Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,3 +1,68 @@
----
-license: apache-2.0
----
+---
+license: apache-2.0
+tags:
+- diffusion
+- vision-language
+- qwen2.5
+- siglip
+---
+
+# DiffusionVL-Qwen2.5
+
+DiffusionVL model with SigLIP vision encoder, PoolerProjector, and Qwen2.5 LLM with BD3LM diffusion-based generation.
+
+## Usage
+
+```python
+from transformers import AutoModelForCausalLM, AutoProcessor
+import torch
+from PIL import Image
+
+# Load model
+model = AutoModelForCausalLM.from_pretrained(
+    "path/to/model",
+    torch_dtype=torch.bfloat16,
+    device_map="auto",
+    trust_remote_code=True
+)
+
+# Load processor
+processor = AutoProcessor.from_pretrained("path/to/model", trust_remote_code=True)
+
+# Prepare inputs
+image = Image.open("image.jpg").convert("RGB")
+messages = [
+    {"role": "user", "content": [
+        {"type": "image"},
+        {"type": "text", "text": "Describe this image."}
+    ]}
+]
+text = processor.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+inputs = processor(text=[text], images=[image], return_tensors="pt", padding=True)
+inputs = {k: v.to(model.device) if hasattr(v, 'to') else v for k, v in inputs.items()}
+
+# Generate
+output_ids = model.generate(
+    inputs=inputs["input_ids"],
+    images=inputs.get("pixel_values"),
+    gen_length=256,
+    steps=8,
+    temperature=0.0,
+    remasking_strategy="low_confidence_static",
+)
+
+# Decode
+output_text = processor.decode(output_ids[0], skip_special_tokens=True)
+print(output_text)
+```
+
+## Model Configuration
+
+- **Architecture**: DiffusionVL_Qwen2_5_ForConditionalGeneration
+- **Vision Encoder**: SigLIP (384x384, patch_size=14)
+- **MM Projector**: PoolerProjector (Conv2d + MLP)
+- **LLM**: Qwen2.5 (standard RoPE)
+- **BD3LM Enabled**: True
+- **Block Size**: 8
+- **Hidden Size**: 3584
+- **Num Layers**: 28

added_tokens.json

@@ -0,0 +1,24 @@
+{
+  "</tool_call>": 151658,
+  "<tool_call>": 151657,
+  "<|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,
+  "<|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.jinja

@@ -0,0 +1,7 @@
+{% 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
+You are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>
+{% endif %}<|im_start|>{{ message['role'] }}
+{% if message['content'] is string %}{{ message['content'] }}<|im_end|>
+{% 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 %}<image>{% elif content['type'] == 'video' or 'video' in content %}{% set video_count.value = video_count.value + 1 %}<video>{% elif 'text' in content %}{{ content['text'] }}{% endif %}{% endfor %}<|im_end|>
+{% endif %}{% endfor %}{% if add_generation_prompt %}<|im_start|>assistant
+{% endif %}

config.json

@@ -0,0 +1,273 @@
+{
+  "add_faster_video": false,
+  "add_time_instruction": false,
+  "anneal_start_block_size": 1,
+  "architectures": [
+    "DiffusionVL_Qwen2_5_ForConditionalGeneration"
+  ],
+  "attention_dropout": 0.0,
+  "bd3lm_antithetic_sampling": true,
+  "bd3lm_attn_backend": "sdpa",
+  "bd3lm_block_aligned_eos": true,
+  "bd3lm_block_size": 8,
+  "bd3lm_complementary_mask": false,
+  "bd3lm_cross_attn": true,
+  "bd3lm_ignore_bos": true,
+  "bd3lm_noise_granularity": "block",
+  "bd3lm_noise_type": "loglinear",
+  "bd3lm_parameterization": "subs",
+  "bd3lm_resample": false,
+  "bd3lm_sampling_eps_max": 1.0,
+  "bd3lm_sampling_eps_min": 0.001,
+  "bd3lm_time_conditioning": false,
+  "bd3lm_token_shift_prediction": false,
+  "bd3lm_var_min": true,
+  "bos_token_id": 151643,
+  "enable_bd3lm": true,
+  "enable_block_size_annealing": false,
+  "enable_noise_level_annealing": false,
+  "eos_token_id": 151645,
+  "faster_token_stride": 10,
+  "force_sample": false,
+  "hidden_act": "silu",
+  "hidden_size": 3584,
+  "image_aspect_ratio": "anyres_max_4",
+  "image_crop_resolution": null,
+  "image_grid_pinpoints": [
+    [
+      384,
+      384
+    ],
+    [
+      384,
+      768
+    ],
+    [
+      384,
+      1152
+    ],
+    [
+      384,
+      1536
+    ],
+    [
+      384,
+      1920
+    ],
+    [
+      384,
+      2304
+    ],
+    [
+      768,
+      384
+    ],
+    [
+      768,
+      768
+    ],
+    [
+      768,
+      1152
+    ],
+    [
+      768,
+      1536
+    ],
+    [
+      768,
+      1920
+    ],
+    [
+      768,
+      2304
+    ],
+    [
+      1152,
+      384
+    ],
+    [
+      1152,
+      768
+    ],
+    [
+      1152,
+      1152
+    ],
+    [
+      1152,
+      1536
+    ],
+    [
+      1152,
+      1920
+    ],
+    [
+      1152,
+      2304
+    ],
+    [
+      1536,
+      384
+    ],
+    [
+      1536,
+      768
+    ],
+    [
+      1536,
+      1152
+    ],
+    [
+      1536,
+      1536
+    ],
+    [
+      1536,
+      1920
+    ],
+    [
+      1536,
+      2304
+    ],
+    [
+      1920,
+      384
+    ],
+    [
+      1920,
+      768
+    ],
+    [
+      1920,
+      1152
+    ],
+    [
+      1920,
+      1536
+    ],
+    [
+      1920,
+      1920
+    ],
+    [
+      1920,
+      2304
+    ],
+    [
+      2304,
+      384
+    ],
+    [
+      2304,
+      768
+    ],
+    [
+      2304,
+      1152
+    ],
+    [
+      2304,
+      1536
+    ],
+    [
+      2304,
+      1920
+    ],
+    [
+      2304,
+      2304
+    ]
+  ],
+  "image_split_resolution": null,
+  "initializer_range": 0.02,
+  "intermediate_size": 18944,
+  "layer_types": [
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention",
+    "full_attention"
+  ],
+  "max_pixels": 262144,
+  "max_position_embeddings": 32768,
+  "max_window_layers": 28,
+  "min_pixels": 147456,
+  "mm_hidden_size": 1152,
+  "mm_newline_position": "grid",
+  "mm_patch_merge_type": "spatial_unpad",
+  "mm_projector_lr": null,
+  "mm_projector_type": "mlp2x_gelu",
+  "mm_resampler_type": null,
+  "mm_spatial_pool_mode": "bilinear",
+  "mm_spatial_pool_stride": null,
+  "mm_tunable_parts": "mm_vision_tower,mm_mlp_adapter,mm_language_model",
+  "mm_use_im_patch_token": false,
+  "mm_use_im_start_end": false,
+  "mm_vision_select_feature": "patch",
+  "mm_vision_select_layer": -2,
+  "mm_vision_tower": "/data/minimax-dialogue/users/qingke/results/hf_models/siglip2-so400m-patch14-384",
+  "mm_vision_tower_lr": 2e-06,
+  "model_max_length": 8192,
+  "model_type": "diffusionvl_qwen2_5",
+  "num_attention_heads": 28,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 4,
+  "pos_skipping_range": 4096,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 1000000.0,
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "tokenizer_model_max_length": 8192,
+  "tokenizer_padding_side": "right",
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.55.0",
+  "use_cache": true,
+  "use_mm_proj": true,
+  "use_pos_skipping": false,
+  "use_sliding_window": false,
+  "vision_tower_pretrained": null,
+  "vocab_size": 152064,
+  "mask_token_id": 151671,
+  "vision_config": {
+    "hidden_size": 1152,
+    "intermediate_size": 4304,
+    "num_hidden_layers": 26,
+    "num_attention_heads": 16,
+    "num_channels": 3,
+    "image_size": 384,
+    "patch_size": 14,
+    "hidden_act": "gelu_pytorch_tanh",
+    "layer_norm_eps": 1e-06,
+    "attention_dropout": 0.0
+  },
+  "auto_map": {
+    "AutoConfig": "configuration_diffusionvl_qwen2_5.DiffusionVL_Qwen2_5_Config",
+    "AutoModelForCausalLM": "modeling_diffusionvl_qwen2_5.DiffusionVL_Qwen2_5_ForConditionalGeneration",
+    "AutoProcessor": "processing_diffusionvl_qwen2_5.DiffusionVL_Qwen2_5_Processor"
+  }
+}

