KyleShao commited on
Commit
a63d81a
·
verified ·
1 Parent(s): 0d29169

Upload folder using huggingface_hub

Browse files
README.md ADDED
@@ -0,0 +1,178 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Cosmos-Predict2.5-2B (Diffusers Format)
2
+
3
+ This is the NVIDIA Cosmos Predict 2.5 2B model in Diffusers-compatible format for use with FastVideo.
4
+
5
+ ## Model Components
6
+
7
+ This model consists of the following components:
8
+
9
+ ### 1. Transformer (DiT)
10
+ - **Class**: `Cosmos25Transformer3DModel`
11
+ - **Architecture**: 28 layers, 16 attention heads, 128 head dim
12
+ - **Parameters**: ~2B parameters
13
+ - **Input channels**: 16 (latent space)
14
+ - **Patch size**: (1, 2, 2) for temporal and spatial dimensions
15
+ - **Features**:
16
+ - AdaLN-LoRA conditioning (dim=256)
17
+ - RoPE positional embeddings with 3D scaling
18
+ - Cross-attention projection for text conditioning
19
+ - RMS normalization for Q/K
20
+
21
+ ### 2. VAE (Wan2.1)
22
+ - **Class**: `AutoencoderKLWan`
23
+ - **Latent channels**: 16
24
+ - **Compression**: 8x spatial, 4x temporal
25
+ - **Architecture**: 4-stage encoder/decoder with residual blocks
26
+ - **Features**:
27
+ - Feature caching for efficiency
28
+ - Configurable tiling support
29
+ - Clip output to [-1, 1]
30
+
31
+ ### 3. Scheduler
32
+ - **Class**: `FlowUniPCMultistepScheduler`
33
+ - **Type**: Multi-step flow matching solver (UniPC)
34
+ - **Order**: 2 (predictor-corrector)
35
+ - **Configuration**:
36
+ - Training timesteps: 1000
37
+ - Shift: 1
38
+ - No dynamic shifting
39
+ - Solver type: bh2 (recommended for >10 steps)
40
+
41
+ ### 4. Text Encoder & Tokenizer
42
+ - **Note**: Text encoder and tokenizer are not included in this directory
43
+ - **Official Implementation**: Uses Reason1 or official TextEncoder from `cosmos_predict2`
44
+ - **Expected format**: Text embeddings with shape (batch, 512, 100352)
45
+
46
+ ## Directory Structure
47
+
48
+ ```
49
+ models--nvidia--Cosmos-Predict2.5-2B-Diffusers/
50
+ ├── model_index.json # Pipeline component registry
51
+ ├── README.md # This file
52
+ ├── transformer/
53
+ │ ├── config.json # Transformer configuration
54
+ │ └── 81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt # Model weights
55
+ ├── vae/
56
+ │ ├── config.json # VAE configuration
57
+ │ └── tokenizer.pth # VAE weights
58
+ └── scheduler/
59
+ └── scheduler_config.json # Scheduler configuration
60
+ ```
61
+
62
+ ## Usage with FastVideo
63
+
64
+ ### Option 1: Using FastVideo Pipeline (Recommended)
65
+
66
+ ```python
67
+ from fastvideo import FastVideoArgs
68
+ from fastvideo.pipelines.basic.cosmos.cosmos2_5_pipeline import Cosmos2_5Pipeline
69
+
70
+ # Initialize pipeline
71
+ args = FastVideoArgs.from_cli_args(model="nvidia/Cosmos-Predict2.5-2B-Diffusers")
72
+ pipeline = Cosmos2_5Pipeline(args)
73
+
74
+ # Generate video
75
+ output = pipeline(
76
+ prompt="A robot welding in an industrial setting",
77
+ height=480,
78
+ width=832,
79
+ num_frames=121,
80
+ num_inference_steps=35,
81
+ guidance_scale=7.0,
82
+ )
83
+ ```
84
+
85
+ ### Option 2: Manual Component Loading
86
+
87
+ ```python
88
+ from fastvideo.models.dits.cosmos2_5 import Cosmos25Transformer3DModel
89
+ from fastvideo.models.vaes.wanvae import AutoencoderKLWan
90
+ from fastvideo.models.schedulers.scheduling_flow_unipc_multistep import FlowUniPCMultistepScheduler
91
+
92
+ # Load components
93
+ transformer = Cosmos25Transformer3DModel.from_pretrained(
94
+ "nvidia/Cosmos-Predict2.5-2B-Diffusers",
95
+ subfolder="transformer"
96
+ )
97
+
98
+ vae = AutoencoderKLWan.from_pretrained(
99
+ "nvidia/Cosmos-Predict2.5-2B-Diffusers",
100
+ subfolder="vae"
101
+ )
102
+
103
+ scheduler = FlowUniPCMultistepScheduler.from_pretrained(
104
+ "nvidia/Cosmos-Predict2.5-2B-Diffusers",
105
+ subfolder="scheduler"
106
+ )
107
+ ```
108
+
109
+ ## Key Differences from Official
110
+
111
+ 1. **Scheduler**: This model uses `FlowUniPCMultistepScheduler` (multi-step) which matches the official Cosmos 2.5 implementation, NOT `FlowMatchEulerDiscreteScheduler` (single-step) used in some FastVideo examples.
112
+
113
+ 2. **Weight Format**: Uses FastVideo-compatible weight format with proper key mapping.
114
+
115
+ 3. **Configuration**: All hyperparameters match the official Cosmos 2.5 2B model.
116
+
117
+ ## Inference Parameters
118
+
119
+ Recommended settings for best quality:
120
+
121
+ - **Resolution**: 480x832 (or multiples of 16)
122
+ - **Frames**: 121 (or any compatible length)
123
+ - **Steps**: 35 (with UniPC scheduler)
124
+ - **Guidance Scale**: 7.0
125
+ - **Scheduler Shift**: 5.0 (dynamic, applied during inference)
126
+ - **FPS**: 24.0
127
+
128
+ ## Model Information
129
+
130
+ - **Model Size**: ~2B parameters (transformer only)
131
+ - **Precision**: BFloat16
132
+ - **Context**: Trained for video prediction/generation
133
+ - **License**: Check NVIDIA's official license for Cosmos models
134
+
135
+ ## Citation
136
+
137
+ If you use this model, please cite:
138
+
139
+ ```bibtex
140
+ @misc{cosmos2024,
141
+ title={Cosmos: Foundation Models for Video Generation},
142
+ author={NVIDIA},
143
+ year={2024}
144
+ }
145
+ ```
146
+
147
+ ## Notes
148
+
149
+ 1. This is a Diffusers-compatible format but uses FastVideo classes, not standard Diffusers classes.
150
+ 2. The text encoder component needs to be loaded separately from the official cosmos_predict2 package.
151
+ 3. For best results, use the same scheduler (FlowUniPCMultistepScheduler) that the official model uses.
152
+ 4. The model expects text embeddings in the shape (batch, 512, 100352) - make sure your text encoder produces this format.
153
+
154
+
155
+
156
+
157
+
158
+
159
+
160
+
161
+
162
+
163
+
164
+
165
+
166
+
167
+
168
+
169
+
170
+
171
+
172
+
173
+
174
+
175
+
176
+
177
+
178
+
SETUP_INSTRUCTIONS.md ADDED
@@ -0,0 +1,118 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Setup Instructions for Cosmos-Predict2.5-2B-Diffusers
2
+
3
+ ## Missing Directories
4
+
5
+ You need to create the `text_encoder` directory:
6
+
7
+ ```bash
8
+ mkdir -p text_encoder
9
+ ```
10
+
11
+ Then create the following config files:
12
+
13
+ ## text_encoder/config.json
14
+
15
+ ```json
16
+ {
17
+ "_class_name": "Reason1TextEncoder",
18
+ "_diffusers_version": "0.34.0",
19
+ "_name_or_path": "Qwen/Qwen2.5-VL-7B-Instruct",
20
+ "tokenizer_type": "Qwen/Qwen2.5-VL-7B-Instruct",
21
+ "arch_config": {
22
+ "architectures": ["Qwen2_5_VLForConditionalGeneration"],
23
+ "model_type": "qwen2_5_vl",
24
+ "vocab_size": 152064,
25
+ "hidden_size": 3584,
26
+ "num_hidden_layers": 28,
27
+ "num_attention_heads": 28,
28
+ "num_key_value_heads": 4,
29
+ "intermediate_size": 18944,
30
+ "text_len": 512,
31
+ "hidden_state_skip_layer": 0,
32
+ "bos_token_id": 151643,
33
+ "pad_token_id": 151643,
34
+ "eos_token_id": 151645,
35
+ "image_token_id": 151655,
36
+ "video_token_id": 151656,
37
+ "vision_token_id": 151654,
38
+ "vision_start_token_id": 151652,
39
+ "vision_end_token_id": 151653,
40
+ "vision_config": null,
41
+ "rope_theta": 1000000.0,
42
+ "rope_scaling": {
43
+ "type": "mrope",
44
+ "mrope_section": [16, 24, 24]
45
+ },
46
+ "max_position_embeddings": 128000,
47
+ "max_window_layers": 28,
48
+ "embedding_concat_strategy": "mean_pooling",
49
+ "n_layers_per_group": 5,
50
+ "num_embedding_padding_tokens": 512,
51
+ "attention_dropout": 0.0,
52
+ "hidden_act": "silu",
53
+ "initializer_range": 0.02,
54
+ "rms_norm_eps": 1e-6,
55
+ "use_sliding_window": false,
56
+ "sliding_window": 32768,
57
+ "tie_word_embeddings": false,
58
+ "use_cache": false,
59
+ "output_hidden_states": true,
60
+ "torch_dtype": "bfloat16",
61
+ "_attn_implementation": "flash_attention_2"
62
+ }
63
+ }
64
+ ```
65
+
66
+ ## Complete Directory Structure
67
+
68
+ After setup, your model folder should look like this:
69
+
70
+ ```
71
+ models--nvidia--Cosmos-Predict2.5-2B-Diffusers/
72
+ ├── model_index.json
73
+ ├── README.md
74
+ ├── SETUP_INSTRUCTIONS.md (this file)
75
+ ├── text_encoder/
76
+ │ └── config.json
77
+ ├── tokenizer/
78
+ │ └── config.json
79
+ ├── transformer/
80
+ │ ├── config.json
81
+ │ └── 81edfebe-bd6a-4039-8c1d-737df1a790bf_ema_bf16.pt
82
+ ├── vae/
83
+ │ ├── config.json
84
+ │ └── tokenizer.pth
85
+ └── scheduler/
86
+ └── scheduler_config.json
87
+ ```
88
+
89
+ ## Notes
90
+
91
+ - The `text_encoder` and `tokenizer` configs reference `Qwen/Qwen2.5-VL-7B-Instruct`
92
+ - The actual Qwen model weights (~7B parameters) need to be downloaded separately from HuggingFace
93
+ - The Reason1TextEncoder will automatically load from the Qwen checkpoint path specified in the config
94
+
95
+
96
+
97
+
98
+
99
+
100
+
101
+
102
+
103
+
104
+
105
+
106
+
107
+
108
+
109
+
110
+
111
+
112
+
113
+
114
+
115
+
116
+
117
+
118
+
model_index.json ADDED
@@ -0,0 +1,46 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_class_name": "Cosmos2_5Pipeline",
3
+ "_diffusers_version": "0.34.0",
4
+ "_fastvideo_config": "Cosmos25Config",
5
+ "transformer": [
6
+ "fastvideo.models.dits.cosmos2_5",
7
+ "Cosmos25Transformer3DModel"
8
+ ],
9
+ "vae": [
10
+ "diffusers",
11
+ "AutoencoderKLWan"
12
+ ],
13
+ "text_encoder": [
14
+ "fastvideo.models.encoders.reason1",
15
+ "Reason1TextEncoder"
16
+ ],
17
+ "tokenizer": [
18
+ "transformers",
19
+ "AutoProcessor"
20
+ ],
21
+ "scheduler": [
22
+ "fastvideo.models.schedulers.scheduling_flow_unipc_multistep",
23
+ "FlowUniPCMultistepScheduler"
24
+ ],
25
+ "safety_checker": [
26
+ null,
27
+ null
28
+ ]
29
+ }
30
+
31
+
32
+
33
+
34
+
35
+
36
+
37
+
38
+
39
+
40
+
41
+
42
+
43
+
44
+
45
+
46
+
scheduler/scheduler_config.json ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_class_name": "FlowUniPCMultistepScheduler",
3
+ "_diffusers_version": "0.34.0",
4
+ "num_train_timesteps": 1000,
5
+ "solver_order": 2,
6
+ "prediction_type": "flow_prediction",
7
+ "shift": 1.0,
8
+ "use_dynamic_shifting": false,
9
+ "final_sigmas_type": "zero",
10
+ "predict_x0": true,
11
+ "solver_type": "bh2",
12
+ "lower_order_final": true,
13
+ "timestep_spacing": "linspace",
14
+ "steps_offset": 0
15
+ }
16
+
17
+
18
+
19
+
20
+
21
+
22
+
23
+
24
+
25
+
26
+
27
+
28
+
29
+
30
+
31
+
32
+
33
+
34
+
35
+
