+
+    def batch_decode(self, *args, **kwargs):
+        """
+        This method forwards all its arguments to DreamTokenizer's [`~PreTrainedTokenizer.batch_decode`]. Please
+        refer to the docstring of this method for more information.
+        """
+        return self.tokenizer.batch_decode(*args, **kwargs)
+
+    def decode(self, *args, **kwargs):
+        """
+        This method forwards all its arguments to DreamTokenizer's [`~PreTrainedTokenizer.decode`]. Please refer to
+        the docstring of this method for more information.
+        """
+        return self.tokenizer.decode(*args, **kwargs)
+
+    @property
+    def model_input_names(self):
+        tokenizer_input_names = self.tokenizer.model_input_names
+        image_processor_input_names = self.image_processor.model_input_names
+        return list(dict.fromkeys(tokenizer_input_names + image_processor_input_names))

processor_config.json

@@ -0,0 +1,6 @@
+{
+  "auto_map": {
+    "AutoProcessor": "processing_dreamvl.DreamVLProcessor"
+  },
+  "processor_class": "DreamVLProcessor"
+}

special_tokens_map.json

@@ -0,0 +1,34 @@
+{
+  "additional_special_tokens": [
+    "<|beginoftext|>",
+    "<|mask|>"
+  ],
+  "bos_token": {
+    "content": "<|beginoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "mask_token": {
+    "content": "<|mask|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenization_dream.py