configuration_diffusionvl_qwen2_5.py

@@ -0,0 +1,189 @@
+# coding=utf-8
+# Copyright 2025 The HustVL 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.
+"""DiffusionVL-Qwen2.5 model configuration."""
+
+from typing import List, Optional, Union
+
+from transformers.configuration_utils import PretrainedConfig
+
+
+class DiffusionVL_Qwen2_5_VisionConfig(PretrainedConfig):
+    """
+    Configuration for SigLIP vision encoder used in DiffusionVL-Qwen2.5.
+
+    Args:
+        hidden_size: Dimension of the encoder layers (1152 for SigLIP-SO400M).
+        intermediate_size: Dimension of the MLP layers.
+        num_hidden_layers: Number of transformer layers.
+        num_attention_heads: Number of attention heads.
+        num_channels: Number of input channels.
+        image_size: Input image resolution.
+        patch_size: Patch size for patch embedding.
+        hidden_act: Activation function.
+        layer_norm_eps: Layer normalization epsilon.
+        attention_dropout: Attention dropout probability.
+    """
+
+    model_type = "diffusionvl_qwen2_5_vision"
+    base_config_key = "vision_config"
+
+    def __init__(
+        self,
+        hidden_size: int = 1152,
+        intermediate_size: int = 4304,
+        num_hidden_layers: int = 26,  # SigLIP uses 27 layers, but last one is removed
+        num_attention_heads: int = 16,
+        num_channels: int = 3,
+        image_size: int = 384,
+        patch_size: int = 14,
+        hidden_act: str = "gelu_pytorch_tanh",
+        layer_norm_eps: float = 1e-6,
+        attention_dropout: float = 0.0,
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.num_channels = num_channels
+        self.image_size = image_size
+        self.patch_size = patch_size
+        self.hidden_act = hidden_act
+        self.layer_norm_eps = layer_norm_eps
+        self.attention_dropout = attention_dropout
+
+
+class DiffusionVL_Qwen2_5_Config(PretrainedConfig):
+    """
+    Configuration for DiffusionVL-Qwen2.5 model.
+
+    This model uses:
+    - SigLIP as the vision encoder (external ViT)
+    - PoolerProjector as the MM projector (Conv2d + MLP)
+    - Qwen2.5 as the LLM backbone (standard RoPE, not M-RoPE)
+    - BD3LM for diffusion-based generation
+
+    Args:
+        vocab_size: Vocabulary size.
+        hidden_size: Dimension of the hidden representations.
+        intermediate_size: Dimension of the MLP representations.
+        num_hidden_layers: Number of hidden layers.
+        num_attention_heads: Number of attention heads.
+        num_key_value_heads: Number of key-value heads for GQA.
+        hidden_act: Activation function.
+        max_position_embeddings: Maximum sequence length.
+        initializer_range: Standard deviation for weight initialization.
+        rms_norm_eps: Epsilon for RMS normalization.
+        use_cache: Whether to use KV cache.
+        tie_word_embeddings: Whether to tie input and output embeddings.
+        attention_dropout: Attention dropout probability.
+        vision_config: Vision encoder configuration.
+        mm_hidden_size: Vision encoder hidden size for projector.
+        enable_bd3lm: Whether to enable BD3LM.
+        bd3lm_block_size: Block size for BD3LM.
+        bd3lm_cross_attn: Whether to use cross-attention in BD3LM.
+        mask_token_id: Token ID for mask token.
+        rope_theta: RoPE base period.
+        sliding_window: Sliding window size for attention.
+    """
+
+    model_type = "diffusionvl_qwen2_5"
+    sub_configs = {"vision_config": DiffusionVL_Qwen2_5_VisionConfig}
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    def __init__(
+        self,
+        # Text model parameters (Qwen2.5 compatible)
+        vocab_size: int = 152064,
+        hidden_size: int = 3584,
+        intermediate_size: int = 18944,
+        num_hidden_layers: int = 28,
+        num_attention_heads: int = 28,
+        num_key_value_heads: int = 4,
+        hidden_act: str = "silu",
+        max_position_embeddings: int = 32768,
+        initializer_range: float = 0.02,
+        rms_norm_eps: float = 1e-6,
+        use_cache: bool = True,
+        tie_word_embeddings: bool = False,
+        attention_dropout: float = 0.0,
+        # Vision configuration
+        vision_config: Optional[Union[DiffusionVL_Qwen2_5_VisionConfig, dict]] = None,
+        # MM projector
+        mm_hidden_size: int = 1152,  # SigLIP hidden size
+        # BD3LM diffusion parameters
+        enable_bd3lm: bool = True,
+        bd3lm_block_size: int = 8,
+        bd3lm_cross_attn: bool = True,
+        bd3lm_antithetic_sampling: bool = True,
+        bd3lm_sampling_eps_min: float = 1e-3,
+        bd3lm_sampling_eps_max: float = 1.0,
+        mask_token_id: int = 151671,
+        # RoPE parameters (standard RoPE, not M-RoPE)
+        rope_theta: float = 1000000.0,
+        rope_scaling: Optional[dict] = None,
+        # Sliding window attention
+        sliding_window: int = 32768,
+        max_window_layers: int = 28,
+        use_sliding_window: bool = False,
+        **kwargs,
+    ):
+        # Text model configuration
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.num_key_value_heads = num_key_value_heads
+        self.hidden_act = hidden_act
+        self.max_position_embeddings = max_position_embeddings
+        self.initializer_range = initializer_range
+        self.rms_norm_eps = rms_norm_eps
+        self.use_cache = use_cache
+        self.attention_dropout = attention_dropout
+        self.rope_theta = rope_theta
+        self.rope_scaling = rope_scaling
+        self.sliding_window = sliding_window
+        self.max_window_layers = max_window_layers
+        self.use_sliding_window = use_sliding_window
+
+        # Vision configuration
+        if vision_config is None:
+            self.vision_config = DiffusionVL_Qwen2_5_VisionConfig()
+        elif isinstance(vision_config, dict):
+            self.vision_config = DiffusionVL_Qwen2_5_VisionConfig(**vision_config)
+        elif isinstance(vision_config, DiffusionVL_Qwen2_5_VisionConfig):
+            self.vision_config = vision_config
+        else:
+            self.vision_config = DiffusionVL_Qwen2_5_VisionConfig()
+
+        # MM projector
+        self.mm_hidden_size = mm_hidden_size
+
+        # BD3LM diffusion configuration
+        self.enable_bd3lm = enable_bd3lm
+        self.bd3lm_block_size = bd3lm_block_size
+        self.bd3lm_cross_attn = bd3lm_cross_attn
+        self.bd3lm_antithetic_sampling = bd3lm_antithetic_sampling
+        self.bd3lm_sampling_eps_min = bd3lm_sampling_eps_min
+        self.bd3lm_sampling_eps_max = bd3lm_sampling_eps_max
+        self.mask_token_id = mask_token_id
+
+        super().__init__(tie_word_embeddings=tie_word_embeddings, **kwargs)
+
+
+__all__ = ["DiffusionVL_Qwen2_5_Config", "DiffusionVL_Qwen2_5_VisionConfig"]

model-00001-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96613e0135affeb37783597f8e0dd15985bfa4a5eb4cae4f6059e3cc2f22f693
+size 4877668008

model-00002-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee0edc9a3304655a5eaed32f7198cc2b037c8d620e6fb2985e6f383002edc061
+size 4932751008

model-00003-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b65fc332179e0afa2f9f154afeecfeea0ea7c4ff1f82b73281a8dbafca5c26a5
+size 4994571904

model-00004-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:565699fa87e0b2bc0eea6a3242f6f64fa92550a0f35bbbad1e8128bf28674ab2
+size 1255812224

model.safetensors.index.json

@@ -0,0 +1,773 @@
+{
+  "metadata": {
+    "total_parameters": 8030348832,
+    "total_size": 16060697672
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00004-of-00004.safetensors",
+    "model.embed_tokens.weight": "model-00001-of-00004.safetensors",
+    "model.image_newline": "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-00002-of-00004.safetensors",
+    "model.layers.16.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00002-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-00002-of-00004.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.v_proj.weight": "model-00002-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-00002-of-00004.safetensors",
+    "model.layers.18.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.v_proj.weight": "model-00002-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-00003-of-00004.safetensors",
+    "model.layers.26.mlp.down_proj.weight": "model-00003-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-00003-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-00003-of-00004.safetensors",
+    "model.layers.27.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.v_proj.weight": "model-00003-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-00001-of-00004.safetensors",
+    "model.layers.5.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00001-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-00001-of-00004.safetensors",
+    "model.layers.6.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.v_proj.weight": "model-00001-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-00001-of-00004.safetensors",
+    "model.layers.8.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.8.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.8.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.8.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.8.self_attn.v_proj.weight": "model-00001-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.mm_projector.0.bias": "model-00004-of-00004.safetensors",
+    "model.mm_projector.0.weight": "model-00004-of-00004.safetensors",
+    "model.mm_projector.2.bias": "model-00004-of-00004.safetensors",
+    "model.mm_projector.2.weight": "model-00004-of-00004.safetensors",
+    "model.norm.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.embeddings.patch_embedding.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.embeddings.patch_embedding.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.embeddings.position_embedding.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.0.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.0.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.0.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.0.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.1.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.1.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.1.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.1.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.10.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.10.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.10.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.10.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.11.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.11.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.11.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.11.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.12.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.12.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.12.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.12.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.13.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.13.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.13.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.13.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.14.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.14.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.14.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.14.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.15.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.15.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.15.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.15.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.16.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.16.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.16.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.16.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.17.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.17.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.17.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.17.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.18.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.18.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.18.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.18.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.19.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.19.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.19.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.19.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.2.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.2.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.2.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.2.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.20.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.20.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.20.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.20.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.21.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.21.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.21.layer_norm2.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.21.layer_norm2.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.22.layer_norm1.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.22.layer_norm1.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.22.layer_norm2.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.22.layer_norm2.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.23.layer_norm1.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.23.layer_norm1.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.23.layer_norm2.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.23.layer_norm2.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.24.layer_norm1.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.24.layer_norm1.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.24.layer_norm2.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.24.layer_norm2.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.25.layer_norm1.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.25.layer_norm1.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.25.layer_norm2.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.25.layer_norm2.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.3.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.3.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.3.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.3.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.4.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.4.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.4.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.4.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.5.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.5.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.5.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.5.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.6.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.6.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.6.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.6.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.7.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.7.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.7.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.7.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.8.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.8.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.8.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.8.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.9.layer_norm1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.9.layer_norm1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.9.layer_norm2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.9.layer_norm2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.post_layernorm.bias": "model-00004-of-00004.safetensors",
+    "model.vision_tower.vision_tower.vision_model.post_layernorm.weight": "model-00004-of-00004.safetensors"
+  }
+}