36
+
37
+
38
+
39
+
text_encoder/.gitattributes ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ *.7z filter=lfs diff=lfs merge=lfs -text
2
+ *.arrow filter=lfs diff=lfs merge=lfs -text
3
+ *.bin filter=lfs diff=lfs merge=lfs -text
4
+ *.bz2 filter=lfs diff=lfs merge=lfs -text
5
+ *.ckpt filter=lfs diff=lfs merge=lfs -text
6
+ *.ftz filter=lfs diff=lfs merge=lfs -text
7
+ *.gz filter=lfs diff=lfs merge=lfs -text
8
+ *.h5 filter=lfs diff=lfs merge=lfs -text
9
+ *.joblib filter=lfs diff=lfs merge=lfs -text
10
+ *.lfs.* filter=lfs diff=lfs merge=lfs -text
11
+ *.mlmodel filter=lfs diff=lfs merge=lfs -text
12
+ *.model filter=lfs diff=lfs merge=lfs -text
13
+ *.msgpack filter=lfs diff=lfs merge=lfs -text
14
+ *.npy filter=lfs diff=lfs merge=lfs -text
15
+ *.npz filter=lfs diff=lfs merge=lfs -text
16
+ *.onnx filter=lfs diff=lfs merge=lfs -text
17
+ *.ot filter=lfs diff=lfs merge=lfs -text
18
+ *.parquet filter=lfs diff=lfs merge=lfs -text
19
+ *.pb filter=lfs diff=lfs merge=lfs -text
20
+ *.pickle filter=lfs diff=lfs merge=lfs -text
21
+ *.pkl filter=lfs diff=lfs merge=lfs -text
22
+ *.pt filter=lfs diff=lfs merge=lfs -text
23
+ *.pth filter=lfs diff=lfs merge=lfs -text
24
+ *.rar filter=lfs diff=lfs merge=lfs -text
25
+ *.safetensors filter=lfs diff=lfs merge=lfs -text
26
+ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
27
+ *.tar.* filter=lfs diff=lfs merge=lfs -text
28
+ *.tar filter=lfs diff=lfs merge=lfs -text
29
+ *.tflite filter=lfs diff=lfs merge=lfs -text
30
+ *.tgz filter=lfs diff=lfs merge=lfs -text
31
+ *.wasm filter=lfs diff=lfs merge=lfs -text
32
+ *.xz filter=lfs diff=lfs merge=lfs -text
33
+ *.zip filter=lfs diff=lfs merge=lfs -text
34
+ *.zst filter=lfs diff=lfs merge=lfs -text
35
+ *tfevents* filter=lfs diff=lfs merge=lfs -text
text_encoder/README.md ADDED
@@ -0,0 +1,384 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: other
3
+ license_name: nvidia-open-model-license
4
+ license_link: >-
5
+ https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license
6
+ datasets:
7
+ - nvidia/Cosmos-Reason1-SFT-Dataset
8
+ - nvidia/Cosmos-Reason1-RL-Dataset
9
+ - nvidia/Cosmos-Reason1-Benchmark
10
+ library_name: transformers
11
+ language:
12
+ - en
13
+ base_model:
14
+ - Qwen/Qwen2.5-VL-7B-Instruct
15
+ tags:
16
+ - nvidia
17
+ - cosmos
18
+ pipeline_tag: image-text-to-text
19
+ ---
20
+
21
+ # **Cosmos-Reason1: Physical AI Common Sense and Embodied Reasoning Models**
22
+
23
+ [**Cosmos**](https://huggingface.co/collections/nvidia/cosmos-reason1-67c9e926206426008f1da1b7) | [**Code**](https://github.com/nvidia-cosmos/cosmos-reason1) | [**Paper**](https://arxiv.org/abs/2503.15558) | [**Paper Website**](https://research.nvidia.com/labs/dir/cosmos-reason1)
24
+
25
+ # Model Overview
26
+
27
+ ## Description:
28
+
29
+ NVIDIA Cosmos Reason – an open, customizable, 7B-parameter reasoning vision language model (VLM) for physical AI and robotics - enables robots and vision AI agents to reason like humans, using prior knowledge, physics understanding and common sense to understand and act in the real world. This model understands space, time, and fundamental physics, and can serve as a planning model to reason what steps an embodied agent might take next.
30
+
31
+ Cosmos Reason excels at navigating the long tail of diverse scenarios of the physical world with spatial-temporal understanding. Cosmos Reason is post-trained with physical common sense and embodied reasoning data with supervised fine-tuning and reinforcement learning. It uses chain-of-thought reasoning capabilities to understand world dynamics without human annotations.
32
+
33
+ Given a video/image and a text prompt, the model first converts the video/image into tokens using a vision encoder and a special translator called a projector. These video tokens are combined with the text prompt and fed into the core model, which uses a mix of LLM modules and techniques. This enables the model to think step-by-step and provide detailed, logical responses.
34
+
35
+ Cosmos Reason can be used for robotics and physical AI applications including:
36
+ - Data curation and annotation — Enable developers to automate high-quality curation and annotation of massive, diverse training datasets.
37
+ - Robot planning and reasoning — Act as the brain for deliberate, methodical decision-making in a robot vision language action (VLA) model. Now robots such as humanoids and autonomous vehicles can interpret environments and given complex commands, break them down into tasks and execute them using common sense, even in unfamiliar environments.
38
+ - Video analytics AI agents — Extract valuable insights and perform root-cause analysis on massive volumes of video data. These agents can be used to analyze and understand recorded or live video streams across city and industrial operations.
39
+
40
+ The model is ready for commercial use.
41
+
42
+ **Model Developer**: NVIDIA
43
+
44
+ ## Model Versions
45
+
46
+ The Cosmos-Reason1 includes the following model:
47
+
48
+ - [Cosmos-Reason1-7B](https://huggingface.co/nvidia/Cosmos-Reason1-7B): Given a text prompt and an input video, think and generate the answer with respect to the input text prompt and video.
49
+
50
+ ### License:
51
+
52
+ This model is released under the [NVIDIA Open Model License](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license). Additional Information: [Apache License 2.0](https://huggingface.co/datasets/choosealicense/licenses/blob/main/markdown/apache-2.0.md).
53
+
54
+ For a custom license, please contact [cosmos-license@nvidia.com](mailto:cosmos-license@nvidia.com).
55
+
56
+ Under the NVIDIA Open Model License, NVIDIA confirms:
57
+
58
+ * Models are commercially usable.
59
+ * You are free to create and distribute Derivative Models.
60
+ * NVIDIA does not claim ownership to any outputs generated using the Models or Derivative Models.
61
+
62
+ **Important Note**: If You bypass, disable, reduce the efficacy of, or circumvent any technical limitation, safety guardrail or associated safety guardrail hyperparameter, encryption, security, digital rights management, or authentication mechanism (collectively “Guardrail”) contained in the Model without a substantially similar Guardrail appropriate for your use case, your rights under this Agreement [NVIDIA Open Model License Agreement](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license) will automatically terminate.
63
+
64
+ ### Deployment Geography:
65
+
66
+ Global
67
+
68
+ ### Use Case:
69
+
70
+ Physical AI: Space, time, fundamental physics understanding and embodied reasoning, encompassing robotics, and autonomous vehicles (AV).
71
+
72
+ ### Release Date:
73
+
74
+ * Github: [05/17/2025](https://github.com/nvidia-cosmos/cosmos-reason1)
75
+ * Huggingface:
76
+ * [08/01/2025](https://huggingface.co/nvidia/Cosmos-Reason1-7B/commit/0caf724f837efea5e25bf6d5818dcdeec0a36604). Shipped a few improvements which include captions with temporal timestamp, Set of Mark prompting.
77
+ * [06/10/2025](https://huggingface.co/nvidia/Cosmos-Reason1-7B/commit/2464fff43c5c0bfb1916ac8c009feda4aed81be9). Enhanced critic capability for physical plausibility.
78
+ * [05/17/2025](https://huggingface.co/nvidia/Cosmos-Reason1-7B/commit/098a5bb62a1f4fc05e5c4ac89aae8005e301aa18). Initial release.
79
+
80
+ ## Model Architecture:
81
+
82
+ Architecture Type: A Multi-modal LLM consists of a Vision Transformer (ViT) for vision encoder and a Dense Transformer model for LLM.
83
+ Network Architecture: Qwen2.5-VL-7B-Instruct.
84
+
85
+ Cosmos-Reason-7B is post-trained based on [Qwen2.5-VL-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-VL-7B-Instruct) and follows the same model architecture.
86
+
87
+
88
+ **Number of model parameters:**
89
+
90
+ Cosmos-Reason1-7B:<br>
91
+ * Vision Transformer (ViT): 675.76M (675,759,104)
92
+ * Language Model (LLM): 7.07B (7,070,619,136)
93
+ * Other components (output projection layer): 545.00M (544,997,376)
94
+
95
+
96
+ ## Computational Load:
97
+
98
+ * Cumulative Compute: 3.2603016e+21 FLOPS
99
+ * Estimated Energy and Emissions for Model Training:
100
+ * Total kWh = 16658432
101
+ * Total Emissions (tCO2e) = 5380.674
102
+
103
+ ## Input
104
+
105
+ **Input Type(s)**: Text+Video/Image
106
+
107
+ **Input Format(s)**:
108
+ * Text: String
109
+ * Video: mp4
110
+ * Image: jpg
111
+
112
+ **Input Parameters**:
113
+ * Text: One-dimensional (1D)
114
+ * Video: Three-dimensional (3D)
115
+ * Image: Two-dimensional (2D)
116
+
117
+ **Other Properties Related to Input**:
118
+ * Use `FPS=4` for input video to match the training setup.
119
+ * Append `Answer the question in the following format: <think>\nyour reasoning\n</think>\n\n<answer>\nyour answer\n</answer>.` in the system prompt to encourage long chain-of-thought reasoning response.
120
+
121
+ ## Output
122
+
123
+ **Output Type(s)**: Text
124
+
125
+ **Output Format**: String
126
+
127
+ **Output Parameters**: Text: One-dimensional (1D)
128
+
129
+ **Other Properties Related to Output**:
130
+ * Recommend using 4096 or more output max tokens to avoid truncation of long chain-of-thought response.
131
+
132
+ * Our AI model recognizes timestamps added at the bottom of each frame for accurate temporal localization.
133
+
134
+ * Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated systems. By leveraging NVIDIA’s hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions. <br>
135
+
136
+
137
+ ## Software Integration
138
+
139
+ **Runtime Engine(s):**
140
+
141
+ * [vLLM](https://github.com/vllm-project/vllm)
142
+
143
+ **Supported Hardware Microarchitecture Compatibility:**
144
+
145
+ * NVIDIA Blackwell
146
+ * NVIDIA Hopper
147
+
148
+ **Note**: We have only tested doing inference with BF16 precision.
149
+
150
+ **Operating System(s):**
151
+
152
+ * Linux (We have not tested on other operating systems.)
153
+
154
+
155
+ # Usage
156
+
157
+ See [Cosmos-Reason1](https://github.com/nvidia-cosmos/cosmos-reason1) for details.
158
+ * Post Training: [Cosmos-Reason1](https://github.com/nvidia-cosmos/cosmos-reason1) provides examples of supervised fine-tuning and reinforcement learning on embodied reasoning datasets.