@@ -0,0 +1,331 @@
+# coding=utf-8
+# Copyright 2024 The Dream team, HKUNLP Group and The HuggingFace Inc. team. All rights reserved.
+#
+# This code is based on Qwen's implementations in this library.
+# 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.
+"""Tokenization classes for Dream."""
+
+import json
+import os
+import unicodedata
+from functools import lru_cache
+from typing import Optional, Tuple
+
+import regex as re
+
+from transformers.tokenization_utils import AddedToken, PreTrainedTokenizer
+from transformers.utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+VOCAB_FILES_NAMES = {
+    "vocab_file": "vocab.json",
+    "merges_file": "merges.txt",
+}
+
+
+MAX_MODEL_INPUT_SIZES = {"dream/dream-tokenizer": 32768}
+
+PRETOKENIZE_REGEX = r"""(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\r\n\p{L}\p{N}]?\p{L}+|\p{N}| ?[^\s\p{L}\p{N}]+[\r\n]*|\s*[\r\n]+|\s+(?!\S)|\s+"""
+
+
+@lru_cache()
+# Copied from transformers.models.gpt2.tokenization_gpt2.bytes_to_unicode
+def bytes_to_unicode():
+    """
+    Returns list of utf-8 byte and a mapping to unicode strings. We specifically avoids mapping to whitespace/control
+    characters the bpe code barfs on.
+    The reversible bpe codes work on unicode strings. This means you need a large # of unicode characters in your vocab
+    if you want to avoid UNKs. When you're at something like a 10B token dataset you end up needing around 5K for
+    decent coverage. This is a significant percentage of your normal, say, 32K bpe vocab. To avoid that, we want lookup
+    tables between utf-8 bytes and unicode strings.
+    """
+    bs = (
+        list(range(ord("!"), ord("~") + 1)) + list(range(ord("¡"), ord("¬") + 1)) + list(range(ord("®"), ord("ÿ") + 1))
+    )
+    cs = bs[:]
+    n = 0
+    for b in range(2**8):
+        if b not in bs:
+            bs.append(b)
+            cs.append(2**8 + n)
+            n += 1
+    cs = [chr(n) for n in cs]
+    return dict(zip(bs, cs))
+
+
+# Copied from transformers.models.gpt2.tokenization_gpt2.get_pairs
+def get_pairs(word):
+    """
+    Return set of symbol pairs in a word.
+    Word is represented as tuple of symbols (symbols being variable-length strings).
+    """
+    pairs = set()
+    prev_char = word[0]
+    for char in word[1:]:
+        pairs.add((prev_char, char))
+        prev_char = char
+    return pairs
+
+
+class DreamTokenizer(PreTrainedTokenizer):
+    """
+    Construct a Dream tokenizer. Based on byte-level Byte-Pair-Encoding.
+    Same with GPT2Tokenizer, this tokenizer has been trained to treat spaces like parts of the tokens so a word will
+    be encoded differently whether it is at the beginning of the sentence (without space) or not:
+    ```python
+    >>> from transformers import AutoTokenizer
+    >>> tokenizer = AutoTokenizer.from_pretrained("Dream-org/Dream-v0-Base-7B", trust_remote_code=True)
+    >>> tokenizer("Hello world")["input_ids"]
+    [9707, 1879]
+    >>> tokenizer(" Hello world")["input_ids"]
+    [21927, 1879]
+    ```
+    This is expected.
+    You should not use GPT2Tokenizer instead, because of the different pretokenization rules.
+    This tokenizer inherits from [`PreTrainedTokenizer`] which contains most of the main methods. Users should refer to
+    this superclass for more information regarding those methods.
+    Args:
+        vocab_file (`str`):
+            Path to the vocabulary file.
+        merges_file (`str`):
+            Path to the merges file.
+        errors (`str`, *optional*, defaults to `"replace"`):
+            Paradigm to follow when decoding bytes to UTF-8. See
+            [bytes.decode](https://docs.python.org/3/library/stdtypes.html#bytes.decode) for more information.
+        unk_token (`str`, *optional*, defaults to `"<|endoftext|>"`):
+            The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this
+            token instead.
+        bos_token (`str`, *optional*):
+            The beginning of sequence token. Not applicable for this tokenizer.
+        eos_token (`str`, *optional*, defaults to `"<|endoftext|>"`):
+            The end of sequence token.
+        pad_token (`str`, *optional*, defaults to `"<|endoftext|>"`):
+            The token used for padding, for example when batching sequences of different lengths.
+        clean_up_tokenization_spaces (`bool`, *optional*, defaults to `False`):
+            Whether or not the model should cleanup the spaces that were added when splitting the input text during the
+            tokenization process. Not applicable to this tokenizer, since tokenization does not add spaces.
+        split_special_tokens (`bool`, *optional*, defaults to `False`):
+            Whether or not the special tokens should be split during the tokenization process. The default behavior is
+            to not split special tokens. This means that if `<|endoftext|>` is the `eos_token`, then `tokenizer.tokenize("<|endoftext|>") =
+            ['<|endoftext|>`]. Otherwise, if `split_special_tokens=True`, then `tokenizer.tokenize("<|endoftext|>")` will be give `['<',
+            '|', 'endo', 'ft', 'ext', '|', '>']`. This argument is only supported for `slow` tokenizers for the moment.
+    """
+
+    vocab_files_names = VOCAB_FILES_NAMES
+    model_input_names = ["input_ids", "attention_mask"]
+
+    def __init__(
+        self,
+        vocab_file,
+        merges_file,
+        errors="replace",
+        unk_token="<|endoftext|>",
+        bos_token=None,
+        eos_token="<|endoftext|>",
+        pad_token="<|endoftext|>",
+        clean_up_tokenization_spaces=False,
+        split_special_tokens=False,
+        **kwargs,
+    ):
+        # Dream vocab does not contain control tokens; added tokens need to be special
+        bos_token = (
+            AddedToken(bos_token, lstrip=False, rstrip=False, special=True, normalized=False)
+            if isinstance(bos_token, str)
+            else bos_token
+        )
+        eos_token = (
+            AddedToken(eos_token, lstrip=False, rstrip=False, special=True, normalized=False)
+            if isinstance(eos_token, str)
+            else eos_token
+        )
+        unk_token = (
+            AddedToken(unk_token, lstrip=False, rstrip=False, special=True, normalized=False)
+            if isinstance(unk_token, str)
+            else unk_token
+        )
+        pad_token = (
+            AddedToken(pad_token, lstrip=False, rstrip=False, special=True, normalized=False)
+            if isinstance(pad_token, str)
+            else pad_token
+        )
+
+        with open(vocab_file, encoding="utf-8") as vocab_handle:
+            self.encoder = json.load(vocab_handle)
+        self.decoder = {v: k for k, v in self.encoder.items()}
+        self.errors = errors  # how to handle errors in decoding
+        self.byte_encoder = bytes_to_unicode()
+        self.byte_decoder = {v: k for k, v in self.byte_encoder.items()}