modeling_diffusionvl_qwen2_5.py

@@ -0,0 +1,1225 @@
+# coding=utf-8
+# Copyright 2025 The HustVL 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.
+
+"""
+DiffusionVL-Qwen2.5 model implementation.
+
+This model uses:
+- SigLIP as the vision encoder (external ViT)
+- mlp2x_gelu as the MM projector (2-layer MLP with GELU)
+- Qwen2.5 as the LLM backbone (standard RoPE)
+- BD3LM for diffusion-based generation
+"""
+
+import math
+from typing import Callable, Dict, List, Optional, Tuple, Union
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from transformers.activations import ACT2FN
+from transformers.modeling_utils import PreTrainedModel
+from transformers.cache_utils import Cache, DynamicCache
+from transformers.modeling_outputs import BaseModelOutputWithPast, CausalLMOutputWithPast
+from transformers.utils import logging
+from transformers.modeling_utils import ALL_ATTENTION_FUNCTIONS
+
+from .configuration_diffusionvl_qwen2_5 import DiffusionVL_Qwen2_5_Config, DiffusionVL_Qwen2_5_VisionConfig
+
+logger = logging.get_logger(__name__)
+
+IMAGE_TOKEN_INDEX = -200
+
+
+# ============================================================================
+# Image Processing Utilities (matching training code)
+# ============================================================================
+
+def select_best_resolution(original_size, possible_resolutions):
+    """
+    Selects the best resolution from a list of possible resolutions based on the original size.
+    """
+    original_width, original_height = original_size
+    best_fit = None
+    max_effective_resolution = 0
+    min_wasted_resolution = float("inf")
+
+    for width, height in possible_resolutions:
+        scale = min(width / original_width, height / original_height)
+        downscaled_width, downscaled_height = int(original_width * scale), int(original_height * scale)
+        effective_resolution = min(downscaled_width * downscaled_height, original_width * original_height)
+        wasted_resolution = (width * height) - effective_resolution
+
+        if effective_resolution > max_effective_resolution or (effective_resolution == max_effective_resolution and wasted_resolution < min_wasted_resolution):
+            max_effective_resolution = effective_resolution
+            min_wasted_resolution = wasted_resolution
+            best_fit = (width, height)
+
+    return best_fit
+
+
+def get_anyres_image_grid_shape(image_size, grid_pinpoints, patch_size):
+    """
+    Calculate the shape of the image patch grid after the preprocessing for images of any resolution.
+    """
+    import re
+    import ast
+
+    if isinstance(grid_pinpoints, str) and "x" in grid_pinpoints:
+        assert patch_size in [224, 336, 384, 448, 512], "patch_size should be in [224, 336, 384, 448, 512]"
+        matches = re.findall(r"\((\d+)x(\d+)\)", grid_pinpoints)
+        range_start = tuple(map(int, matches[0]))
+        range_end = tuple(map(int, matches[-1]))
+        grid_pinpoints = [(i, j) for i in range(range_start[0], range_end[0] + 1) for j in range(range_start[1], range_end[1] + 1)]
+        grid_pinpoints = [[dim * patch_size for dim in pair] for pair in grid_pinpoints]
+    if type(grid_pinpoints) is list:
+        possible_resolutions = grid_pinpoints
+    else:
+        possible_resolutions = ast.literal_eval(grid_pinpoints)
+    width, height = select_best_resolution(image_size, possible_resolutions)
+    return width // patch_size, height // patch_size
+
+
+def unpad_image(tensor, original_size):
+    """
+    Unpads a PyTorch tensor of a padded and resized image.
+
+    Args:
+        tensor: The image tensor in CxHxW format.
+        original_size: The original size of the image (width, height).
+
+    Returns:
+        The unpadded image tensor.
+    """
+    original_width, original_height = original_size
+    current_height, current_width = tensor.shape[1:]
+
+    original_aspect_ratio = original_width / original_height
+    current_aspect_ratio = current_width / current_height
+
+    if original_aspect_ratio > current_aspect_ratio:
+        scale_factor = current_width / original_width
+        new_height = int(original_height * scale_factor)
+        padding = (current_height - new_height) // 2
+        unpadded_tensor = tensor[:, padding : current_height - padding, :]
+    else:
+        scale_factor = current_height / original_height
+        new_width = int(original_width * scale_factor)
+        padding = (current_width - new_width) // 2
+        unpadded_tensor = tensor[:, :, padding : current_width - padding]
+
+    return unpadded_tensor
+
+
+# ============================================================================
+# Vision Encoder (SigLIP)
+# ============================================================================
+
+class SigLipVisionEmbeddings(nn.Module):
+    """Patch embedding for SigLIP vision encoder."""
+
+    def __init__(self, config: DiffusionVL_Qwen2_5_VisionConfig):
+        super().__init__()
+        self.config = config
+        self.embed_dim = config.hidden_size
+        self.image_size = config.image_size
+        self.patch_size = config.patch_size
+
+        self.patch_embedding = nn.Conv2d(
+            in_channels=config.num_channels,
+            out_channels=self.embed_dim,
+            kernel_size=self.patch_size,
+            stride=self.patch_size,
+            padding="valid",
+        )
+
+        self.num_patches = (self.image_size // self.patch_size) ** 2
+        self.num_positions = self.num_patches
+        self.position_embedding = nn.Embedding(self.num_positions, self.embed_dim)
+        self.register_buffer("position_ids", torch.arange(self.num_positions).expand((1, -1)), persistent=False)
+
+    def forward(self, pixel_values: torch.FloatTensor) -> torch.Tensor:
+        patch_embeds = self.patch_embedding(pixel_values)
+        embeddings = patch_embeds.flatten(2).transpose(1, 2)
+        embeddings = embeddings + self.position_embedding(self.position_ids)
+        return embeddings
+
+
+class SigLipAttention(nn.Module):
+    """Multi-headed attention for SigLIP."""
+
+    def __init__(self, config: DiffusionVL_Qwen2_5_VisionConfig):
+        super().__init__()
+        self.config = config
+        self.embed_dim = config.hidden_size
+        self.num_heads = config.num_attention_heads
+        self.head_dim = self.embed_dim // self.num_heads
+        self.scale = self.head_dim ** -0.5
+        self.dropout = config.attention_dropout
+
+        self.k_proj = nn.Linear(self.embed_dim, self.embed_dim)
+        self.v_proj = nn.Linear(self.embed_dim, self.embed_dim)
+        self.q_proj = nn.Linear(self.embed_dim, self.embed_dim)
+        self.out_proj = nn.Linear(self.embed_dim, self.embed_dim)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        output_attentions: Optional[bool] = False,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor]]:
+        batch_size, 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(batch_size, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        key_states = key_states.view(batch_size, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+        value_states = value_states.view(batch_size, q_len, self.num_heads, self.head_dim).transpose(1, 2)
+
+        attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) * self.scale
+
+        if attention_mask is not None:
+            attn_weights = attn_weights + attention_mask
+
+        attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype)
+        attn_weights = nn.functional.dropout(attn_weights, p=self.dropout, training=self.training)
+        attn_output = torch.matmul(attn_weights, value_states)
+
+        attn_output = attn_output.transpose(1, 2).contiguous()
+        attn_output = attn_output.reshape(batch_size, q_len, self.embed_dim)
+        attn_output = self.out_proj(attn_output)
+
+        return attn_output, attn_weights
+
+
+class SigLipMLP(nn.Module):
+    """MLP for SigLIP."""
+
+    def __init__(self, config: DiffusionVL_Qwen2_5_VisionConfig):
+        super().__init__()
+        self.config = config
+        self.activation_fn = ACT2FN[config.hidden_act]
+        self.fc1 = nn.Linear(config.hidden_size, config.intermediate_size)
+        self.fc2 = nn.Linear(config.intermediate_size, config.hidden_size)
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        hidden_states = self.fc1(hidden_states)
+        hidden_states = self.activation_fn(hidden_states)
+        hidden_states = self.fc2(hidden_states)
+        return hidden_states
+
+
+class SigLipEncoderLayer(nn.Module):
+    """Transformer encoder layer for SigLIP."""
+
+    def __init__(self, config: DiffusionVL_Qwen2_5_VisionConfig):
+        super().__init__()
+        self.embed_dim = config.hidden_size
+        self.self_attn = SigLipAttention(config)
+        self.layer_norm1 = nn.LayerNorm(self.embed_dim, eps=config.layer_norm_eps)
+        self.mlp = SigLipMLP(config)
+        self.layer_norm2 = nn.LayerNorm(self.embed_dim, eps=config.layer_norm_eps)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        output_attentions: Optional[bool] = False,
+    ) -> Tuple[torch.FloatTensor]:
+        residual = hidden_states
+        hidden_states = self.layer_norm1(hidden_states)
+        hidden_states, attn_weights = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            output_attentions=output_attentions,
+        )
+        hidden_states = residual + hidden_states
+
+        residual = hidden_states
+        hidden_states = self.layer_norm2(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        hidden_states = residual + hidden_states
+
+        outputs = (hidden_states,)
+        if output_attentions:
+            outputs += (attn_weights,)
+        return outputs
+
+
+class SigLipEncoder(nn.Module):
+    """Transformer encoder for SigLIP."""
+
+    def __init__(self, config: DiffusionVL_Qwen2_5_VisionConfig):
+        super().__init__()
+        self.config = config
+        self.layers = nn.ModuleList([SigLipEncoderLayer(config) for _ in range(config.num_hidden_layers)])
+
+    def forward(
+        self,
+        inputs_embeds: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+    ) -> Tuple:
+        hidden_states = inputs_embeds
+        all_hidden_states = () if output_hidden_states else None
+
+        for encoder_layer in self.layers:
+            if output_hidden_states:
+                all_hidden_states = all_hidden_states + (hidden_states,)
+            layer_outputs = encoder_layer(hidden_states, attention_mask, output_attentions)
+            hidden_states = layer_outputs[0]
+
+        if output_hidden_states:
+            all_hidden_states = all_hidden_states + (hidden_states,)
+
+        return hidden_states, all_hidden_states
+
+
+class SigLipVisionTransformer(nn.Module):
+    """SigLIP Vision Transformer."""
+
+    def __init__(self, config: DiffusionVL_Qwen2_5_VisionConfig):
+        super().__init__()
+        self.config = config
+        self.embeddings = SigLipVisionEmbeddings(config)
+        self.encoder = SigLipEncoder(config)
+        self.post_layernorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)
+
+    def forward(
+        self,
+        pixel_values: torch.FloatTensor,
+        output_hidden_states: Optional[bool] = True,
+    ) -> torch.Tensor:
+        hidden_states = self.embeddings(pixel_values)
+        hidden_states, all_hidden_states = self.encoder(
+            inputs_embeds=hidden_states,
+            output_hidden_states=output_hidden_states,
+        )
+        # Return the last hidden state (before post_layernorm, matching SigLIP behavior)
+        # Use hidden_states from the last layer
+        if output_hidden_states and all_hidden_states:
+            return all_hidden_states[-1]
+        return hidden_states
+
+
+class SigLipVisionModel(nn.Module):
+    """Wrapper to match training code structure: vision_model contains the transformer."""
+
+    def __init__(self, config: DiffusionVL_Qwen2_5_VisionConfig):
+        super().__init__()
+        self.vision_model = SigLipVisionTransformer(config)
+        self.config = config
+