159
+
160
+ ## Training and Evaluation Sections:
161
+ ### 05/17/2025
162
+ Please see our [technical paper](https://arxiv.org/pdf/2503.15558) for detailed evaluations on physical common sense and embodied reasoning. Part of the evaluation datasets are released under [Cosmos-Reason1-Benchmark](https://huggingface.co/datasets/nvidia/Cosmos-Reason1-Benchmark). The embodied reasoning datasets and benchmarks focus on the following areas: robotics (RoboVQA, BridgeDataV2, Agibot, RobFail), ego-centric human demonstration (HoloAssist), and Autonomous Vehicle (AV) driving video data. The AV dataset is collected and annotated by NVIDIA.
163
+
164
+ All datasets go through the data annotation process described in the technical paper to prepare training and evaluation data and annotations.
165
+
166
+ ### 08/01/2025
167
+ We enhance the model capability with the augmented training data. PLM-Video-Human and Nexar are used to enable dense temporal captioning. Describe Anything is added to enhance a set of mark (SoM) prompting. We enrich data in intelligent transportation systems (ITS) and warehouse applications. Lastly, Visual Critics dataset contains a collection of AI generated videos from Cosmos-Predict2 and Wan2.1 with human annotations to describe the physical correctness in AI videos.
168
+
169
+
170
+ ## Training Datasets:
171
+
172
+ **Data Collection Method**:
173
+ * RoboVQA: Hybrid: Automatic/Sensors
174
+ * BridgeDataV2: Automatic/Sensors
175
+ * AgiBot: Automatic/Sensors
176
+ * RoboFail: Automatic/Sensors
177
+ * HoloAssist: Human
178
+ * AV: Automatic/Sensors
179
+ * PLM-Video-Human: Human
180
+ * Nexar: Automatic/Sensors
181
+ * Describe Anything: Human
182
+ * ITS / Warehouse: Human, Automatic
183
+ * Visual Critics: Automatic
184
+
185
+ **Labeling Method**:
186
+ * RoboVQA: Hybrid: Human,Automated
187
+ * BridgeDataV2: Hybrid: Human,Automated
188
+ * AgiBot: Hybrid: Human,Automated
189
+ * RoboFail: Hybrid: Human,Automated
190
+ * HoloAssist: Hybrid: Human,Automated
191
+ * AV: Hybrid: Human,Automated
192
+ * PLM-Video-Human: Human,Automated
193
+ * Nexar: Human
194
+ * Describe Anything: Human,Automated
195
+ * ITS / Warehouse: Human, Automated
196
+ * Visual Critics: Human,Automated
197
+
198
+
199
+ # Evaluation Datasets:
200
+
201
+ **Data Collection Method**:
202
+ * RoboVQA: Hybrid: Automatic/Sensors
203
+ * BridgeDataV2: Automatic/Sensors
204
+ * AgiBot: Automatic/Sensors
205
+ * RoboFail: Automatic/Sensors
206
+ * HoloAssist: Human
207
+ * AV: Automatic/Sensors
208
+
209
+
210
+ **Labeling Method**:
211
+ * RoboVQA: Hybrid: Human,Automated
212
+ * BridgeDataV2: Hybrid: Human,Automated
213
+ * AgiBot: Hybrid: Human,Automated
214
+ * RoboFail: Hybrid: Human,Automated
215
+ * HoloAssist: Hybrid: Human,Automated
216
+ * AV: Hybrid: Human,Automated
217
+
218
+
219
+ **Metrics**:
220
+ We report the model accuracy on the embodied reasoning benchmark introduced in [Cosmos-Reason1](https://arxiv.org/abs/2503.15558). The results differ from those presented in Table 9 due to additional training aimed at supporting a broader range of Physical AI tasks beyond the benchmark.
221
+ | | [RoboVQA](https://robovqa.github.io/) | AV | [BridgeDataV2](https://rail-berkeley.github.io/bridgedata/)| [Agibot](https://github.com/OpenDriveLab/AgiBot-World)| [HoloAssist](https://holoassist.github.io/) | [RoboFail](https://robot-reflect.github.io/) | Average |
222
+ |--------------------|---------------------------------------------|----------|------------------------------------------------------|------------------------------------------------|------------------------------------------------|------------------------------------------------|------------------------------------------------|
223
+ | **Accuracy** | 87.3 | 70.8 | 63.7 | 48.9 | 62.7 | 57.2 | 65.1 |
224
+
225
+ ## Dataset Format
226
+ Modality: Video (mp4) and Text
227
+
228
+ ## Dataset Quantification
229
+ ### 05/17/2025
230
+ We release the embodied reasoning data and benchmarks. Each data sample is a pair of video and text. The text annotations include understanding and reasoning annotations described in the Cosmos-Reason1 paper. Each video may have multiple text annotations. The quantity of the video and text pairs is described in the table below.
231
+ **The AV data is currently unavailable and will be uploaded soon!**
232
+
233
+ | | [RoboVQA](https://robovqa.github.io/) | AV | [BridgeDataV2](https://rail-berkeley.github.io/bridgedata/)| [Agibot](https://github.com/OpenDriveLab/AgiBot-World)| [HoloAssist](https://holoassist.github.io/) | [RoboFail](https://robot-reflect.github.io/) | Total Storage Size |
234
+ |--------------------|---------------------------------------------|----------|------------------------------------------------------|------------------------------------------------|------------------------------------------------|------------------------------------------------|--------------------|
235
+ | **SFT Data** | 1.14m | 24.7k | 258k | 38.9k | 273k | N/A | **300.6GB** |
236
+ | **RL Data** | 252 | 200 | 240 | 200 | 200 | N/A | **2.6GB** |
237
+ | **Benchmark Data** | 110 | 100 | 100 | 100 | 100 | 100 | **1.5GB** |
238
+
239
+ We release text annotations for all embodied reasoning datasets and videos for RoboVQA and AV datasets. For other datasets, users may download the source videos from the original data source and find corresponding video sources via the video names. The held-out RoboFail benchmark is released for measuring the generalization capability.
240
+
241
+ ### 08/01/2025
242
+ | | [PLM-Video-Human](https://huggingface.co/datasets/facebook/PLM-Video-Human) | Nexar | [Describe Anything](https://huggingface.co/datasets/nvidia/describe-anything-dataset)| [ITS / Warehouse] | Visual Critics | Total Storage Size |
243
+ |------------------ |-----------------------------------------------------------------------------|-------------|--------------------------------------------------------------------------------------|-------------------------|--------------------------------------------|--------------------|
244
+ | **SFT Data** | 39k | 240k | 178k | 24k | 24k | **2.6TB** |
245
+
246
+
247
+
248
+ ## Inference:
249
+ **Test Hardware:** H100, A100, GB200 <br>
250
+ > [!NOTE]
251
+ > We suggest using `fps=4` for the input video and `max_tokens=4096` to avoid truncated response.
252
+ ```python
253
+ from transformers import AutoProcessor
254
+ from vllm import LLM, SamplingParams
255
+ from qwen_vl_utils import process_vision_info
256
+
257
+ # You can also replace the MODEL_PATH by a safetensors folder path mentioned above
258
+ MODEL_PATH = "nvidia/Cosmos-Reason1-7B"
259
+
260
+ llm = LLM(
261
+ model=MODEL_PATH,
262
+ limit_mm_per_prompt={"image": 10, "video": 10},
263
+ )
264
+
265
+ sampling_params = SamplingParams(
266
+ temperature=0.6,
267
+ top_p=0.95,
268
+ repetition_penalty=1.05,
269
+ max_tokens=4096,
270
+ )
271
+
272
+ video_messages = [
273
+ {"role": "system", "content": "You are a helpful assistant. Answer the question in the following format: <think>\nyour reasoning\n</think>\n\n<answer>\nyour answer\n</answer>."},
274
+ {"role": "user", "content": [
275
+ {"type": "text", "text": (
276
+ "Is it safe to turn right?"
277
+ )
278
+ },
279
+ {
280
+ "type": "video",
281
+ "video": "file:///path/to/your/video.mp4",
282
+ "fps": 4,
283
+ }
284
+ ]
285
+ },
286
+ ]
287
+
288
+ # Here we use video messages as a demonstration
289
+ messages = video_messages
290
+
291
+ processor = AutoProcessor.from_pretrained(MODEL_PATH)
292
+ prompt = processor.apply_chat_template(
293
+ messages,
294
+ tokenize=False,
295
+ add_generation_prompt=True,
296
+ )
297
+ image_inputs, video_inputs, video_kwargs = process_vision_info(messages, return_video_kwargs=True)
298
+
299
+ mm_data = {}
300
+ if image_inputs is not None:
301
+ mm_data["image"] = image_inputs
302
+ if video_inputs is not None:
303
+ mm_data["video"] = video_inputs
304
+
305
+ llm_inputs = {
306
+ "prompt": prompt,
307
+ "multi_modal_data": mm_data,
308
+
309
+ # FPS will be returned in video_kwargs
310
+ "mm_processor_kwargs": video_kwargs,
311
+ }
312
+
313
+ outputs = llm.generate([llm_inputs], sampling_params=sampling_params)
314
+ generated_text = outputs[0].outputs[0].text
315
+
316
+ print(generated_text)
317
+ ```
318
+
319
+
320
+ ## Ethical Considerations
321
+
322
+ NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.
323
+
324
+ Users are responsible for model inputs and outputs. Users are responsible for ensuring safe integration of this model, including implementing guardrails as well as other safety mechanisms, prior to deployment.
325
+
326
+ For more detailed information on ethical considerations for this model, please see the subcards of Explainability, Bias, Safety & Security, and Privacy below.
327
+
328
+ Please report security vulnerabilities or NVIDIA AI Concerns [here](https://www.nvidia.com/en-us/support/submit-security-vulnerability/).
329
+
330
+ ### Plus Plus (++) Promise
331
+
332
+ We value you, the datasets, the diversity they represent, and what we have been entrusted with. This model and its associated data have been:
333
+
334
+ * Verified to comply with current applicable disclosure laws, regulations, and industry standards.
335
+ * Verified to comply with applicable privacy labeling requirements.
336
+ * Annotated to describe the collector/source (NVIDIA or a third-party).
337
+ * Characterized for technical limitations.
338
+ * Reviewed to ensure proper disclosure is accessible to, maintained for, and in compliance with NVIDIA data subjects and their requests.
339
+ * Reviewed before release.
340
+ * Tagged for known restrictions and potential safety implications.