+        bpe_merges = []
+        with open(merges_file, encoding="utf-8") as merges_handle:
+            for i, line in enumerate(merges_handle):
+                line = line.strip()
+                if (i == 0 and line.startswith("#version:")) or not line:
+                    continue
+                bpe_merges.append(tuple(line.split()))
+        self.bpe_ranks = dict(zip(bpe_merges, range(len(bpe_merges))))
+        # NOTE: the cache can grow without bound and will get really large for long running processes
+        # (esp. for texts of language that do not use space between word, e.g. Chinese); technically
+        # not a memory leak but appears as one.
+        # GPT2Tokenizer has the same problem, so let's be consistent.
+        self.cache = {}
+
+        self.pat = re.compile(PRETOKENIZE_REGEX)
+
+        if kwargs.get("add_prefix_space", False):
+            logger.warning_once(
+                f"{self.__class__.__name} does not support `add_prefix_space`, setting it to True has no effect."
+            )
+
+        super().__init__(
+            errors=errors,
+            bos_token=bos_token,
+            eos_token=eos_token,
+            pad_token=pad_token,
+            unk_token=unk_token,
+            clean_up_tokenization_spaces=clean_up_tokenization_spaces,
+            split_special_tokens=split_special_tokens,
+            **kwargs,
+        )
+
+    @property
+    def vocab_size(self) -> int:
+        return len(self.encoder)
+
+    # Copied from transformers.models.gpt2.tokenization_gpt2.GPT2Tokenizer.get_vocab
+    def get_vocab(self):
+        return dict(self.encoder, **self.added_tokens_encoder)
+
+    # Copied from transformers.models.gpt2.tokenization_gpt2.GPT2Tokenizer.bpe
+    def bpe(self, token):
+        if token in self.cache:
+            return self.cache[token]
+        word = tuple(token)
+        pairs = get_pairs(word)
+
+        if not pairs:
+            return token
+
+        while True:
+            bigram = min(pairs, key=lambda pair: self.bpe_ranks.get(pair, float("inf")))
+            if bigram not in self.bpe_ranks:
+                break
+            first, second = bigram
+            new_word = []
+            i = 0
+            while i < len(word):
+                try:
+                    j = word.index(first, i)
+                except ValueError:
+                    new_word.extend(word[i:])
+                    break
+                else:
+                    new_word.extend(word[i:j])
+                    i = j
+
+                if word[i] == first and i < len(word) - 1 and word[i + 1] == second:
+                    new_word.append(first + second)
+                    i += 2
+                else:
+                    new_word.append(word[i])
+                    i += 1
+            new_word = tuple(new_word)
+            word = new_word
+            if len(word) == 1:
+                break
+            else:
+                pairs = get_pairs(word)
+        word = " ".join(word)
+        self.cache[token] = word
+        return word
+
+    # Copied from transformers.models.gpt2.tokenization_gpt2.GPT2Tokenizer._tokenize
+    def _tokenize(self, text):
+        """Tokenize a string."""
+        bpe_tokens = []
+        for token in re.findall(self.pat, text):
+            token = "".join(
+                self.byte_encoder[b] for b in token.encode("utf-8")
+            )  # Maps all our bytes to unicode strings, avoiding control tokens of the BPE (spaces in our case)
+            bpe_tokens.extend(bpe_token for bpe_token in self.bpe(token).split(" "))
+        return bpe_tokens
+
+    # Copied from transformers.models.gpt2.tokenization_gpt2.GPT2Tokenizer._convert_token_to_id
+    def _convert_token_to_id(self, token):
+        """Converts a token (str) in an id using the vocab."""
+        return self.encoder.get(token, self.encoder.get(self.unk_token))
+
+    # Copied from transformers.models.gpt2.tokenization_gpt2.GPT2Tokenizer._convert_id_to_token
+    def _convert_id_to_token(self, index):
+        """Converts an index (integer) in a token (str) using the vocab."""
+        return self.decoder.get(index)
+
+    # Copied from transformers.models.gpt2.tokenization_gpt2.GPT2Tokenizer.convert_tokens_to_string
+    def convert_tokens_to_string(self, tokens):
+        """Converts a sequence of tokens (string) in a single string."""
+        text = "".join(tokens)
+        text = bytearray([self.byte_decoder[c] for c in text]).decode("utf-8", errors=self.errors)
+        return text
+
+    def decode(
+        self,
+        token_ids,
+        skip_special_tokens: bool = False,
+        clean_up_tokenization_spaces: Optional[bool] = False,
+        spaces_between_special_tokens: bool = False,
+        **kwargs,
+    ) -> str:
+        # `spaces_between_special_tokens` defaults to True for _decode in slow tokenizers
+        # and cannot be configured elsewhere, but it should default to False for DreamTokenizer
+        return super().decode(
+            token_ids,
+            skip_special_tokens=skip_special_tokens,
+            clean_up_tokenization_spaces=clean_up_tokenization_spaces,
+            spaces_between_special_tokens=spaces_between_special_tokens,
+            **kwargs,
+        )
+
+    # Copied from transformers.models.gpt2.tokenization_gpt2.GPT2Tokenizer.save_vocabulary
+    def save_vocabulary(self, save_directory: str, filename_prefix: Optional[str] = None) -> Tuple[str]:
+        if not os.path.isdir(save_directory):
+            logger.error(f"Vocabulary path ({save_directory}) should be a directory")
+            return
+        vocab_file = os.path.join(
+            save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["vocab_file"]
+        )
+        merge_file = os.path.join(
+            save_directory, (filename_prefix + "-" if filename_prefix else "") + VOCAB_FILES_NAMES["merges_file"]
+        )
+
+        with open(vocab_file, "w", encoding="utf-8") as f:
+            f.write(json.dumps(self.encoder, indent=2, sort_keys=True, ensure_ascii=False) + "\n")
+
+        index = 0
+        with open(merge_file, "w", encoding="utf-8") as writer:
+            writer.write("#version: 0.2\n")
+            for bpe_tokens, token_index in sorted(self.bpe_ranks.items(), key=lambda kv: kv[1]):
+                if index != token_index:
+                    logger.warning(
+                        f"Saving vocabulary to {merge_file}: BPE merge indices are not consecutive."
+                        " Please check that the tokenizer is not corrupted!"
+                    )
+                    index = token_index
+                writer.write(" ".join(bpe_tokens) + "\n")
+                index += 1
+
+        return vocab_file, merge_file
+
+    def prepare_for_tokenization(self, text, **kwargs):
+        text = unicodedata.normalize("NFC", text)
+        return (text, kwargs)