+    def forward(self, pixel_values: torch.FloatTensor, output_hidden_states: bool = True) -> torch.Tensor:
+        return self.vision_model(pixel_values, output_hidden_states=output_hidden_states)
+
+
+class DiffusionVL_Qwen2_5_VisionTower(nn.Module):
+    """Vision tower wrapping SigLIP. Matches training code: vision_tower.vision_tower.vision_model.xxx"""
+
+    def __init__(self, config: DiffusionVL_Qwen2_5_VisionConfig):
+        super().__init__()
+        self.vision_tower = SigLipVisionModel(config)
+        self.config = config
+
+    @property
+    def dtype(self):
+        return self.vision_tower.vision_model.embeddings.patch_embedding.weight.dtype
+
+    @property
+    def device(self):
+        return self.vision_tower.vision_model.embeddings.patch_embedding.weight.device
+
+    def forward(self, pixel_values: torch.FloatTensor) -> torch.Tensor:
+        # Match training code: convert to vision tower dtype/device
+        pixel_values = pixel_values.to(device=self.device, dtype=self.dtype)
+        return self.vision_tower(pixel_values, output_hidden_states=True)
+
+
+# ============================================================================
+# MM Projector (mlp2x_gelu - matches training code)
+# ============================================================================
+
+def build_mm_projector(config: DiffusionVL_Qwen2_5_Config) -> nn.Module:
+    """
+    Build MM projector matching training code's mlp2x_gelu structure.
+
+    Structure:
+        0: nn.Linear(mm_hidden_size, hidden_size)  # 1152 -> 3584
+        1: nn.GELU()
+        2: nn.Linear(hidden_size, hidden_size)     # 3584 -> 3584
+    """
+    return nn.Sequential(
+        nn.Linear(config.mm_hidden_size, config.hidden_size),
+        nn.GELU(),
+        nn.Linear(config.hidden_size, config.hidden_size),
+    )
+
+
+# ============================================================================
+# LLM Components (Qwen2.5 based)
+# ============================================================================
+
+class DiffusionVL_Qwen2_5_RMSNorm(nn.Module):
+    def __init__(self, hidden_size, eps=1e-6):
+        super().__init__()
+        self.weight = nn.Parameter(torch.ones(hidden_size))
+        self.variance_epsilon = eps
+
+    def forward(self, hidden_states):
+        input_dtype = hidden_states.dtype
+        hidden_states = hidden_states.to(torch.float32)
+        variance = hidden_states.pow(2).mean(-1, keepdim=True)
+        hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
+        return self.weight * hidden_states.to(input_dtype)
+
+
+def rotate_half(x: torch.Tensor) -> torch.Tensor:
+    """Rotates half the hidden dims of the input."""
+    x1 = x[..., : x.shape[-1] // 2]
+    x2 = x[..., x.shape[-1] // 2 :]
+    return torch.cat((-x2, x1), dim=-1)
+
+
+def apply_rotary_pos_emb(q, k, cos, sin, unsqueeze_dim=1):
+    """Apply standard rotary position embedding (not M-RoPE)."""
+    cos = cos.unsqueeze(unsqueeze_dim)
+    sin = sin.unsqueeze(unsqueeze_dim)
+    q_embed = (q * cos) + (rotate_half(q) * sin)
+    k_embed = (k * cos) + (rotate_half(k) * sin)
+    return q_embed, k_embed
+
+
+class DiffusionVL_Qwen2_5_RotaryEmbedding(nn.Module):
+    """Standard rotary position embedding for Qwen2.5."""
+
+    def __init__(self, config: DiffusionVL_Qwen2_5_Config):
+        super().__init__()
+        self.config = config
+        dim = config.hidden_size // config.num_attention_heads
+        inv_freq = 1.0 / (config.rope_theta ** (torch.arange(0, dim, 2, dtype=torch.float32) / dim))
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+
+    @torch.no_grad()
+    def forward(self, x, position_ids):
+        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1)
+        position_ids_expanded = position_ids[:, None, :].float()
+
+        device_type = x.device.type
+        with torch.autocast(device_type=device_type, enabled=False):
+            freqs = (inv_freq_expanded @ position_ids_expanded).transpose(1, 2)
+            emb = torch.cat((freqs, freqs), dim=-1)
+            cos = emb.cos()
+            sin = emb.sin()
+
+        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
+
+
+def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
+    """Repeat key/value heads for GQA."""
+    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 DiffusionVL_Qwen2_5_MLP(nn.Module):
+    def __init__(self, config: DiffusionVL_Qwen2_5_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_states):
+        return self.down_proj(self.act_fn(self.gate_proj(hidden_states)) * self.up_proj(hidden_states))
+
+
+class DiffusionVL_Qwen2_5_Attention(nn.Module):
+    """Attention with BD3LM store_kv support."""
+
+    def __init__(self, config: DiffusionVL_Qwen2_5_Config, layer_idx: int):
+        super().__init__()
+        self.config = config
+        self.layer_idx = layer_idx
+
+        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.scaling = self.head_dim ** -0.5
+        self.attention_dropout = config.attention_dropout
+        self.is_causal = False  # BD3LM uses block causal mask
+
+        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)
+
+        # Sliding window
+        self.sliding_window = config.sliding_window if (
+            config.use_sliding_window and layer_idx < config.max_window_layers
+        ) else None
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
+        store_kv: bool = True,
+        **kwargs,
+    ) -> Tuple[torch.Tensor, Optional[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)
+
+        # Apply rotary embeddings
+        cos, sin = position_embeddings
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
+
+        # Handle KV cache
+        if past_key_values is not None:
+            if store_kv:
+                cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
+                key_states, value_states = past_key_values.update(
+                    key_states, value_states, self.layer_idx, cache_kwargs
+                )
+            else:
+                # Read-only: concatenate with cached KV
+                if self.layer_idx < len(past_key_values):
+                    past_key_states, past_value_states = past_key_values[self.layer_idx]
+                    key_states = torch.cat([past_key_states, key_states], dim=2)
+                    value_states = torch.cat([past_value_states, value_states], dim=2)
+
+        # GQA: repeat KV heads
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+        # Compute attention with SDPA
+        if attention_mask is not None:
+            attn_output = F.scaled_dot_product_attention(
+                query_states,
+                key_states,
+                value_states,
+                attn_mask=attention_mask,
+                dropout_p=0.0,
+                is_causal=False,
+                scale=self.scaling,
+            )
+        else:
+            attn_output = F.scaled_dot_product_attention(
+                query_states,
+                key_states,
+                value_states,
+                dropout_p=0.0,
+                is_causal=False,
+                scale=self.scaling,
+            )
+
+        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)
+
+        return attn_output, None
+
+
+class DiffusionVL_Qwen2_5_DecoderLayer(nn.Module):
+    def __init__(self, config: DiffusionVL_Qwen2_5_Config, layer_idx: int):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+        self.self_attn = DiffusionVL_Qwen2_5_Attention(config, layer_idx)
+        self.mlp = DiffusionVL_Qwen2_5_MLP(config)
+        self.input_layernorm = DiffusionVL_Qwen2_5_RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.post_attention_layernorm = DiffusionVL_Qwen2_5_RMSNorm(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_values: Optional[Cache] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
+        store_kv: bool = True,
+        **kwargs,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor]]:
+        residual = hidden_states
+        hidden_states = self.input_layernorm(hidden_states)
+
+        hidden_states, attn_weights = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            cache_position=cache_position,
+            position_embeddings=position_embeddings,
+            store_kv=store_kv,
+            **kwargs,
+        )
+        hidden_states = residual + hidden_states
+
+        residual = hidden_states
+        hidden_states = self.post_attention_layernorm(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        hidden_states = residual + hidden_states
+
+        return hidden_states, attn_weights
+
+
+# ============================================================================
+# Main Model Classes
+# ============================================================================
+
+class DiffusionVL_Qwen2_5_PreTrainedModel(PreTrainedModel):
+    config_class = DiffusionVL_Qwen2_5_Config
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["DiffusionVL_Qwen2_5_DecoderLayer", "SigLipEncoderLayer"]
+
+    def _init_weights(self, module: nn.Module) -> None:
+        std = self.config.initializer_range
+        if isinstance(module, nn.Linear):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.Embedding):
+            module.weight.data.normal_(mean=0.0, std=std)
+
+
+class DiffusionVL_Qwen2_5_Model(DiffusionVL_Qwen2_5_PreTrainedModel):
+    """Base model with vision tower, projector, and LLM layers."""
+
+    def __init__(self, config: DiffusionVL_Qwen2_5_Config):
+        super().__init__(config)
+        self.config = config
+
+        # Vision components
+        self.vision_tower = DiffusionVL_Qwen2_5_VisionTower(config.vision_config)
+        self.mm_projector = build_mm_projector(config)
+        self.image_newline = nn.Parameter(torch.zeros(config.hidden_size))
+
+        # LLM components
+        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size)
+        self.layers = nn.ModuleList([
+            DiffusionVL_Qwen2_5_DecoderLayer(config, layer_idx)
+            for layer_idx in range(config.num_hidden_layers)
+        ])
+        self.norm = DiffusionVL_Qwen2_5_RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.rotary_emb = DiffusionVL_Qwen2_5_RotaryEmbedding(config)
+
+        # BD3LM components
+        self.block_size = config.bd3lm_block_size
+        self.mask_token_id = config.mask_token_id
+
+        self.gradient_checkpointing = False
+        self.post_init()
+
+    @property
+    def num_patches_per_side(self):
+        """Number of patches per side for the vision encoder."""
+        image_size = getattr(self.config.vision_config, "image_size", 384)
+        patch_size = getattr(self.config.vision_config, "patch_size", 14)
+        return image_size // patch_size
+
+    def get_image_features(self, pixel_values: torch.Tensor) -> torch.Tensor:
+        """Get image features through vision tower and projector."""
+        vision_output = self.vision_tower(pixel_values)
+        image_features = self.mm_projector(vision_output)
+        return image_features
+
+    def forward(
+        self,
+        input_ids: Optional[torch.LongTensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        store_kv: bool = True,
+        **kwargs,
+    ):
+        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 True
+
+        if inputs_embeds is None:
+            inputs_embeds = self.embed_tokens(input_ids)
+
+        past_key_values_length = past_key_values.get_seq_length() if past_key_values is not None else 0
+
+        if cache_position is None:
+            cache_position = torch.arange(
+                past_key_values_length, past_key_values_length + inputs_embeds.shape[1],
+                device=inputs_embeds.device
+            )
+
+        if position_ids is None:
+            position_ids = cache_position.unsqueeze(0)
+
+        hidden_states = inputs_embeds
+        position_embeddings = self.rotary_emb(hidden_states, position_ids)
+
+        for decoder_layer in self.layers:
+            hidden_states, _ = decoder_layer(
+                hidden_states,
+                attention_mask=attention_mask,
+                position_ids=position_ids,