341
+
342
+ ### Bias
343
+
344
+ | Field | Response |
345
+ | :--------------------------------------------------------------------------------------------------------------------------------------------------------------- | :------- |
346
+ | Participation considerations from adversely impacted groups [protected classes](https://www.senate.ca.gov/content/protected-classes) in model design and testing: | None |
347
+ | Measures taken to mitigate against unwanted bias: | The training video sources contain multiple physical embodiments and environments including human, car, single arm robot, bimanual robot in indoor and outdoor environments. By training on numerous and various physical interactions and curated datasets, we strive to provide a model that does not possess biases towards certain embodiments or environments. |
348
+
349
+ ### Explainability
350
+
351
+ | Field | Response |
352
+ | :-------------------------------------------------------- | :------------------------------------------------------------------------------------------------------------------- |
353
+ | Intended Application & Domain: | Physical AI Reasoning |
354
+ | Model Type: | Transformer |
355
+ | Intended Users: | Physical AI developers |
356
+ | Output: | Text |
357
+ | Describe how the model works: | Given a video/image and a text prompt, the model first converts the video/image into tokens using a vision encoder and a special translator called a projector. These video tokens are combined with the text prompt and fed into the core model, which uses a mix of LLM modules and techniques. This enables the model to think step-by-step and provide detailed, logical responses. |
358
+ | Technical Limitations: | The model may not follow the video or text input accurately in challenging cases, where the input video shows complex scene composition and temporal dynamics. Examples of challenging scenes include: fast camera movements, overlapping human-object interactions, low lighting with high motion blur, and multiple people performing different actions simultaneously. |
359
+ | Verified to have met prescribed NVIDIA quality standards: | Yes |
360
+ | Performance Metrics: | Quantitative and Qualitative Evaluation. Cosmos-Reason1 proposes the embodied reasoning benchmark and physical common sense benchmark to evaluate accuracy with visual question answering. |
361
+ | Potential Known Risks: | The model's output can generate all forms of texts, including what may be considered toxic, offensive, or indecent. |
362
+ | Licensing: | [NVIDIA Open Model License](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license). Additional Information: [Apache License 2.0](https://huggingface.co/datasets/choosealicense/licenses/blob/main/markdown/apache-2.0.md). |
363
+
364
+ ### Privacy
365
+
366
+ | Field | Response |
367
+ | :------------------------------------------------------------------ | :------------- |
368
+ | Generatable or reverse engineerable personal information? | None Known |
369
+ | Protected class data used to create this model? | None Known |
370
+ | Was consent obtained for any personal data used? | None Known |
371
+ | How often is dataset reviewed? | Before Release |
372
+ | Is there provenance for all datasets used in training? | Yes |
373
+ | Does data labeling (annotation, metadata) comply with privacy laws? | Yes |
374
+ | Applicable Privacy Policy | [NVIDIA Privacy Policy](https://www.nvidia.com/en-us/about-nvidia/privacy-policy) |
375
+
376
+
377
+ ### Safety
378
+
379
+ | Field | Response |
380
+ | :---------------------------------------------- | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
381
+ | Model Application(s): | Physical AI common sense understanding and embodied reasoning |
382
+ | Describe the life critical impact (if present). | None Known |
383
+ | Use Case Restrictions: | [NVIDIA Open Model License](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license). Additional Information: [Apache License 2.0](https://huggingface.co/datasets/choosealicense/licenses/blob/main/markdown/apache-2.0.md). |
384
+ | Model and dataset restrictions: | The Principle of least privilege (PoLP) is applied limiting access for dataset generation and model development. Restrictions enforce dataset access during training, and dataset license constraints adhered to. Model checkpoints are made available on Hugging Face, and may become available on cloud providers' model catalog. |
text_encoder/chat_template.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ {
2
+ "chat_template": "{% set image_count = namespace(value=0) %}{% set video_count = namespace(value=0) %}{% for message in messages %}{% if loop.first and message['role'] != 'system' %}<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n{% endif %}<|im_start|>{{ message['role'] }}\n{% if message['content'] is string %}{{ message['content'] }}<|im_end|>\n{% else %}{% for content in message['content'] %}{% if content['type'] == 'image' or 'image' in content or 'image_url' in content %}{% set image_count.value = image_count.value + 1 %}{% if add_vision_id %}Picture {{ image_count.value }}: {% endif %}<|vision_start|><|image_pad|><|vision_end|>{% elif content['type'] == 'video' or 'video' in content %}{% set video_count.value = video_count.value + 1 %}{% if add_vision_id %}Video {{ video_count.value }}: {% endif %}<|vision_start|><|video_pad|><|vision_end|>{% elif 'text' in content %}{{ content['text'] }}{% endif %}{% endfor %}<|im_end|>\n{% endif %}{% endfor %}{% if add_generation_prompt %}<|im_start|>assistant\n{% endif %}"
3
+ }
text_encoder/config.json ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen2_5_VLForConditionalGeneration"
4
+ ],
5
+ "attention_dropout": 0.0,
6
+ "bos_token_id": 151643,
7
+ "eos_token_id": 151645,
8
+ "vision_start_token_id": 151652,
9
+ "vision_end_token_id": 151653,
10
+ "vision_token_id": 151654,
11
+ "image_token_id": 151655,
12
+ "video_token_id": 151656,
13
+ "hidden_act": "silu",
14
+ "hidden_size": 3584,
15
+ "initializer_range": 0.02,
16
+ "intermediate_size": 18944,
17
+ "max_position_embeddings": 128000,
18
+ "max_window_layers": 28,
19
+ "model_type": "qwen2_5_vl",
20
+ "num_attention_heads": 28,
21
+ "num_hidden_layers": 28,
22
+ "num_key_value_heads": 4,
23
+ "rms_norm_eps": 1e-06,
24
+ "rope_theta": 1000000.0,
25
+ "sliding_window": 32768,
26
+ "tie_word_embeddings": false,
27
+ "torch_dtype": "bfloat16",
28
+ "transformers_version": "4.41.2",
29
+ "use_cache": true,
30
+ "use_sliding_window": false,
31
+ "vision_config": {
32
+ "depth": 32,
33
+ "hidden_act": "silu",
34
+ "hidden_size": 1280,
35
+ "intermediate_size": 3420,
36
+ "num_heads": 16,
37
+ "in_chans": 3,
38
+ "out_hidden_size": 3584,
39
+ "patch_size": 14,
40
+ "spatial_merge_size": 2,
41
+ "spatial_patch_size": 14,
42
+ "window_size": 112,
43
+ "fullatt_block_indexes": [
44
+ 7,
45
+ 15,
46
+ 23,
47
+ 31
48
+ ],
49
+ "tokens_per_second": 2,
50
+ "temporal_patch_size": 2
51
+ },
52
+ "rope_scaling": {
53
+ "type": "mrope",
54
+ "mrope_section": [
55
+ 16,
56
+ 24,
57
+ 24
58
+ ]
59
+ },
60
+ "vocab_size": 152064
61
+ }
text_encoder/generation_config.json ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 151643,
3
+ "pad_token_id": 151643,
4
+ "do_sample": true,
5
+ "eos_token_id": [
6
+ 151645,
7
+ 151643
8
+ ],
9
+ "repetition_penalty": 1.05,
10
+ "temperature": 0.000001,
11
+ "transformers_version": "4.37.0"
12
+ }
text_encoder/model-00001-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c28404126221997ae8eb70a23b919c96174d42e35ae1d537e0c95093d50b359a
3
+ size 4968243304
text_encoder/model-00002-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f281081864c10992d3e03874c79d526c84407e049d713747f19eb9c79cd16db3
3
+ size 4991495816
text_encoder/model-00003-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3bb3a62a8d0e83c6283388ddea99395b221f908f9181b8edd0f7f91d02260ebe
3
+ size 4932751040
text_encoder/model-00004-of-00004.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:91bacbe1ad798e16daa05023b4e4bec70b53c8cd7d757db86c5bc76c4e0bbf15
3
+ size 1691924384
text_encoder/model.safetensors.index.json ADDED
@@ -0,0 +1,736 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "metadata": {
3
+ "total_size": 16584333312
4
+ },
5
+ "weight_map": {
6
+ "visual.patch_embed.proj.weight": "model-00001-of-00004.safetensors",
7
+ "visual.blocks.0.norm1.weight": "model-00001-of-00004.safetensors",
8
+ "visual.blocks.0.norm2.weight": "model-00001-of-00004.safetensors",
9
+ "visual.blocks.0.attn.qkv.weight": "model-00001-of-00004.safetensors",
10
+ "visual.blocks.0.attn.qkv.bias": "model-00001-of-00004.safetensors",
11
+ "visual.blocks.0.attn.proj.weight": "model-00001-of-00004.safetensors",
12
+ "visual.blocks.0.attn.proj.bias": "model-00001-of-00004.safetensors",
13
+ "visual.blocks.0.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
14
+ "visual.blocks.0.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
15
+ "visual.blocks.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
16
+ "visual.blocks.0.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
17
+ "visual.blocks.0.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
18
+ "visual.blocks.0.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
19
+ "visual.blocks.1.norm1.weight": "model-00001-of-00004.safetensors",
20
+ "visual.blocks.1.norm2.weight": "model-00001-of-00004.safetensors",
21
+ "visual.blocks.1.attn.qkv.weight": "model-00001-of-00004.safetensors",
22
+ "visual.blocks.1.attn.qkv.bias": "model-00001-of-00004.safetensors",
23
+ "visual.blocks.1.attn.proj.weight": "model-00001-of-00004.safetensors",
24
+ "visual.blocks.1.attn.proj.bias": "model-00001-of-00004.safetensors",
25
+ "visual.blocks.1.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
26
+ "visual.blocks.1.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
27
+ "visual.blocks.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
28
+ "visual.blocks.1.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
29
+ "visual.blocks.1.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
30
+ "visual.blocks.1.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
31
+ "visual.blocks.2.norm1.weight": "model-00001-of-00004.safetensors",
32
+ "visual.blocks.2.norm2.weight": "model-00001-of-00004.safetensors",
33
+ "visual.blocks.2.attn.qkv.weight": "model-00001-of-00004.safetensors",
34
+ "visual.blocks.2.attn.qkv.bias": "model-00001-of-00004.safetensors",
35
+ "visual.blocks.2.attn.proj.weight": "model-00001-of-00004.safetensors",
36
+ "visual.blocks.2.attn.proj.bias": "model-00001-of-00004.safetensors",
37
+ "visual.blocks.2.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
38
+ "visual.blocks.2.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
39
+ "visual.blocks.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
40
+ "visual.blocks.2.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
41
+ "visual.blocks.2.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
42
+ "visual.blocks.2.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
43
+ "visual.blocks.3.norm1.weight": "model-00001-of-00004.safetensors",
44
+ "visual.blocks.3.norm2.weight": "model-00001-of-00004.safetensors",
45
+ "visual.blocks.3.attn.qkv.weight": "model-00001-of-00004.safetensors",
46
+ "visual.blocks.3.attn.qkv.bias": "model-00001-of-00004.safetensors",
47
+ "visual.blocks.3.attn.proj.weight": "model-00001-of-00004.safetensors",
48
+ "visual.blocks.3.attn.proj.bias": "model-00001-of-00004.safetensors",
49
+ "visual.blocks.3.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
50
+ "visual.blocks.3.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
51
+ "visual.blocks.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
52
+ "visual.blocks.3.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
53
+ "visual.blocks.3.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
54
+ "visual.blocks.3.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
55
+ "visual.blocks.4.norm1.weight": "model-00001-of-00004.safetensors",
56
+ "visual.blocks.4.norm2.weight": "model-00001-of-00004.safetensors",
57
+ "visual.blocks.4.attn.qkv.weight": "model-00001-of-00004.safetensors",
58
+ "visual.blocks.4.attn.qkv.bias": "model-00001-of-00004.safetensors",
59
+ "visual.blocks.4.attn.proj.weight": "model-00001-of-00004.safetensors",
60
+ "visual.blocks.4.attn.proj.bias": "model-00001-of-00004.safetensors",
61
+ "visual.blocks.4.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
62
+ "visual.blocks.4.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
63
+ "visual.blocks.4.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
64
+ "visual.blocks.4.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
65
+ "visual.blocks.4.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
66
+ "visual.blocks.4.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
67
+ "visual.blocks.5.norm1.weight": "model-00001-of-00004.safetensors",
68
+ "visual.blocks.5.norm2.weight": "model-00001-of-00004.safetensors",
69
+ "visual.blocks.5.attn.qkv.weight": "model-00001-of-00004.safetensors",
70
+ "visual.blocks.5.attn.qkv.bias": "model-00001-of-00004.safetensors",
71
+ "visual.blocks.5.attn.proj.weight": "model-00001-of-00004.safetensors",
72
+ "visual.blocks.5.attn.proj.bias": "model-00001-of-00004.safetensors",
73
+ "visual.blocks.5.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
74
+ "visual.blocks.5.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
75
+ "visual.blocks.5.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
76
+ "visual.blocks.5.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
77
+ "visual.blocks.5.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
78
+ "visual.blocks.5.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
79
+ "visual.blocks.6.norm1.weight": "model-00001-of-00004.safetensors",
80
+ "visual.blocks.6.norm2.weight": "model-00001-of-00004.safetensors",
81
+ "visual.blocks.6.attn.qkv.weight": "model-00001-of-00004.safetensors",
82
+ "visual.blocks.6.attn.qkv.bias": "model-00001-of-00004.safetensors",
83
+ "visual.blocks.6.attn.proj.weight": "model-00001-of-00004.safetensors",
84
+ "visual.blocks.6.attn.proj.bias": "model-00001-of-00004.safetensors",
85
+ "visual.blocks.6.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
86
+ "visual.blocks.6.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
87
+ "visual.blocks.6.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
88
+ "visual.blocks.6.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
89
+ "visual.blocks.6.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
90
+ "visual.blocks.6.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
91
+ "visual.blocks.7.norm1.weight": "model-00001-of-00004.safetensors",
92
+ "visual.blocks.7.norm2.weight": "model-00001-of-00004.safetensors",
93
+ "visual.blocks.7.attn.qkv.weight": "model-00001-of-00004.safetensors",
94
+ "visual.blocks.7.attn.qkv.bias": "model-00001-of-00004.safetensors",