tokenizer_config.json

@@ -0,0 +1,223 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151665": {
+      "content": "<|beginoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151666": {
+      "content": "<|mask|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<|beginoftext|>",
+    "<|mask|>"
+  ],
+  "auto_map": {
+    "AutoProcessor": "processing_dreamvl.DreamVLProcessor",
+    "AutoTokenizer": [
+      "tokenization_dream.DreamTokenizer",
+      null
+    ]
+  },
+  "bos_token": "<|beginoftext|>",
+  "chat_template": "{% set image_count = namespace(value=0) %}{% set video_count = namespace(value=0) %}{% for message in messages %}{% if loop.first and message['role'] != 'system' %}<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n{% endif %}<|im_start|>{{ message['role'] }}\n{% if message['content'] is string %}{{ message['content'] }}<|im_end|>\n{% else %}{% for content in message['content'] %}{% if content['type'] == 'image' or 'image' in content or 'image_url' in content %}{% set image_count.value = image_count.value + 1 %}{% if add_vision_id %}Picture {{ image_count.value }}: {% endif %}<|image_pad|>{% elif content['type'] == 'video' or 'video' in content %}{% set video_count.value = video_count.value + 1 %}{% if add_vision_id %}Video {{ video_count.value }}: {% endif %}<|video_pad|>{% elif 'text' in content %}{{ content['text'] }}{% endif %}{% endfor %}<|im_end|>\n{% endif %}{% endfor %}{% if add_generation_prompt %}<|im_start|>assistant\n{% endif %}",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "mask_token": "<|mask|>",
+  "model_max_length": 8192,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "processor_class": "DreamVLProcessor",
+  "split_special_tokens": false,
+  "tokenizer_class": "DreamTokenizer",
+  "unk_token": null
+}