+                past_key_values=past_key_values,
+                cache_position=cache_position,
+                position_embeddings=position_embeddings,
+                store_kv=store_kv,
+            )
+
+        hidden_states = self.norm(hidden_states)
+
+        if not return_dict:
+            return (hidden_states, past_key_values if use_cache else None)
+
+        return BaseModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=past_key_values if use_cache else None,
+        )
+
+
+class DiffusionVL_Qwen2_5_ForConditionalGeneration(DiffusionVL_Qwen2_5_PreTrainedModel):
+    """DiffusionVL-Qwen2.5 with LM head for diffusion-based generation."""
+
+    _tied_weights_keys = ["lm_head.weight"]
+
+    def __init__(self, config: DiffusionVL_Qwen2_5_Config):
+        super().__init__(config)
+        self.model = DiffusionVL_Qwen2_5_Model(config)
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+
+        # BD3LM attributes
+        self.block_size = config.bd3lm_block_size
+        self.mask_token_id = config.mask_token_id
+
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.model.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.model.embed_tokens = value
+
+    def get_output_embeddings(self):
+        return self.lm_head
+
+    def set_output_embeddings(self, new_embeddings):
+        self.lm_head = new_embeddings
+
+    @torch.no_grad()
+    def generate(
+        self,
+        inputs: Optional[torch.Tensor] = None,
+        images: Optional[torch.Tensor] = None,
+        image_sizes: Optional[List[Tuple[int, int]]] = None,
+        num_patches_per_image: Optional[List[int]] = None,
+        gen_length: int = 256,
+        steps: int = 8,
+        temperature: float = 0.0,
+        **kwargs,
+    ):
+        """Diffusion-based generation."""
+        if images is not None:
+            inputs_embeds = self.prepare_inputs_labels_for_multimodal(
+                input_ids=inputs,
+                images=images,
+                image_sizes=image_sizes,
+                num_patches_per_image=num_patches_per_image,
+            )
+        else:
+            inputs_embeds = self.get_input_embeddings()(inputs)
+
+        return self.generate_with_bd3lm(
+            inputs_embeds=inputs_embeds,
+            gen_length=gen_length,
+            steps=steps,
+            temperature=temperature,
+            **kwargs,
+        )
+
+    def prepare_inputs_labels_for_multimodal(
+        self,
+        input_ids: torch.Tensor,
+        images: torch.Tensor,
+        image_sizes: Optional[List[Tuple[int, int]]] = None,
+        num_patches_per_image: Optional[List[int]] = None,
+    ) -> torch.Tensor:
+        """
+        Prepare inputs by merging text embeddings with image features.
+
+        Implements spatial_unpad merge type matching training code:
+        - For single-patch images: just add image_newline
+        - For multi-patch images (anyres): unpad, interpolate, add newline per row
+
+        Args:
+            input_ids: Token IDs with IMAGE_TOKEN_INDEX placeholders
+            images: Tensor of shape (total_patches, C, H, W) containing all patches
+            image_sizes: List of (width, height) tuples for each original image
+            num_patches_per_image: List of patch counts per image (from processor)
+        """
+        import re as regex_module
+
+        device = input_ids.device
+        batch_size = input_ids.shape[0]
+
+        # Get raw image features from vision tower + projector
+        # images shape: (total_patches, C, H, W)
+        # raw_image_features shape: (total_patches, num_tokens, hidden_size)
+        raw_image_features = self.model.get_image_features(images)
+
+        # Determine split sizes for per-image features
+        if num_patches_per_image is not None:
+            split_sizes = num_patches_per_image
+        else:
+            # Fallback: assume 1 patch per image
+            split_sizes = [1] * images.shape[0]
+
+        # Split features per image
+        image_features_list = list(torch.split(raw_image_features, split_sizes, dim=0))
+
+        # Process image features with spatial_unpad logic
+        mm_patch_merge_type = getattr(self.config, "mm_patch_merge_type", "spatial_unpad")
+        image_aspect_ratio = getattr(self.config, "image_aspect_ratio", "anyres_max_4")
+
+        processed_image_features = []
+        for image_idx, image_feature in enumerate(image_features_list):
+            # image_feature shape: (num_patches, num_tokens, hidden_size)
+
+            if image_feature.shape[0] > 1:  # Multi-patch (anyres) image
+                base_image_feature = image_feature[0]  # Base patch: (num_tokens, hidden_size)
+                image_feature = image_feature[1:]  # Additional patches: (num_grid_patches, num_tokens, hidden_size)
+
+                # Get patch grid dimensions
+                height = width = self.model.num_patches_per_side  # e.g., 27 for SigLIP 384
+
+                # Get max num patches for interpolation
+                max_num_patches = 4
+                if "anyres_max" in image_aspect_ratio:
+                    matched = regex_module.match(r"anyres_max_(\d+)", image_aspect_ratio)
+                    if matched:
+                        max_num_patches = int(matched.group(1))
+
+                # Determine grid shape - matching training code logic exactly
+                num_grid_patches = image_feature.shape[0]  # Actual grid patch count
+
+                if image_sizes is not None and image_idx < len(image_sizes):
+                    image_size = image_sizes[image_idx]
+                    grid_pinpoints = getattr(self.config, "image_grid_pinpoints", "(1x1),...,(2x2)")
+                    vision_tower_image_size = 384  # SigLIP patch size
+
+                    try:
+                        num_patch_width, num_patch_height = get_anyres_image_grid_shape(
+                            image_size, grid_pinpoints, vision_tower_image_size
+                        )
+                        # Verify calculated shape matches actual patch count
+                        expected_patches = num_patch_width * num_patch_height
+                        if expected_patches != num_grid_patches:
+                            logger.warning(
+                                f"Grid shape mismatch! image_size={image_size}, "
+                                f"expected {num_patch_width}x{num_patch_height}={expected_patches} patches, "
+                                f"but got {num_grid_patches} patches. Using actual count."
+                            )
+                            # Infer grid shape from actual patch count
+                            # Try to find factors that match the image aspect ratio
+                            img_w, img_h = image_size
+                            for h in range(1, num_grid_patches + 1):
+                                if num_grid_patches % h == 0:
+                                    w = num_grid_patches // h
+                                    # Check if this matches aspect ratio direction
+                                    if (img_w >= img_h and w >= h) or (img_w < img_h and w < h):
+                                        num_patch_height, num_patch_width = h, w
+                                        break
+                            else:
+                                num_patch_height = num_patch_width = int(math.sqrt(num_grid_patches))
+                    except Exception as e:
+                        # Fallback to 2x2, matching training code
+                        logger.warning(f"get_anyres_image_grid_shape error: {e}, fallback to 2x2")
+                        num_patch_width, num_patch_height = 2, 2
+                else:
+                    # Fallback to 2x2, matching training code
+                    num_patch_width, num_patch_height = 2, 2
+
+                # Reshape: (num_grid_patches, num_tokens, hidden) -> (patch_h, patch_w, h, w, hidden)
+                image_feature = image_feature.view(num_patch_height, num_patch_width, height, width, -1)
+
+                if "unpad" in mm_patch_merge_type:
+                    # Permute to (hidden, patch_h, h, patch_w, w) then flatten
+                    image_feature = image_feature.permute(4, 0, 2, 1, 3).contiguous()
+                    image_feature = image_feature.flatten(1, 2).flatten(2, 3)  # (hidden, H, W)
+
+                    # Unpad if image_sizes available
+                    if image_sizes is not None and image_idx < len(image_sizes):
+                        image_feature = unpad_image(image_feature, image_sizes[image_idx])
+
+                    c, h, w = image_feature.shape
+
+                    # Interpolate if too large
+                    if "anyres_max" in image_aspect_ratio:
+                        unit = height  # num_patches_per_side
+                        times = math.sqrt(h * w / (max_num_patches * unit**2))
+                        if times > 1.1:
+                            image_feature = image_feature[None]
+                            image_feature = F.interpolate(
+                                image_feature,
+                                [int(h // times), int(w // times)],
+                                mode="bilinear"
+                            )[0]
+
+                    # Add image_newline per row
+                    # image_feature: (hidden, h, w) -> add newline: (hidden, h, w+1)
+                    image_feature = torch.cat([
+                        image_feature,
+                        self.model.image_newline[:, None, None].expand(image_feature.shape[0], image_feature.shape[1], 1).to(image_feature.device)
+                    ], dim=-1)
+                    # Flatten and transpose: (hidden, h, w+1) -> (h*(w+1), hidden)
+                    image_feature = image_feature.flatten(1, 2).transpose(0, 1)
+                else:
+                    # Flat merge without unpad
+                    image_feature = image_feature.permute(0, 2, 1, 3, 4).contiguous()
+                    image_feature = image_feature.flatten(0, 3)
+
+                # Concatenate base image feature
+                image_feature = torch.cat([base_image_feature, image_feature], dim=0)
+                processed_image_features.append(image_feature)
+
+            else:  # Single-patch image
+                image_feature = image_feature[0]  # Remove batch dim: (num_tokens, hidden_size)
+                if "unpad" in mm_patch_merge_type:
+                    image_feature = torch.cat([
+                        image_feature,
+                        self.model.image_newline[None].to(image_feature.device)
+                    ], dim=0)
+                processed_image_features.append(image_feature)
+
+        # Build embeddings with image tokens replaced
+        new_input_embeds_list = []
+
+        for batch_idx in range(batch_size):
+            cur_input_ids = input_ids[batch_idx]
+            num_images = (cur_input_ids == IMAGE_TOKEN_INDEX).sum().item()
+
+            if num_images == 0:
+                cur_input_embeds = self.get_input_embeddings()(cur_input_ids)
+                new_input_embeds_list.append(cur_input_embeds)
+                continue
+
+            # LLaVA format: IMAGE_TOKEN_INDEX (-200) as placeholder
+            image_token_indices = [-1] + torch.where(cur_input_ids == IMAGE_TOKEN_INDEX)[0].tolist() + [len(cur_input_ids)]
+
+            cur_input_ids_noim = []
+            for idx in range(len(image_token_indices) - 1):
+                start = image_token_indices[idx] + 1
+                end = image_token_indices[idx + 1]
+                if start < end:
+                    cur_input_ids_noim.append(cur_input_ids[start:end])
+
+            if cur_input_ids_noim:
+                cur_input_embeds_noim = self.get_input_embeddings()(torch.cat(cur_input_ids_noim).to(device))
+                split_sizes_text = [x.shape[0] for x in cur_input_ids_noim]
+                cur_input_embeds_noim_split = list(torch.split(cur_input_embeds_noim, split_sizes_text))
+            else:
+                cur_input_embeds_noim_split = []
+