95
+ "visual.blocks.7.attn.proj.weight": "model-00001-of-00004.safetensors",
96
+ "visual.blocks.7.attn.proj.bias": "model-00001-of-00004.safetensors",
97
+ "visual.blocks.7.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
98
+ "visual.blocks.7.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
99
+ "visual.blocks.7.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
100
+ "visual.blocks.7.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
101
+ "visual.blocks.7.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
102
+ "visual.blocks.7.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
103
+ "visual.blocks.8.norm1.weight": "model-00001-of-00004.safetensors",
104
+ "visual.blocks.8.norm2.weight": "model-00001-of-00004.safetensors",
105
+ "visual.blocks.8.attn.qkv.weight": "model-00001-of-00004.safetensors",
106
+ "visual.blocks.8.attn.qkv.bias": "model-00001-of-00004.safetensors",
107
+ "visual.blocks.8.attn.proj.weight": "model-00001-of-00004.safetensors",
108
+ "visual.blocks.8.attn.proj.bias": "model-00001-of-00004.safetensors",
109
+ "visual.blocks.8.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
110
+ "visual.blocks.8.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
111
+ "visual.blocks.8.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
112
+ "visual.blocks.8.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
113
+ "visual.blocks.8.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
114
+ "visual.blocks.8.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
115
+ "visual.blocks.9.norm1.weight": "model-00001-of-00004.safetensors",
116
+ "visual.blocks.9.norm2.weight": "model-00001-of-00004.safetensors",
117
+ "visual.blocks.9.attn.qkv.weight": "model-00001-of-00004.safetensors",
118
+ "visual.blocks.9.attn.qkv.bias": "model-00001-of-00004.safetensors",
119
+ "visual.blocks.9.attn.proj.weight": "model-00001-of-00004.safetensors",
120
+ "visual.blocks.9.attn.proj.bias": "model-00001-of-00004.safetensors",
121
+ "visual.blocks.9.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
122
+ "visual.blocks.9.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
123
+ "visual.blocks.9.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
124
+ "visual.blocks.9.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
125
+ "visual.blocks.9.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
126
+ "visual.blocks.9.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
127
+ "visual.blocks.10.norm1.weight": "model-00001-of-00004.safetensors",
128
+ "visual.blocks.10.norm2.weight": "model-00001-of-00004.safetensors",
129
+ "visual.blocks.10.attn.qkv.weight": "model-00001-of-00004.safetensors",
130
+ "visual.blocks.10.attn.qkv.bias": "model-00001-of-00004.safetensors",
131
+ "visual.blocks.10.attn.proj.weight": "model-00001-of-00004.safetensors",
132
+ "visual.blocks.10.attn.proj.bias": "model-00001-of-00004.safetensors",
133
+ "visual.blocks.10.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
134
+ "visual.blocks.10.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
135
+ "visual.blocks.10.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
136
+ "visual.blocks.10.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
137
+ "visual.blocks.10.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
138
+ "visual.blocks.10.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
139
+ "visual.blocks.11.norm1.weight": "model-00001-of-00004.safetensors",
140
+ "visual.blocks.11.norm2.weight": "model-00001-of-00004.safetensors",
141
+ "visual.blocks.11.attn.qkv.weight": "model-00001-of-00004.safetensors",
142
+ "visual.blocks.11.attn.qkv.bias": "model-00001-of-00004.safetensors",
143
+ "visual.blocks.11.attn.proj.weight": "model-00001-of-00004.safetensors",
144
+ "visual.blocks.11.attn.proj.bias": "model-00001-of-00004.safetensors",
145
+ "visual.blocks.11.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
146
+ "visual.blocks.11.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
147
+ "visual.blocks.11.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
148
+ "visual.blocks.11.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
149
+ "visual.blocks.11.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
150
+ "visual.blocks.11.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
151
+ "visual.blocks.12.norm1.weight": "model-00001-of-00004.safetensors",
152
+ "visual.blocks.12.norm2.weight": "model-00001-of-00004.safetensors",
153
+ "visual.blocks.12.attn.qkv.weight": "model-00001-of-00004.safetensors",
154
+ "visual.blocks.12.attn.qkv.bias": "model-00001-of-00004.safetensors",
155
+ "visual.blocks.12.attn.proj.weight": "model-00001-of-00004.safetensors",
156
+ "visual.blocks.12.attn.proj.bias": "model-00001-of-00004.safetensors",
157
+ "visual.blocks.12.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
158
+ "visual.blocks.12.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
159
+ "visual.blocks.12.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
160
+ "visual.blocks.12.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
161
+ "visual.blocks.12.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
162
+ "visual.blocks.12.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
163
+ "visual.blocks.13.norm1.weight": "model-00001-of-00004.safetensors",
164
+ "visual.blocks.13.norm2.weight": "model-00001-of-00004.safetensors",
165
+ "visual.blocks.13.attn.qkv.weight": "model-00001-of-00004.safetensors",
166
+ "visual.blocks.13.attn.qkv.bias": "model-00001-of-00004.safetensors",
167
+ "visual.blocks.13.attn.proj.weight": "model-00001-of-00004.safetensors",
168
+ "visual.blocks.13.attn.proj.bias": "model-00001-of-00004.safetensors",
169
+ "visual.blocks.13.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
170
+ "visual.blocks.13.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
171
+ "visual.blocks.13.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
172
+ "visual.blocks.13.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
173
+ "visual.blocks.13.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
174
+ "visual.blocks.13.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
175
+ "visual.blocks.14.norm1.weight": "model-00001-of-00004.safetensors",
176
+ "visual.blocks.14.norm2.weight": "model-00001-of-00004.safetensors",
177
+ "visual.blocks.14.attn.qkv.weight": "model-00001-of-00004.safetensors",
178
+ "visual.blocks.14.attn.qkv.bias": "model-00001-of-00004.safetensors",
179
+ "visual.blocks.14.attn.proj.weight": "model-00001-of-00004.safetensors",
180
+ "visual.blocks.14.attn.proj.bias": "model-00001-of-00004.safetensors",
181
+ "visual.blocks.14.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
182
+ "visual.blocks.14.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
183
+ "visual.blocks.14.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
184
+ "visual.blocks.14.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
185
+ "visual.blocks.14.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
186
+ "visual.blocks.14.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
187
+ "visual.blocks.15.norm1.weight": "model-00001-of-00004.safetensors",
188
+ "visual.blocks.15.norm2.weight": "model-00001-of-00004.safetensors",
189
+ "visual.blocks.15.attn.qkv.weight": "model-00001-of-00004.safetensors",
190
+ "visual.blocks.15.attn.qkv.bias": "model-00001-of-00004.safetensors",
191
+ "visual.blocks.15.attn.proj.weight": "model-00001-of-00004.safetensors",
192
+ "visual.blocks.15.attn.proj.bias": "model-00001-of-00004.safetensors",
193
+ "visual.blocks.15.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
194
+ "visual.blocks.15.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
195
+ "visual.blocks.15.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
196
+ "visual.blocks.15.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
197
+ "visual.blocks.15.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
198
+ "visual.blocks.15.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
199
+ "visual.blocks.16.norm1.weight": "model-00001-of-00004.safetensors",
200
+ "visual.blocks.16.norm2.weight": "model-00001-of-00004.safetensors",
201
+ "visual.blocks.16.attn.qkv.weight": "model-00001-of-00004.safetensors",
202
+ "visual.blocks.16.attn.qkv.bias": "model-00001-of-00004.safetensors",
203
+ "visual.blocks.16.attn.proj.weight": "model-00001-of-00004.safetensors",
204
+ "visual.blocks.16.attn.proj.bias": "model-00001-of-00004.safetensors",
205
+ "visual.blocks.16.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
206
+ "visual.blocks.16.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
207
+ "visual.blocks.16.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
208
+ "visual.blocks.16.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
209
+ "visual.blocks.16.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
210
+ "visual.blocks.16.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
211
+ "visual.blocks.17.norm1.weight": "model-00001-of-00004.safetensors",
212
+ "visual.blocks.17.norm2.weight": "model-00001-of-00004.safetensors",
213
+ "visual.blocks.17.attn.qkv.weight": "model-00001-of-00004.safetensors",
214
+ "visual.blocks.17.attn.qkv.bias": "model-00001-of-00004.safetensors",
215
+ "visual.blocks.17.attn.proj.weight": "model-00001-of-00004.safetensors",
216
+ "visual.blocks.17.attn.proj.bias": "model-00001-of-00004.safetensors",
217
+ "visual.blocks.17.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
218
+ "visual.blocks.17.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
219
+ "visual.blocks.17.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
220
+ "visual.blocks.17.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
221
+ "visual.blocks.17.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
222
+ "visual.blocks.17.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
223
+ "visual.blocks.18.norm1.weight": "model-00001-of-00004.safetensors",
224
+ "visual.blocks.18.norm2.weight": "model-00001-of-00004.safetensors",
225
+ "visual.blocks.18.attn.qkv.weight": "model-00001-of-00004.safetensors",
226
+ "visual.blocks.18.attn.qkv.bias": "model-00001-of-00004.safetensors",
227
+ "visual.blocks.18.attn.proj.weight": "model-00001-of-00004.safetensors",
228
+ "visual.blocks.18.attn.proj.bias": "model-00001-of-00004.safetensors",
229
+ "visual.blocks.18.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
230
+ "visual.blocks.18.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
231
+ "visual.blocks.18.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
232
+ "visual.blocks.18.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
233
+ "visual.blocks.18.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
234
+ "visual.blocks.18.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
235
+ "visual.blocks.19.norm1.weight": "model-00001-of-00004.safetensors",
236
+ "visual.blocks.19.norm2.weight": "model-00001-of-00004.safetensors",
237
+ "visual.blocks.19.attn.qkv.weight": "model-00001-of-00004.safetensors",
238
+ "visual.blocks.19.attn.qkv.bias": "model-00001-of-00004.safetensors",
239
+ "visual.blocks.19.attn.proj.weight": "model-00001-of-00004.safetensors",
240
+ "visual.blocks.19.attn.proj.bias": "model-00001-of-00004.safetensors",
241
+ "visual.blocks.19.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
242
+ "visual.blocks.19.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
243
+ "visual.blocks.19.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
244
+ "visual.blocks.19.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
245
+ "visual.blocks.19.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
246
+ "visual.blocks.19.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
247
+ "visual.blocks.20.norm1.weight": "model-00001-of-00004.safetensors",
248
+ "visual.blocks.20.norm2.weight": "model-00001-of-00004.safetensors",
249
+ "visual.blocks.20.attn.qkv.weight": "model-00001-of-00004.safetensors",
250
+ "visual.blocks.20.attn.qkv.bias": "model-00001-of-00004.safetensors",
251
+ "visual.blocks.20.attn.proj.weight": "model-00001-of-00004.safetensors",
252
+ "visual.blocks.20.attn.proj.bias": "model-00001-of-00004.safetensors",
253
+ "visual.blocks.20.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
254
+ "visual.blocks.20.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
255
+ "visual.blocks.20.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
256
+ "visual.blocks.20.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
257
+ "visual.blocks.20.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
258
+ "visual.blocks.20.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
259
+ "visual.blocks.21.norm1.weight": "model-00001-of-00004.safetensors",
260
+ "visual.blocks.21.norm2.weight": "model-00001-of-00004.safetensors",
261
+ "visual.blocks.21.attn.qkv.weight": "model-00001-of-00004.safetensors",
262
+ "visual.blocks.21.attn.qkv.bias": "model-00001-of-00004.safetensors",
263
+ "visual.blocks.21.attn.proj.weight": "model-00001-of-00004.safetensors",
264
+ "visual.blocks.21.attn.proj.bias": "model-00001-of-00004.safetensors",
265
+ "visual.blocks.21.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
266
+ "visual.blocks.21.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
267
+ "visual.blocks.21.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
268
+ "visual.blocks.21.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
269
+ "visual.blocks.21.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
270
+ "visual.blocks.21.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
271
+ "visual.blocks.22.norm1.weight": "model-00001-of-00004.safetensors",
272
+ "visual.blocks.22.norm2.weight": "model-00001-of-00004.safetensors",
273
+ "visual.blocks.22.attn.qkv.weight": "model-00001-of-00004.safetensors",
274
+ "visual.blocks.22.attn.qkv.bias": "model-00001-of-00004.safetensors",
275
+ "visual.blocks.22.attn.proj.weight": "model-00001-of-00004.safetensors",
276
+ "visual.blocks.22.attn.proj.bias": "model-00001-of-00004.safetensors",
277
+ "visual.blocks.22.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
278
+ "visual.blocks.22.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
279
+ "visual.blocks.22.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
280
+ "visual.blocks.22.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
281
+ "visual.blocks.22.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
282
+ "visual.blocks.22.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
283
+ "visual.blocks.23.norm1.weight": "model-00001-of-00004.safetensors",
284
+ "visual.blocks.23.norm2.weight": "model-00001-of-00004.safetensors",
285
+ "visual.blocks.23.attn.qkv.weight": "model-00001-of-00004.safetensors",
286
+ "visual.blocks.23.attn.qkv.bias": "model-00001-of-00004.safetensors",
287
+ "visual.blocks.23.attn.proj.weight": "model-00001-of-00004.safetensors",
288