+            cur_new_input_embeds = []
+            cur_image_idx = 0
+
+            for idx in range(num_images + 1):
+                if idx < len(cur_input_embeds_noim_split):
+                    cur_new_input_embeds.append(cur_input_embeds_noim_split[idx].to(device))
+                if idx < num_images and cur_image_idx < len(processed_image_features):
+                    cur_new_input_embeds.append(processed_image_features[cur_image_idx].to(device))
+                    cur_image_idx += 1
+
+            if cur_new_input_embeds:
+                cur_new_input_embeds = torch.cat(cur_new_input_embeds, dim=0)
+            else:
+                cur_new_input_embeds = self.get_input_embeddings()(cur_input_ids.to(device))
+
+            new_input_embeds_list.append(cur_new_input_embeds)
+
+        # Pad to same length
+        max_len = max(x.shape[0] for x in new_input_embeds_list)
+        hidden_size = new_input_embeds_list[0].shape[-1]
+        dtype = new_input_embeds_list[0].dtype
+
+        inputs_embeds = torch.zeros(batch_size, max_len, hidden_size, dtype=dtype, device=device)
+        for i, embed in enumerate(new_input_embeds_list):
+            inputs_embeds[i, :embed.shape[0]] = embed
+
+        return inputs_embeds
+
+    @torch.no_grad()
+    def generate_with_bd3lm(
+        self,
+        inputs_embeds: torch.FloatTensor,
+        gen_length: int = 256,
+        steps: int = 8,
+        temperature: float = 0.0,
+        top_k: int = 0,
+        top_p: float = 1.0,
+        remasking_strategy: str = 'low_confidence_static',
+        use_kv_cache: bool = True,
+        confidence_threshold: float = 0.85,
+        **kwargs,
+    ):
+        """BD3LM generation with KV-cache support."""
+        device = inputs_embeds.device
+        batch_size = inputs_embeds.shape[0]
+        prompt_len = inputs_embeds.shape[1]
+        block_size = self.block_size
+        mask_id = self.mask_token_id
+
+        num_blocks = (prompt_len + gen_length + block_size - 1) // block_size
+        total_length = num_blocks * block_size
+
+        # Initialize with mask tokens
+        x_ids = torch.full((batch_size, total_length), mask_id, dtype=torch.long, device=device)
+        mask_embed = self.get_input_embeddings()(torch.tensor([mask_id], device=device))
+        x_embeds = mask_embed.repeat(batch_size, total_length, 1)
+        x_embeds[:, :prompt_len] = inputs_embeds.clone()
+
+        # Reconstruct prompt IDs
+        prompt_logits = self.lm_head(inputs_embeds)
+        prompt_ids = torch.argmax(prompt_logits, dim=-1)
+        x_ids[:, :prompt_len] = prompt_ids
+
+        # Block causal mask
+        dtype = inputs_embeds.dtype
+        block_mask = torch.tril(torch.ones(num_blocks, num_blocks, device=device, dtype=dtype))
+        block_diffusion_mask = block_mask.repeat_interleave(block_size, dim=0).repeat_interleave(block_size, dim=1)
+        block_diffusion_mask = block_diffusion_mask.unsqueeze(0).unsqueeze(1)
+        block_diffusion_mask = torch.where(
+            block_diffusion_mask == 0.,
+            torch.tensor(float('-inf'), device=device, dtype=dtype),
+            torch.tensor(0., device=device, dtype=dtype)
+        )
+
+        position_ids = torch.arange(total_length, device=device).unsqueeze(0).expand(batch_size, -1)
+
+        # KV-cache prefill
+        prefill_blocks = prompt_len // block_size
+        prefill_length = prefill_blocks * block_size
+
+        past_key_values = DynamicCache() if use_kv_cache else None
+
+        if use_kv_cache and prefill_length > 0:
+            prefill_embeds = x_embeds[:, :prefill_length]
+            prefill_mask = block_diffusion_mask[:, :, :prefill_length, :prefill_length]
+            prefill_pos_ids = position_ids[:, :prefill_length]
+
+            prefill_outputs = self.model(
+                inputs_embeds=prefill_embeds,
+                attention_mask=prefill_mask,
+                position_ids=prefill_pos_ids,
+                past_key_values=past_key_values,
+                use_cache=True,
+                store_kv=True,
+            )
+            past_key_values = prefill_outputs.past_key_values
+
+        num_transfer_tokens = self._get_num_transfer_tokens(block_size, steps)
+
+        # Generate block by block
+        for block_idx in range(prefill_blocks, num_blocks):
+            block_start = block_idx * block_size
+            block_end = block_start + block_size
+
+            cur_block_embeds = x_embeds[:, block_start:block_end].clone()
+            cur_block_ids = x_ids[:, block_start:block_end]
+            cur_mask = block_diffusion_mask[:, :, block_start:block_end, :block_end]
+            cur_pos_ids = position_ids[:, block_start:block_end]
+
+            for step in range(steps + 1):
+                is_mask = torch.all(torch.abs(cur_block_embeds - mask_embed) < 1e-5, dim=-1)
+                if not is_mask.any():
+                    if use_kv_cache:
+                        _ = self.model(
+                            inputs_embeds=cur_block_embeds,
+                            attention_mask=cur_mask,
+                            position_ids=cur_pos_ids,
+                            past_key_values=past_key_values,
+                            use_cache=True,
+                            store_kv=True,
+                        )
+                    break
+
+                if use_kv_cache:
+                    outputs = self.model(
+                        inputs_embeds=cur_block_embeds,
+                        attention_mask=cur_mask,
+                        position_ids=cur_pos_ids,
+                        past_key_values=past_key_values,
+                        use_cache=True,
+                        store_kv=False,
+                    )
+                    logits = self.lm_head(outputs.last_hidden_state).float()
+                else:
+                    context_embeds = x_embeds[:, :block_end].clone()
+                    context_embeds[:, block_start:block_end] = cur_block_embeds
+                    context_mask = block_diffusion_mask[:, :, :block_end, :block_end]
+                    context_pos_ids = position_ids[:, :block_end]
+
+                    outputs = self.model(
+                        inputs_embeds=context_embeds,
+                        attention_mask=context_mask,
+                        position_ids=context_pos_ids,
+                        past_key_values=None,
+                        use_cache=False,
+                        store_kv=False,
+                    )
+                    logits = self.lm_head(outputs.last_hidden_state[:, block_start:block_end]).float()
+
+                x0, x0_p = self._sample_with_temperature(logits, temperature, top_k, top_p)
+
+                # Ensure tensors are on the same device (for device_map="auto")
+                output_device = x0.device
+                is_mask_on_device = is_mask.to(output_device)
+
+                num_to_transfer = num_transfer_tokens[step].item()
+                transfer_mask = self._get_transfer_mask(
+                    is_mask_on_device, x0_p, num_to_transfer, remasking_strategy, confidence_threshold, output_device
+                )
+
+                cur_block_ids = torch.where(transfer_mask, x0, cur_block_ids)
+                x0_embeds = self.get_input_embeddings()(x0)
+                cur_block_embeds = torch.where(transfer_mask.unsqueeze(-1), x0_embeds, cur_block_embeds)
+
+            x_embeds[:, block_start:block_end] = cur_block_embeds
+            x_ids[:, block_start:block_end] = cur_block_ids
+
+            # EOS check: stop generation if EOS token is generated
+            if block_end > prompt_len:
+                gen_start_in_block = max(prompt_len, block_start)
+                gen_ids_check = x_ids[:, gen_start_in_block:block_end]
+                eos_token_id = self.config.eos_token_id if hasattr(self.config, 'eos_token_id') else 151645
+                if eos_token_id in gen_ids_check:
+                    break
+
+        return x_ids[:, prompt_len:prompt_len + gen_length]
+
+    def _get_num_transfer_tokens(self, block_length: int, steps: int) -> torch.Tensor:
+        if steps == 0:
+            return torch.zeros(0, dtype=torch.int64)
+        base = block_length // steps
+        remainder = block_length % steps
+        num_transfer_tokens = torch.zeros(steps + 1, dtype=torch.int64) + base
+        num_transfer_tokens[:remainder] += 1
+        return num_transfer_tokens
+
+    def _sample_with_temperature(self, logits, temperature, top_k, top_p):
+        vocab_size = logits.shape[-1]
+        logits_2d = logits.reshape(-1, vocab_size)
+        probs_original = F.softmax(logits_2d, dim=-1)
+
+        if temperature == 0:
+            token = torch.argmax(logits_2d, dim=-1, keepdim=True)
+        else:
+            logits_modified = logits_2d / temperature
+            if top_k > 0:
+                logits_modified = self._top_k_logits(logits_modified, top_k)
+            if top_p < 1.0:
+                logits_modified = self._top_p_logits(logits_modified, top_p)
+            probs_modified = F.softmax(logits_modified, dim=-1)
+            token = torch.multinomial(probs_modified, num_samples=1)
+
+        token_prob = torch.gather(probs_original, -1, token)
+        orig_shape = logits.shape[:-1]
+        return token.view(*orig_shape), token_prob.view(*orig_shape)
+
+    @staticmethod
+    def _top_k_logits(logits, k):
+        if k <= 0:
+            return logits
+        values, _ = torch.topk(logits, k)
+        min_values = values[..., -1, None]
+        return torch.where(logits < min_values, float('-inf'), logits)
+
+    @staticmethod
+    def _top_p_logits(logits, p):
+        sorted_logits, sorted_indices = torch.sort(logits, descending=True)
+        cumulative_probs = torch.cumsum(F.softmax(sorted_logits, dim=-1), dim=-1)
+        sorted_mask = cumulative_probs > p
+        sorted_mask[..., 1:] = sorted_mask[..., :-1].clone()
+        sorted_mask[..., 0] = False
+        mask_indices = torch.scatter(
+            torch.full_like(logits, False, dtype=torch.bool),
+            -1, sorted_indices, sorted_mask
+        )
+        return logits.masked_fill(mask_indices, float('-inf'))
+
+    def _get_transfer_mask(self, is_mask, x0_p, num_to_transfer, strategy, threshold, device):
+        transfer_mask = torch.zeros_like(is_mask, dtype=torch.bool)
+
+        if strategy == 'sequential':
+            for j in range(is_mask.shape[0]):
+                if is_mask[j].any():
+                    mask_positions = is_mask[j].nonzero(as_tuple=True)[0]
+                    num_to_select = min(num_to_transfer, len(mask_positions))
+                    selected_positions = mask_positions[:num_to_select]
+                    transfer_mask[j, selected_positions] = True
+
+        elif strategy == 'low_confidence_static':
+            confidence = torch.where(is_mask, x0_p, float('-inf'))
+            for j in range(confidence.shape[0]):
+                num_masks = is_mask[j].sum().item()
+                k = min(num_to_transfer, num_masks)
+                if k > 0 and not torch.all(torch.isinf(confidence[j])):
+                    _, idx = torch.topk(confidence[j], k)
+                    transfer_mask[j, idx] = True
+
+        elif strategy == 'low_confidence_dynamic':
+            confidence = torch.where(is_mask, x0_p, float('-inf'))
+            for j in range(confidence.shape[0]):
+                high_conf_mask = confidence[j] > threshold
+                num_high = high_conf_mask.sum().item()
+                if num_high >= num_to_transfer:
+                    transfer_mask[j] = high_conf_mask
+                else:
+                    num_masks = is_mask[j].sum().item()
+                    k = min(num_to_transfer, num_masks)
+                    if k > 0:
+                        _, idx = torch.topk(confidence[j], k)
+                        transfer_mask[j, idx] = True
+
+        return transfer_mask
+
+
+__all__ = [
+    "DiffusionVL_Qwen2_5_Config",
+    "DiffusionVL_Qwen2_5_VisionConfig",
+    "DiffusionVL_Qwen2_5_Model",
+    "DiffusionVL_Qwen2_5_ForConditionalGeneration",
+]

preprocessor_config.json

@@ -0,0 +1,21 @@
+{
+  "image_mean": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "image_std": [
+    0.5,
+    0.5,
+    0.5
+  ],
+  "size": [
+    384,
+    384
+  ],
+  "rescale_factor": 0.00392156862745098,
+  "processor_class": "DiffusionVL_Qwen2_5_Processor",
+  "auto_map": {
+    "AutoProcessor": "processing_diffusionvl_qwen2_5.DiffusionVL_Qwen2_5_Processor"
+  }
+}

processing_diffusionvl_qwen2_5.py

@@ -0,0 +1,485 @@
+# coding=utf-8
+# Copyright 2025 The HustVL 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.
+"""
+DiffusionVL-Qwen2.5 Processor - Self-contained image processing matching training code.
+
+This processor implements the same image processing pipeline as the training code:
+- process_images with anyres support
+- tokenizer_image_token for proper <image> token handling
+- Uses SiglipImageProcessor for the underlying image preprocessing
+"""
+
+import ast
+import math
+import re
+from typing import List, Optional, Tuple, Union
+
+import torch
+import numpy as np
+from PIL import Image
+
+from transformers.feature_extraction_utils import BatchFeature
+from transformers.processing_utils import ProcessorMixin
+from transformers.tokenization_utils_base import PreTokenizedInput, TextInput
+from transformers import SiglipImageProcessor
+
+
+# Image token for LLaVA format
+DEFAULT_IMAGE_TOKEN = "<image>"
+IMAGE_TOKEN_INDEX = -200
+
+
+# ============================================================================
+# Image Processing Utilities (matching training code mm_utils.py)
+# ============================================================================
+
+def select_best_resolution(original_size: Tuple[int, int], possible_resolutions: List[Tuple[int, int]]) -> Tuple[int, int]:
+    """
+    Selects the best resolution from a list of possible resolutions based on the original size.
+    Matching training code: llava/mm_utils.py::select_best_resolution
+    """
+    original_width, original_height = original_size
+    best_fit = None
+    max_effective_resolution = 0
+    min_wasted_resolution = float("inf")
+
+    for width, height in possible_resolutions:
+        scale = min(width / original_width, height / original_height)
+        downscaled_width, downscaled_height = int(original_width * scale), int(original_height * scale)
+        effective_resolution = min(downscaled_width * downscaled_height, original_width * original_height)
+        wasted_resolution = (width * height) - effective_resolution
+
+        if effective_resolution > max_effective_resolution or (effective_resolution == max_effective_resolution and wasted_resolution < min_wasted_resolution):
+            max_effective_resolution = effective_resolution
+            min_wasted_resolution = wasted_resolution
+            best_fit = (width, height)
+
+    return best_fit
+
+
+def resize_and_pad_image(image: Image.Image, target_resolution: Tuple[int, int]) -> Image.Image:
+    """
+    Resize and pad an image to a target resolution while maintaining aspect ratio.
+    Matching training code: llava/mm_utils.py::resize_and_pad_image
+    """
+    original_width, original_height = image.size
+    target_width, target_height = target_resolution
+
+    scale_w = target_width / original_width
+    scale_h = target_height / original_height
+
+    if scale_w < scale_h:
+        new_width = target_width
+        new_height = min(math.ceil(original_height * scale_w), target_height)
+    else:
+        new_height = target_height
+        new_width = min(math.ceil(original_width * scale_h), target_width)
+
+    resized_image = image.resize((new_width, new_height))
+    new_image = Image.new("RGB", (target_width, target_height), (0, 0, 0))
+    paste_x = (target_width - new_width) // 2
+    paste_y = (target_height - new_height) // 2
+    new_image.paste(resized_image, (paste_x, paste_y))
+
+    return new_image
+
+
+def divide_to_patches(image: Image.Image, patch_size: int) -> List[Image.Image]:
+    """
+    Divides an image into patches of a specified size.
+    Matching training code: llava/mm_utils.py::divide_to_patches
+    """
+    patches = []
+    width, height = image.size
+    for i in range(0, height, patch_size):
+        for j in range(0, width, patch_size):
+            box = (j, i, j + patch_size, i + patch_size)
+            patch = image.crop(box)
+            patches.append(patch)
+    return patches
+
+
+def expand2square(pil_img: Image.Image, background_color: Tuple[int, int, int]) -> Image.Image:
+    """
+    Expand image to square by padding.
+    Matching training code: llava/mm_utils.py::expand2square
+    """
+    width, height = pil_img.size
+    if width == height:
+        return pil_img
+    elif width > height:
+        result = Image.new(pil_img.mode, (width, width), background_color)
+        result.paste(pil_img, (0, (width - height) // 2))
+        return result
+    else:
+        result = Image.new(pil_img.mode, (height, height), background_color)
+        result.paste(pil_img, ((height - width) // 2, 0))
+        return result
+
+
+def get_anyres_image_grid_shape(image_size: Tuple[int, int], grid_pinpoints, patch_size: int) -> Tuple[int, int]:
+    """
+    Calculate the shape of the image patch grid after the preprocessing for images of any resolution.
+    Matching training code: llava/mm_utils.py::get_anyres_image_grid_shape
+    """
+    if isinstance(grid_pinpoints, str) and "x" in grid_pinpoints:
+        assert patch_size in [224, 336, 384, 448, 512], "patch_size should be in [224, 336, 384, 448, 512]"
+        matches = re.findall(r"\((\d+)x(\d+)\)", grid_pinpoints)
+        range_start = tuple(map(int, matches[0]))
+        range_end = tuple(map(int, matches[-1]))
+        grid_pinpoints = [(i, j) for i in range(range_start[0], range_end[0] + 1) for j in range(range_start[1], range_end[1] + 1)]
+        grid_pinpoints = [[dim * patch_size for dim in pair] for pair in grid_pinpoints]
+    if isinstance(grid_pinpoints, list):
+        possible_resolutions = grid_pinpoints
+    else:
+        possible_resolutions = ast.literal_eval(grid_pinpoints)
+    width, height = select_best_resolution(image_size, possible_resolutions)
+    return width // patch_size, height // patch_size
+
+
+def process_anyres_image(image: Image.Image, processor: SiglipImageProcessor, grid_pinpoints: str) -> torch.Tensor:
+    """
+    Process an image with variable resolutions (anyres).
+    Matching training code: llava/mm_utils.py::process_anyres_image
+
+    Returns: torch.Tensor of shape (num_patches, C, H, W) where num_patches = 1 + grid_patches
+    """
+    # Get patch size from processor
+    if isinstance(processor.size, dict):
+        patch_size = processor.size.get("shortest_edge", processor.size.get("height", 384))
+    else:
+        patch_size = processor.size[0] if hasattr(processor.size, '__getitem__') else 384
+
+    crop_size = processor.crop_size.get("height", patch_size) if hasattr(processor, 'crop_size') else patch_size
+
+    # Parse grid pinpoints
+    if isinstance(grid_pinpoints, str) and "x" in grid_pinpoints:
+        assert patch_size in [224, 336, 384, 448, 512], f"patch_size {patch_size} should be in [224, 336, 384, 448, 512]"
+        matches = re.findall(r"\((\d+)x(\d+)\)", grid_pinpoints)
+        range_start = tuple(map(int, matches[0]))
+        range_end = tuple(map(int, matches[-1]))
+        grid_pinpoints_list = [(i, j) for i in range(range_start[0], range_end[0] + 1) for j in range(range_start[1], range_end[1] + 1)]
+        possible_resolutions = [[dim * patch_size for dim in pair] for pair in grid_pinpoints_list]
+    elif isinstance(grid_pinpoints, list):
+        possible_resolutions = grid_pinpoints
+    else:
+        possible_resolutions = ast.literal_eval(grid_pinpoints)
+
+    best_resolution = select_best_resolution(image.size, possible_resolutions)
+    image_padded = resize_and_pad_image(image, best_resolution)
+    patches = divide_to_patches(image_padded, crop_size)
+
+    # Base image (resized to patch size) - matching training code behavior
+    if isinstance(processor.size, dict):
+        shortest_edge = processor.size.get("shortest_edge", processor.size.get("height", 384))
+    else:
+        shortest_edge = min(processor.size) if hasattr(processor.size, '__iter__') else 384
+    image_original_resize = image.resize((shortest_edge, shortest_edge))
+
+    # Combine: base image + grid patches (same order as training code)
+    image_patches = [image_original_resize] + patches
+
+    # Preprocess all patches using the HF processor
+    processed_patches = [processor.preprocess(patch, return_tensors="pt")["pixel_values"][0] for patch in image_patches]
+
+    return torch.stack(processed_patches, dim=0)
+
+
+def process_images(images: List[Image.Image], image_processor: SiglipImageProcessor, model_cfg) -> torch.Tensor:
+    """
+    Process images matching the training code pipeline.
+    Matching training code: llava/mm_utils.py::process_images
+
+    Args:
+        images: List of PIL Images
+        image_processor: SiglipImageProcessor instance
+        model_cfg: Model config with image_aspect_ratio and image_grid_pinpoints
+
+    Returns:
+        torch.Tensor or List[torch.Tensor] of processed image patches
+    """
+    image_aspect_ratio = getattr(model_cfg, "image_aspect_ratio", None)
+    new_images = []
+
+    if image_aspect_ratio == "anyres" or (image_aspect_ratio and "anyres" in image_aspect_ratio):
+        grid_pinpoints = getattr(model_cfg, "image_grid_pinpoints", "(1x1),...,(2x2)")
+        for image in images:
+            processed = process_anyres_image(image, image_processor, grid_pinpoints)
+            new_images.append(processed)
+    elif image_aspect_ratio == "pad":
+        for image in images:
+            image = expand2square(image, tuple(int(x * 255) for x in image_processor.image_mean))
+            processed = image_processor.preprocess(image, return_tensors="pt")["pixel_values"][0]
+            new_images.append(processed)
+    else:
+        # Default: simple preprocessing
+        return image_processor.preprocess(images, return_tensors="pt")["pixel_values"]