+ "visual.blocks.23.attn.proj.bias": "model-00001-of-00004.safetensors",
289
+ "visual.blocks.23.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
290
+ "visual.blocks.23.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
291
+ "visual.blocks.23.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
292
+ "visual.blocks.23.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
293
+ "visual.blocks.23.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
294
+ "visual.blocks.23.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
295
+ "visual.blocks.24.norm1.weight": "model-00001-of-00004.safetensors",
296
+ "visual.blocks.24.norm2.weight": "model-00001-of-00004.safetensors",
297
+ "visual.blocks.24.attn.qkv.weight": "model-00001-of-00004.safetensors",
298
+ "visual.blocks.24.attn.qkv.bias": "model-00001-of-00004.safetensors",
299
+ "visual.blocks.24.attn.proj.weight": "model-00001-of-00004.safetensors",
300
+ "visual.blocks.24.attn.proj.bias": "model-00001-of-00004.safetensors",
301
+ "visual.blocks.24.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
302
+ "visual.blocks.24.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
303
+ "visual.blocks.24.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
304
+ "visual.blocks.24.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
305
+ "visual.blocks.24.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
306
+ "visual.blocks.24.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
307
+ "visual.blocks.25.norm1.weight": "model-00001-of-00004.safetensors",
308
+ "visual.blocks.25.norm2.weight": "model-00001-of-00004.safetensors",
309
+ "visual.blocks.25.attn.qkv.weight": "model-00001-of-00004.safetensors",
310
+ "visual.blocks.25.attn.qkv.bias": "model-00001-of-00004.safetensors",
311
+ "visual.blocks.25.attn.proj.weight": "model-00001-of-00004.safetensors",
312
+ "visual.blocks.25.attn.proj.bias": "model-00001-of-00004.safetensors",
313
+ "visual.blocks.25.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
314
+ "visual.blocks.25.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
315
+ "visual.blocks.25.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
316
+ "visual.blocks.25.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
317
+ "visual.blocks.25.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
318
+ "visual.blocks.25.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
319
+ "visual.blocks.26.norm1.weight": "model-00001-of-00004.safetensors",
320
+ "visual.blocks.26.norm2.weight": "model-00001-of-00004.safetensors",
321
+ "visual.blocks.26.attn.qkv.weight": "model-00001-of-00004.safetensors",
322
+ "visual.blocks.26.attn.qkv.bias": "model-00001-of-00004.safetensors",
323
+ "visual.blocks.26.attn.proj.weight": "model-00001-of-00004.safetensors",
324
+ "visual.blocks.26.attn.proj.bias": "model-00001-of-00004.safetensors",
325
+ "visual.blocks.26.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
326
+ "visual.blocks.26.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
327
+ "visual.blocks.26.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
328
+ "visual.blocks.26.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
329
+ "visual.blocks.26.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
330
+ "visual.blocks.26.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
331
+ "visual.blocks.27.norm1.weight": "model-00001-of-00004.safetensors",
332
+ "visual.blocks.27.norm2.weight": "model-00001-of-00004.safetensors",
333
+ "visual.blocks.27.attn.qkv.weight": "model-00001-of-00004.safetensors",
334
+ "visual.blocks.27.attn.qkv.bias": "model-00001-of-00004.safetensors",
335
+ "visual.blocks.27.attn.proj.weight": "model-00001-of-00004.safetensors",
336
+ "visual.blocks.27.attn.proj.bias": "model-00001-of-00004.safetensors",
337
+ "visual.blocks.27.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
338
+ "visual.blocks.27.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
339
+ "visual.blocks.27.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
340
+ "visual.blocks.27.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
341
+ "visual.blocks.27.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
342
+ "visual.blocks.27.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
343
+ "visual.blocks.28.norm1.weight": "model-00001-of-00004.safetensors",
344
+ "visual.blocks.28.norm2.weight": "model-00001-of-00004.safetensors",
345
+ "visual.blocks.28.attn.qkv.weight": "model-00001-of-00004.safetensors",
346
+ "visual.blocks.28.attn.qkv.bias": "model-00001-of-00004.safetensors",
347
+ "visual.blocks.28.attn.proj.weight": "model-00001-of-00004.safetensors",
348
+ "visual.blocks.28.attn.proj.bias": "model-00001-of-00004.safetensors",
349
+ "visual.blocks.28.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
350
+ "visual.blocks.28.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
351
+ "visual.blocks.28.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
352
+ "visual.blocks.28.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
353
+ "visual.blocks.28.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
354
+ "visual.blocks.28.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
355
+ "visual.blocks.29.norm1.weight": "model-00001-of-00004.safetensors",
356
+ "visual.blocks.29.norm2.weight": "model-00001-of-00004.safetensors",
357
+ "visual.blocks.29.attn.qkv.weight": "model-00001-of-00004.safetensors",
358
+ "visual.blocks.29.attn.qkv.bias": "model-00001-of-00004.safetensors",
359
+ "visual.blocks.29.attn.proj.weight": "model-00001-of-00004.safetensors",
360
+ "visual.blocks.29.attn.proj.bias": "model-00001-of-00004.safetensors",
361
+ "visual.blocks.29.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
362
+ "visual.blocks.29.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
363
+ "visual.blocks.29.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
364
+ "visual.blocks.29.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
365
+ "visual.blocks.29.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
366
+ "visual.blocks.29.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
367
+ "visual.blocks.30.norm1.weight": "model-00001-of-00004.safetensors",
368
+ "visual.blocks.30.norm2.weight": "model-00001-of-00004.safetensors",
369
+ "visual.blocks.30.attn.qkv.weight": "model-00001-of-00004.safetensors",
370
+ "visual.blocks.30.attn.qkv.bias": "model-00001-of-00004.safetensors",
371
+ "visual.blocks.30.attn.proj.weight": "model-00001-of-00004.safetensors",
372
+ "visual.blocks.30.attn.proj.bias": "model-00001-of-00004.safetensors",
373
+ "visual.blocks.30.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
374
+ "visual.blocks.30.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
375
+ "visual.blocks.30.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
376
+ "visual.blocks.30.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
377
+ "visual.blocks.30.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
378
+ "visual.blocks.30.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
379
+ "visual.blocks.31.norm1.weight": "model-00001-of-00004.safetensors",
380
+ "visual.blocks.31.norm2.weight": "model-00001-of-00004.safetensors",
381
+ "visual.blocks.31.attn.qkv.weight": "model-00001-of-00004.safetensors",
382
+ "visual.blocks.31.attn.qkv.bias": "model-00001-of-00004.safetensors",
383
+ "visual.blocks.31.attn.proj.weight": "model-00001-of-00004.safetensors",
384
+ "visual.blocks.31.attn.proj.bias": "model-00001-of-00004.safetensors",
385
+ "visual.blocks.31.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
386
+ "visual.blocks.31.mlp.gate_proj.bias": "model-00001-of-00004.safetensors",
387
+ "visual.blocks.31.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
388
+ "visual.blocks.31.mlp.up_proj.bias": "model-00001-of-00004.safetensors",
389
+ "visual.blocks.31.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
390
+ "visual.blocks.31.mlp.down_proj.bias": "model-00001-of-00004.safetensors",
391
+ "visual.merger.ln_q.weight": "model-00001-of-00004.safetensors",
392
+ "visual.merger.mlp.0.weight": "model-00001-of-00004.safetensors",
393
+ "visual.merger.mlp.0.bias": "model-00001-of-00004.safetensors",
394
+ "visual.merger.mlp.2.weight": "model-00001-of-00004.safetensors",
395
+ "visual.merger.mlp.2.bias": "model-00001-of-00004.safetensors",
396
+ "model.embed_tokens.weight": "model-00001-of-00004.safetensors",
397
+ "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
398
+ "model.layers.0.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
399
+ "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
400
+ "model.layers.0.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
401
+ "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
402
+ "model.layers.0.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
403
+ "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
404
+ "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
405
+ "model.layers.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
406
+ "model.layers.0.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
407
+ "model.layers.0.input_layernorm.weight": "model-00001-of-00004.safetensors",
408
+ "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
409
+ "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
410
+ "model.layers.1.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
411
+ "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
412
+ "model.layers.1.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
413
+ "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
414
+ "model.layers.1.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
415
+ "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
416
+ "model.layers.1.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
417
+ "model.layers.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
418
+ "model.layers.1.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
419
+ "model.layers.1.input_layernorm.weight": "model-00001-of-00004.safetensors",
420
+ "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
421
+ "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
422
+ "model.layers.2.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
423
+ "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
424
+ "model.layers.2.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
425
+ "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
426
+ "model.layers.2.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
427
+ "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
428
+ "model.layers.2.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
429
+ "model.layers.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
430
+ "model.layers.2.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
431
+ "model.layers.2.input_layernorm.weight": "model-00001-of-00004.safetensors",
432
+ "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
433
+ "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
434
+ "model.layers.3.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
435
+ "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
436
+ "model.layers.3.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
437
+ "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
438
+ "model.layers.3.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
439
+ "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
440
+ "model.layers.3.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
441
+ "model.layers.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
442
+ "model.layers.3.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
443
+ "model.layers.3.input_layernorm.weight": "model-00001-of-00004.safetensors",
444
+ "model.layers.3.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
445
+ "model.layers.4.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
446
+ "model.layers.4.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
447
+ "model.layers.4.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
448
+ "model.layers.4.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
449
+ "model.layers.4.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
450
+ "model.layers.4.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
451
+ "model.layers.4.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
452
+ "model.layers.4.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
453
+ "model.layers.4.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
454
+ "model.layers.4.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
455
+ "model.layers.4.input_layernorm.weight": "model-00001-of-00004.safetensors",
456
+ "model.layers.4.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
457
+ "model.layers.5.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
458
+ "model.layers.5.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
459
+ "model.layers.5.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
460
+ "model.layers.5.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
461
+ "model.layers.5.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
462
+ "model.layers.5.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
463
+ "model.layers.5.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
464
+ "model.layers.5.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
465
+ "model.layers.5.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
466
+ "model.layers.5.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
467
+ "model.layers.5.input_layernorm.weight": "model-00002-of-00004.safetensors",
468
+ "model.layers.5.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
469
+ "model.layers.6.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
470
+ "model.layers.6.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
471
+ "model.layers.6.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
472
+ "model.layers.6.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
473
+ "model.layers.6.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
474
+ "model.layers.6.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
475
+ "model.layers.6.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
476
+ "model.layers.6.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
477
+ "model.layers.6.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
478
+ "model.layers.6.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
479
+ "model.layers.6.input_layernorm.weight": "model-00002-of-00004.safetensors",
480
+ "model.layers.6.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
481
+ "model.layers.7.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
482
+ "model.layers.7.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
483
+ "model.layers.7.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
484
+ "model.layers.7.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
485
+ "model.layers.7.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
486
+ "model.layers.7.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