+
+    # Stack if all same shape, otherwise return list
+    if all(x.shape == new_images[0].shape for x in new_images):
+        new_images = torch.stack(new_images, dim=0)
+    return new_images
+
+
+def tokenizer_image_token(prompt: str, tokenizer, image_token_index: int = IMAGE_TOKEN_INDEX, return_tensors: str = None):
+    """
+    Tokenize prompt with proper handling of <image> tokens.
+    Matching training code: llava/mm_utils.py::tokenizer_image_token
+
+    Args:
+        prompt: Text prompt containing <image> placeholders
+        tokenizer: Tokenizer instance
+        image_token_index: Index to use for image tokens (default: -200)
+        return_tensors: If "pt", return PyTorch tensor
+
+    Returns:
+        List of token IDs or torch.Tensor
+    """
+    prompt_chunks = [tokenizer(chunk).input_ids for chunk in prompt.split("<image>")]
+
+    def insert_separator(X, sep):
+        return [ele for sublist in zip(X, [sep] * len(X)) for ele in sublist][:-1]
+
+    input_ids = []
+    offset = 0
+    if len(prompt_chunks) > 0 and len(prompt_chunks[0]) > 0 and prompt_chunks[0][0] == tokenizer.bos_token_id:
+        offset = 1
+        input_ids.append(prompt_chunks[0][0])
+
+    for x in insert_separator(prompt_chunks, [image_token_index] * (offset + 1)):
+        input_ids.extend(x[offset:])
+
+    if return_tensors is not None:
+        if return_tensors == "pt":
+            return torch.tensor(input_ids, dtype=torch.long)
+        raise ValueError(f"Unsupported tensor type: {return_tensors}")
+    return input_ids
+
+
+# ============================================================================
+# Conversation Templates (matching training code)
+# ============================================================================
+
+class Conversation:
+    """Simple conversation class matching LLaVA's conv_templates."""
+
+    def __init__(self, system: str, roles: Tuple[str, str], sep: str, sep2: str = None):
+        self.system = system
+        self.roles = roles
+        self.sep = sep
+        self.sep2 = sep2
+        self.messages = []
+
+    def copy(self):
+        return Conversation(
+            system=self.system,
+            roles=self.roles,
+            sep=self.sep,
+            sep2=self.sep2,
+        )
+
+    def append_message(self, role: str, message: str):
+        self.messages.append([role, message])
+
+    def get_prompt(self) -> str:
+        """Build the prompt string."""
+        ret = ""
+        if self.system:
+            ret = f"<|im_start|>system\n{self.system}<|im_end|>\n"
+
+        for role, message in self.messages:
+            if message:
+                ret += f"<|im_start|>{role}\n{message}<|im_end|>\n"
+            else:
+                ret += f"<|im_start|>{role}\n"
+        return ret
+
+
+# Pre-defined conversation template for Qwen2.5
+CONV_QWEN_2_5 = Conversation(
+    system="You are Qwen, created by Alibaba Cloud. You are a helpful assistant.",
+    roles=("user", "assistant"),
+    sep="<|im_end|>",
+    sep2=None,
+)
+
+
+# ============================================================================
+# Main Processor Class
+# ============================================================================
+
+class DiffusionVL_Qwen2_5_Processor(ProcessorMixin):
+    """
+    Processor for DiffusionVL-Qwen2.5 model.
+
+    Self-contained implementation matching the training code pipeline:
+    - Uses SiglipImageProcessor for image preprocessing
+    - Implements process_images with anyres support
+    - Implements tokenizer_image_token for proper <image> token handling
+
+    The processor stores model config for anyres parameters. Config can be:
+    1. Passed during __init__ via config parameter
+    2. Set after loading via set_config() method
+    3. Passed per-call via model_cfg parameter in __call__
+    """
+
+    attributes = ["tokenizer"]
+    tokenizer_class = ("Qwen2Tokenizer", "Qwen2TokenizerFast")
+
+    def __init__(
+        self,
+        tokenizer=None,
+        image_processor=None,
+        config=None,
+        **kwargs
+    ):
+        # Use provided image_processor or create default SiglipImageProcessor
+        if image_processor is None:
+            self.image_processor = SiglipImageProcessor.from_pretrained("google/siglip-so400m-patch14-384")
+        else:
+            self.image_processor = image_processor
+
+        # Store config for anyres processing
+        self._config = config
+
+        super().__init__(tokenizer)
+
+    def set_config(self, config):
+        """Set model config for anyres image processing."""
+        self._config = config
+
+    def __call__(
+        self,
+        text: Optional[Union[TextInput, PreTokenizedInput, List[TextInput], List[PreTokenizedInput]]] = None,
+        images: Optional[Union[Image.Image, List[Image.Image]]] = None,
+        model_cfg=None,
+        return_tensors: Optional[str] = "pt",
+        **kwargs,
+    ) -> BatchFeature:
+        """
+        Process text and images for model input.
+
+        Args:
+            text: Input text or list of texts with <image> placeholder.
+            images: PIL Image or list of PIL Images.
+            model_cfg: Model config (needed for anyres parameters).
+            return_tensors: Return type ("pt" for PyTorch).
+
+        Returns:
+            BatchFeature with input_ids and pixel_values.
+        """
+        if text is None and images is None:
+            raise ValueError("You must provide either text or images.")
+
+        # Process text using tokenizer_image_token
+        if text is not None:
+            if isinstance(text, str):
+                text = [text]
+
+            all_input_ids = []
+            for t in text:
+                input_ids = tokenizer_image_token(t, self.tokenizer, IMAGE_TOKEN_INDEX, return_tensors="pt")
+                all_input_ids.append(input_ids)
+
+            # Pad sequences if multiple
+            if len(all_input_ids) > 1:
+                max_len = max(ids.shape[0] for ids in all_input_ids)
+                padded_input_ids = []
+                for ids in all_input_ids:
+                    if ids.shape[0] < max_len:
+                        padding = torch.full((max_len - ids.shape[0],), self.tokenizer.pad_token_id, dtype=torch.long)
+                        ids = torch.cat([ids, padding])
+                    padded_input_ids.append(ids)
+                input_ids = torch.stack(padded_input_ids)
+            else:
+                input_ids = all_input_ids[0].unsqueeze(0)
+
+            text_inputs = {"input_ids": input_ids}
+        else:
+            text_inputs = {}
+
+        # Process images using process_images
+        if images is not None:
+            if isinstance(images, Image.Image):
+                images = [images]
+
+            # Get image sizes before processing
+            image_sizes = [img.size for img in images]
+
+            # Use model_cfg if provided, otherwise use stored config
+            cfg = model_cfg if model_cfg is not None else self._config
+
+            if cfg is not None:
+                pixel_values = process_images(images, self.image_processor, cfg)
+                # Calculate num_patches_per_image for anyres
+                if isinstance(pixel_values, list):
+                    num_patches_per_image = [t.shape[0] for t in pixel_values]
+                    # Concatenate all patches into single tensor
+                    pixel_values = torch.cat(pixel_values, dim=0)
+                elif pixel_values.dim() == 5:
+                    # Shape: (num_images, num_patches, C, H, W)
+                    num_patches_per_image = [pixel_values.shape[1]] * pixel_values.shape[0]
+                    pixel_values = pixel_values.view(-1, *pixel_values.shape[2:])
+                else:
+                    # Shape: (total_patches, C, H, W) - 1 patch per image
+                    num_patches_per_image = [1] * len(images)
+            else:
+                # Fallback to simple preprocessing if no config
+                pixel_values = self.image_processor.preprocess(images, return_tensors="pt")["pixel_values"]
+                num_patches_per_image = [1] * len(images)
+
+            image_inputs = {
+                "pixel_values": pixel_values,
+                "image_sizes": image_sizes,
+            }
+        else:
+            image_inputs = {}
+            num_patches_per_image = None
+
+        # Create BatchFeature first
+        result = BatchFeature(data={**text_inputs, **image_inputs}, tensor_type=return_tensors)
+
+        # Add num_patches_per_image as plain Python list (not converted to tensor)
+        # This is needed for prepare_inputs_labels_for_multimodal
+        if num_patches_per_image is not None:
+            result["num_patches_per_image"] = num_patches_per_image
+
+        return result
+
+    def batch_decode(self, *args, **kwargs):
+        """Decode token IDs to text."""
+        return self.tokenizer.batch_decode(*args, **kwargs)
+
+    def decode(self, *args, **kwargs):
+        """Decode token IDs to text."""
+        return self.tokenizer.decode(*args, **kwargs)
+
+    @property
+    def model_input_names(self):
+        tokenizer_input_names = self.tokenizer.model_input_names
+        image_processor_input_names = ["pixel_values", "image_sizes", "num_patches_per_image"]
+        return list(dict.fromkeys(tokenizer_input_names + image_processor_input_names))
+
+
+__all__ = [
+    "DiffusionVL_Qwen2_5_Processor",
+    "process_images",
+    "tokenizer_image_token",
+    "get_anyres_image_grid_shape",
+    "Conversation",
+    "CONV_QWEN_2_5",
+    "DEFAULT_IMAGE_TOKEN",
+    "IMAGE_TOKEN_INDEX",
+]

special_tokens_map.json

@@ -0,0 +1,31 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c5ae00e602b8860cbd784ba82a8aa14e8feecec692e7076590d014d7b7fdafa
+size 11421896

tokenizer_config.json

@@ -0,0 +1,208 @@
+{
+  "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
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": {},
+  "model_max_length": 8192,
+  "pad_token": "<|endoftext|>",
+  "padding_side": "right",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}