487
+ "model.layers.7.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
488
+ "model.layers.7.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
489
+ "model.layers.7.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
490
+ "model.layers.7.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
491
+ "model.layers.7.input_layernorm.weight": "model-00002-of-00004.safetensors",
492
+ "model.layers.7.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
493
+ "model.layers.8.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
494
+ "model.layers.8.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
495
+ "model.layers.8.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
496
+ "model.layers.8.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
497
+ "model.layers.8.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
498
+ "model.layers.8.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
499
+ "model.layers.8.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
500
+ "model.layers.8.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
501
+ "model.layers.8.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
502
+ "model.layers.8.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
503
+ "model.layers.8.input_layernorm.weight": "model-00002-of-00004.safetensors",
504
+ "model.layers.8.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
505
+ "model.layers.9.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
506
+ "model.layers.9.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
507
+ "model.layers.9.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
508
+ "model.layers.9.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
509
+ "model.layers.9.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
510
+ "model.layers.9.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
511
+ "model.layers.9.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
512
+ "model.layers.9.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
513
+ "model.layers.9.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
514
+ "model.layers.9.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
515
+ "model.layers.9.input_layernorm.weight": "model-00002-of-00004.safetensors",
516
+ "model.layers.9.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
517
+ "model.layers.10.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
518
+ "model.layers.10.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
519
+ "model.layers.10.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
520
+ "model.layers.10.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
521
+ "model.layers.10.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
522
+ "model.layers.10.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
523
+ "model.layers.10.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
524
+ "model.layers.10.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
525
+ "model.layers.10.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
526
+ "model.layers.10.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
527
+ "model.layers.10.input_layernorm.weight": "model-00002-of-00004.safetensors",
528
+ "model.layers.10.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
529
+ "model.layers.11.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
530
+ "model.layers.11.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
531
+ "model.layers.11.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
532
+ "model.layers.11.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
533
+ "model.layers.11.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
534
+ "model.layers.11.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
535
+ "model.layers.11.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
536
+ "model.layers.11.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
537
+ "model.layers.11.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
538
+ "model.layers.11.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
539
+ "model.layers.11.input_layernorm.weight": "model-00002-of-00004.safetensors",
540
+ "model.layers.11.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
541
+ "model.layers.12.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
542
+ "model.layers.12.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
543
+ "model.layers.12.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
544
+ "model.layers.12.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
545
+ "model.layers.12.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
546
+ "model.layers.12.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
547
+ "model.layers.12.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
548
+ "model.layers.12.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
549
+ "model.layers.12.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
550
+ "model.layers.12.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
551
+ "model.layers.12.input_layernorm.weight": "model-00002-of-00004.safetensors",
552
+ "model.layers.12.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
553
+ "model.layers.13.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
554
+ "model.layers.13.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
555
+ "model.layers.13.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
556
+ "model.layers.13.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
557
+ "model.layers.13.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
558
+ "model.layers.13.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
559
+ "model.layers.13.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
560
+ "model.layers.13.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
561
+ "model.layers.13.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
562
+ "model.layers.13.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
563
+ "model.layers.13.input_layernorm.weight": "model-00002-of-00004.safetensors",
564
+ "model.layers.13.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
565
+ "model.layers.14.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
566
+ "model.layers.14.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
567
+ "model.layers.14.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
568
+ "model.layers.14.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
569
+ "model.layers.14.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
570
+ "model.layers.14.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
571
+ "model.layers.14.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
572
+ "model.layers.14.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
573
+ "model.layers.14.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
574
+ "model.layers.14.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
575
+ "model.layers.14.input_layernorm.weight": "model-00002-of-00004.safetensors",
576
+ "model.layers.14.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
577
+ "model.layers.15.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
578
+ "model.layers.15.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
579
+ "model.layers.15.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
580
+ "model.layers.15.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
581
+ "model.layers.15.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
582
+ "model.layers.15.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
583
+ "model.layers.15.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
584
+ "model.layers.15.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
585
+ "model.layers.15.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
586
+ "model.layers.15.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
587
+ "model.layers.15.input_layernorm.weight": "model-00002-of-00004.safetensors",
588
+ "model.layers.15.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
589
+ "model.layers.16.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
590
+ "model.layers.16.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
591
+ "model.layers.16.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
592
+ "model.layers.16.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
593
+ "model.layers.16.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
594
+ "model.layers.16.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
595
+ "model.layers.16.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
596
+ "model.layers.16.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
597
+ "model.layers.16.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
598
+ "model.layers.16.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
599
+ "model.layers.16.input_layernorm.weight": "model-00003-of-00004.safetensors",
600
+ "model.layers.16.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
601
+ "model.layers.17.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
602
+ "model.layers.17.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
603
+ "model.layers.17.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
604
+ "model.layers.17.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
605
+ "model.layers.17.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
606
+ "model.layers.17.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
607
+ "model.layers.17.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
608
+ "model.layers.17.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
609
+ "model.layers.17.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
610
+ "model.layers.17.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
611
+ "model.layers.17.input_layernorm.weight": "model-00003-of-00004.safetensors",
612
+ "model.layers.17.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
613
+ "model.layers.18.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
614
+ "model.layers.18.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
615
+ "model.layers.18.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
616
+ "model.layers.18.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
617
+ "model.layers.18.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
618
+ "model.layers.18.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
619
+ "model.layers.18.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
620
+ "model.layers.18.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
621
+ "model.layers.18.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
622
+ "model.layers.18.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
623
+ "model.layers.18.input_layernorm.weight": "model-00003-of-00004.safetensors",
624
+ "model.layers.18.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
625
+ "model.layers.19.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
626
+ "model.layers.19.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
627
+ "model.layers.19.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
628
+ "model.layers.19.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
629
+ "model.layers.19.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
630
+ "model.layers.19.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
631
+ "model.layers.19.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
632
+ "model.layers.19.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
633
+ "model.layers.19.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
634
+ "model.layers.19.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
635
+ "model.layers.19.input_layernorm.weight": "model-00003-of-00004.safetensors",
636
+ "model.layers.19.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
637
+ "model.layers.20.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
638
+ "model.layers.20.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
639
+ "model.layers.20.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
640
+ "model.layers.20.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
641
+ "model.layers.20.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
642
+ "model.layers.20.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
643
+ "model.layers.20.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
644
+ "model.layers.20.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
645
+ "model.layers.20.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
646
+ "model.layers.20.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
647
+ "model.layers.20.input_layernorm.weight": "model-00003-of-00004.safetensors",
648
+ "model.layers.20.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
649
+ "model.layers.21.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
650
+ "model.layers.21.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
651
+ "model.layers.21.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
652
+ "model.layers.21.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
653
+ "model.layers.21.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
654
+ "model.layers.21.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
655
+ "model.layers.21.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
656
+ "model.layers.21.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
657
+ "model.layers.21.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
658
+ "model.layers.21.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
659
+ "model.layers.21.input_layernorm.weight": "model-00003-of-00004.safetensors",
660
+ "model.layers.21.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
661
+ "model.layers.22.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
662
+ "model.layers.22.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
663
+ "model.layers.22.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
664
+ "model.layers.22.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
665
+ "model.layers.22.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
666
+ "model.layers.22.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
667
+ "model.layers.22.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
668
+ "model.layers.22.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
669
+ "model.layers.22.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
670
+ "model.layers.22.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
671
+ "model.layers.22.input_layernorm.weight": "model-00003-of-00004.safetensors",
672
+ "model.layers.22.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
673
+ "model.layers.23.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
674
+ "model.layers.23.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
675
+ "model.layers.23.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
676
+ "model.layers.23.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
677
+ "model.layers.23.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
678
+ "model.layers.23.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
679
+ "model.layers.23.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
680
+ "model.layers.23.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
681
+ "model.layers.23.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
682
+ "model.layers.23.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
683
+ "model.layers.23.input_layernorm.weight": "model-00003-of-00004.safetensors",
684
+ "model.layers.23.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
685
+ "model.layers.24.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
686
+ "model.layers.24.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
687
+ "model.layers.24.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
688
+ "model.layers.24.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
689
+ "model.layers.24.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
690
+ "model.layers.24.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
691
+ "model.layers.24.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
692
+ "model.layers.24.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
693
+ "model.layers.24.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
694
+ "model.layers.24.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
695
+ "model.layers.24.input_layernorm.weight": "model-00003-of-00004.safetensors",
696
+ "model.layers.24.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
697
+ "model.layers.25.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
698
+ "model.layers.25.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
699
+ "model.layers.25.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
700
+ "model.layers.25.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
701
+ "model.layers.25.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
702
+ "model.layers.25.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
703
+ "model.layers.25.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
704
+ "model.layers.25.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
705
+ "model.layers.25.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
706
+ "model.layers.25.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
707
+ "model.layers.25.input_layernorm.weight": "model-00003-of-00004.safetensors",
708
+ "model.layers.25.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
709
+ "model.layers.26.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
710
+ "model.layers.26.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
711
+ "model.layers.26.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
712
+ "model.layers.26.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
713
+ "model.layers.26.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
714
+ "model.layers.26.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
715
+ "model.layers.26.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
716
+ "model.layers.26.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
717
+ "model.layers.26.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
718
+ "model.layers.26.mlp.down_proj.weight": "model-00004-of-00004.safetensors",
719
+ "model.layers.26.input_layernorm.weight": "model-00004-of-00004.safetensors",
720
+ "model.layers.26.post_attention_layernorm.weight": "model-00004-of-00004.safetensors",
721
+ "model.layers.27.self_attn.q_proj.weight": "model-00004-of-00004.safetensors",
722
+ "model.layers.27.self_attn.q_proj.bias": "model-00004-of-00004.safetensors",
723
+ "model.layers.27.self_attn.k_proj.weight": "model-00004-of-00004.safetensors",
724
+ "model.layers.27.self_attn.k_proj.bias": "model-00004-of-00004.safetensors",
725
+ "model.layers.27.self_attn.v_proj.weight": "model-00004-of-00004.safetensors",
726
+ "model.layers.27.self_attn.v_proj.bias": "model-00004-of-00004.safetensors",
727
+ "model.layers.27.self_attn.o_proj.weight": "model-00004-of-00004.safetensors",
728
+ "model.layers.27.mlp.gate_proj.weight": "model-00004-of-00004.safetensors",
729
+ "model.layers.27.mlp.up_proj.weight": "model-00004-of-00004.safetensors",
730
+ "model.layers.27.mlp.down_proj.weight": "model-00004-of-00004.safetensors",
731
+ "model.layers.27.input_layernorm.weight": "model-00004-of-00004.safetensors",
732
+ "model.layers.27.post_attention_layernorm.weight": "model-00004-of-00004.safetensors",
733
+ "model.norm.weight": "model-00004-of-00004.safetensors",
734
+ "lm_head.weight": "model-00004-of-00004.safetensors"
735
+ }
736
+ }
text_encoder/preprocessor_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "min_pixels": 3136,
3
+ "max_pixels": 12845056,
4
+ "patch_size": 14,
5
+ "temporal_patch_size": 2,
6
+ "merge_size": 2,
7
+ "image_mean": [
8
+ 0.48145466,
9
+ 0.4578275,
10
+ 0.40821073
11
+ ],
12
+ "image_std": [
13
+ 0.26862954,
14
+ 0.26130258,
15
+ 0.27577711
16
+ ],
17
+ "image_processor_type": "Qwen2VLImageProcessor",
18
+ "processor_class": "Qwen2_5_VLProcessor"
19
+ }
text_encoder/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
text_encoder/tokenizer_config.json ADDED
@@ -0,0 +1,207 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "added_tokens_decoder": {
4
+ "151643": {
5
+ "content": "<|endoftext|>",
6
+ "lstrip": false,
7
+ "normalized": false,
8
+ "rstrip": false,
9
+ "single_word": false,
10
+ "special": true
11
+ },
12
+ "151644": {
13
+ "content": "<|im_start|>",
14
+ "lstrip": false,
15
+ "normalized": false,
16
+ "rstrip": false,
17
+ "single_word": false,
18
+ "special": true
19
+ },
20
+ "151645": {
21
+ "content": "<|im_end|>",
22
+ "lstrip": false,
23
+ "normalized": false,
24
+ "rstrip": false,
25
+ "single_word": false,
26
+ "special": true
27
+ },
28
+ "151646": {
29
+ "content": "<|object_ref_start|>",
30
+ "lstrip": false,
31
+ "normalized": false,
32
+ "rstrip": false,
33
+ "single_word": false,
34
+ "special": true
35
+ },
36
+ "151647": {
37
+ "content": "<|object_ref_end|>",
38
+ "lstrip": false,
39
+ "normalized": false,
40
+ "rstrip": false,
41
+ "single_word": false,
42
+ "special": true
43
+ },
44
+ "151648": {
45
+ "content": "<|box_start|>",
46
+ "lstrip": false,
47
+ "normalized": false,
48
+ "rstrip": false,
49
+ "single_word": false,
50
+ "special": true
51
+ },
52
+ "151649": {
53
+ "content": "<|box_end|>",
54
+ "lstrip": false,
55
+ "normalized": false,
56
+ "rstrip": false,
57
+ "single_word": false,
58
+ "special": true
59
+ },
60
+ "151650": {
61
+ "content": "<|quad_start|>",
62
+ "lstrip": false,
63
+ "normalized": false,
64
+ "rstrip": false,
65
+ "single_word": false,
66
+ "special": true
67
+ },
68
+ "151651": {
69
+ "content": "<|quad_end|>",
70
+ "lstrip": false,
71
+ "normalized": false,
72
+ "rstrip": false,
73
+ "single_word": false,
74
+ "special": true
75
+ },
76
+ "151652": {
77
+ "content": "<|vision_start|>",
78
+ "lstrip": false,
79
+ "normalized": false,
80
+ "rstrip": false,
81
+ "single_word": false,
82
+ "special": true
83
+ },
84
+ "151653": {
85
+ "content": "<|vision_end|>",
86
+ "lstrip": false,
87
+ "normalized": false,
88
+ "rstrip": false,
89
+ "single_word": false,
90
+ "special": true
91
+ },
92
+ "151654": {
93
+ "content": "<|vision_pad|>",
94
+ "lstrip": false,
95
+ "normalized": false,
96
+ "rstrip": false,
97
+ "single_word": false,
98
+ "special": true
99
+ },
100
+ "151655": {
101
+ "content": "<|image_pad|>",
102
+ "lstrip": false,
103
+ "normalized": false,
104
+ "rstrip": false,
105
+ "single_word": false,
106
+ "special": true
107
+ },
108
+ "151656": {
109
+ "content": "<|video_pad|>",
110
+ "lstrip": false,
111
+ "normalized": false,
112
+ "rstrip": false,
113
+ "single_word": false,
114
+ "special": true
115
+ },
116
+ "151657": {
117
+ "content": "<tool_call>",
118
+ "lstrip": false,
119
+ "normalized": false,
120
+ "rstrip": false,
121
+ "single_word": false,
122
+ "special": false
123
+ },
124
+ "151658": {
125
+ "content": "</tool_call>",
126
+ "lstrip": false,
127
+ "normalized": false,
128
+ "rstrip": false,
129
+ "single_word": false,
130
+ "special": false
131
+ },
132
+ "151659": {
133
+ "content": "<|fim_prefix|>",
134
+ "lstrip": false,
135
+ "normalized": false,
136
+ "rstrip": false,
137
+ "single_word": false,
138
+ "special": false
139
+ },
140
+ "151660": {
141
+ "content": "<|fim_middle|>",
142
+ "lstrip": false,
143
+ "normalized": false,
144
+ "rstrip": false,
145
+ "single_word": false,
146
+ "special": false
147
+ },
148
+ "151661": {
149
+ "content": "<|fim_suffix|>",
150
+ "lstrip": false,
151
+ "normalized": false,
152
+ "rstrip": false,
153
+ "single_word": false,
154
+ "special": false
155
+ },
156
+ "151662": {
157
+ "content": "<|fim_pad|>",
158
+ "lstrip": false,
159
+ "normalized": false,
160
+ "rstrip": false,
161
+ "single_word": false,
162
+ "special": false
163
+ },
164
+ "151663": {
165
+ "content": "<|repo_name|>",
166
+ "lstrip": false,
167
+ "normalized": false,
168
+ "rstrip": false,
169
+ "single_word": false,
170
+ "special": false
171
+ },
172
+ "151664": {
173
+ "content": "<|file_sep|>",
174
+ "lstrip": false,
175
+ "normalized": false,
176
+ "rstrip": false,
177
+ "single_word": false,
178
+ "special": false
179
+ }
180
+ },
181
+ "additional_special_tokens": [
182
+ "<|im_start|>",
183
+ "<|im_end|>",
184
+ "<|object_ref_start|>",
185
+ "<|object_ref_end|>",
186
+ "<|box_start|>",
187
+ "<|box_end|>",
188
+ "<|quad_start|>",
189
+ "<|quad_end|>",
190
+ "<|vision_start|>",
191
+ "<|vision_end|>",
192
+ "<|vision_pad|>",
193
+ "<|image_pad|>",
194
+ "<|video_pad|>"
195
+ ],
196
+ "bos_token": null,
197
+ "chat_template": "{% set image_count = namespace(value=0) %}{% set video_count = namespace(value=0) %}{% for message in messages %}{% if loop.first and message['role'] != 'system' %}<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n{% endif %}<|im_start|>{{ message['role'] }}\n{% if message['content'] is string %}{{ message['content'] }}<|im_end|>\n{% else %}{% for content in message['content'] %}{% if content['type'] == 'image' or 'image' in content or 'image_url' in content %}{% set image_count.value = image_count.value + 1 %}{% if add_vision_id %}Picture {{ image_count.value }}: {% endif %}<|vision_start|><|image_pad|><|vision_end|>{% elif content['type'] == 'video' or 'video' in content %}{% set video_count.value = video_count.value + 1 %}{% if add_vision_id %}Video {{ video_count.value }}: {% endif %}<|vision_start|><|video_pad|><|vision_end|>{% elif 'text' in content %}{{ content['text'] }}{% endif %}{% endfor %}<|im_end|>\n{% endif %}{% endfor %}{% if add_generation_prompt %}<|im_start|>assistant\n{% endif %}",
198
+ "clean_up_tokenization_spaces": false,
199
+ "eos_token": "<|im_end|>",
200
+ "errors": "replace",
201
+ "model_max_length": 131072,
202
+ "pad_token": "<|endoftext|>",
203
+ "split_special_tokens": false,
204
+ "tokenizer_class": "Qwen2Tokenizer",
205
+ "unk_token": null,
206
+ "add_bos_token": false
207
+ }
tokenizer/config.json ADDED
@@ -0,0 +1,68 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_class_name": "AutoProcessor",
3
+ "_diffusers_version": "0.34.0",
4
+ "_name_or_path": "Qwen/Qwen2.5-VL-7B-Instruct",
5
+ "processor_type": "Qwen2_5_VLProcessor",
6
+ "auto_map": {
7
+ "AutoProcessor": "processing_qwen2_5_vl.Qwen2_5_VLProcessor"
8
+ },
9
+ "tokenizer_config": {
10
+ "model_type": "qwen2_5_vl",
11
+ "tokenizer_class": "Qwen2Tokenizer",
12
+ "bos_token": "<|endoftext|>",
13
+ "eos_token": "<|im_end|>",
14
+ "pad_token": "<|endoftext|>",
15
+ "unk_token": null,
16
+ "add_bos_token": false,
17
+ "add_eos_token": false,
18
+ "chat_template": "{% set image_count = namespace(value=0) %}{% set video_count = namespace(value=0) %}{% for message in messages %}{% if loop.first and message['role'] != 'system' %}<|im_start|>system\nYou are a helpful assistant.<|im_end|>\n{% endif %}<|im_start|>{{ message['role'] }}\n{% if message['content'] is string %}{{ message['content'] }}<|im_end|>\n{% else %}{% for content in message['content'] %}{% if content['type'] == 'image' or 'image' in content or 'image_url' in content %}{% set image_count.value = image_count.value + 1 %}{% if add_vision_id %}Picture {{ image_count.value }}: {% endif %}<|vision_start|><|image_pad|><|vision_end|>{% elif content['type'] == 'video' or 'video' in content %}{% set video_count.value = video_count.value + 1 %}{% if add_vision_id %}Video {{ video_count.value }}: {% endif %}<|vision_start|><|video_pad|><|vision_end|>{% elif 'text' in content %}{{ content['text'] }}{% endif %}{% endfor %}<|im_end|>\n{% endif %}{% endfor %}{% if add_generation_prompt %}<|im_start|>assistant\n{% endif %}"
19
+ },
20
+ "image_processor_config": {
21
+ "image_processor_type": "Qwen2_5_VLImageProcessor",
22
+ "do_resize": true,
23
+ "size": {
24
+ "min_pixels": 3136,
25
+ "max_pixels": 12845056
26
+ },
27
+ "do_rescale": true,
28
+ "rescale_factor": 0.00392156862745098,
29
+ "do_normalize": true,
30
+ "image_mean": [0.48145466, 0.4578275, 0.40821073],
31
+ "image_std": [0.26862954, 0.26130258, 0.27577711]
32
+ },
33
+ "video_processor_config": {
34
+ "video_processor_type": "Qwen2_5_VLVideoProcessor",
35
+ "do_resize": true,
36
+ "size": {
37
+ "min_pixels": 3136,
38
+ "max_pixels": 12845056
39
+ },
40
+ "fps": 2.0,
41
+ "max_num_frames": 768
42
+ }
43
+ }
44
+
45
+
46
+
47
+
48
+
49
+
50
+
51
+
52
+
53
+
54
+
55
+
56
+
57
+
58
+
59
+
60
+
61
+
62
+
63
+
64
+
65
+
66
+
67
+
68
+
transformer/config.json ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_class_name": "Cosmos25Transformer3DModel",
3
+ "_diffusers_version": "0.31.0",
4
+ "num_attention_heads": 16,
5
+ "attention_head_dim": 128,
6
+ "in_channels": 16,
7
+ "out_channels": 16,
8
+ "num_layers": 28,
9
+ "patch_size": [1, 2, 2],
10
+ "max_size": [128, 240, 240],
11
+ "rope_scale": [1.0, 3.0, 3.0],
12
+ "text_embed_dim": 1024,
13
+ "mlp_ratio": 4.0,
14
+ "adaln_lora_dim": 256,
15
+ "use_adaln_lora": true,
16
+ "concat_padding_mask": true,
17
+ "extra_pos_embed_type": null,
18
+ "use_crossattn_projection": true,
19
+ "rope_enable_fps_modulation": false,
20
+ "qk_norm": "rms_norm"
21
+ }
22
+
23
+
24
+
25
+
26
+
27
+
28
+
29
+
30
+
31
+
32
+
33
+
34
+
35
+
36
+
37
+
38
+
39
+
40
+
41
+
42
+
43
+
44
+
45
+
46
+
47
+
transformer/cosmos2_5_post_trained.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:46b81bd459415c5694bf68e0e980b7e92160a26724b9ab189cff78925556edde
3
+ size 4118430115
vae/config.json ADDED
@@ -0,0 +1,80 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_class_name": "AutoencoderKLWan",
3
+ "_diffusers_version": "0.34.0.dev0",
4
+ "attn_scales": [],
5
+ "base_dim": 96,
6
+ "dim_mult": [
7
+ 1,
8
+ 2,
9
+ 4,
10
+ 4
11
+ ],
12
+ "dropout": 0.0,
13
+ "latents_mean": [
14
+ -0.7571,
15
+ -0.7089,
16
+ -0.9113,
17
+ 0.1075,
18
+ -0.1745,
19
+ 0.9653,
20
+ -0.1517,
21
+ 1.5508,
22
+ 0.4134,
23
+ -0.0715,
24
+ 0.5517,
25
+ -0.3632,
26
+ -0.1922,
27
+ -0.9497,
28
+ 0.2503,
29
+ -0.2921
30
+ ],
31
+ "latents_std": [
32
+ 2.8184,
33
+ 1.4541,
34
+ 2.3275,
35
+ 2.6558,
36
+ 1.2196,
37
+ 1.7708,
38
+ 2.6052,
39
+ 2.0743,
40
+ 3.2687,
41
+ 2.1526,
42
+ 2.8652,
43
+ 1.5579,
44
+ 1.6382,
45
+ 1.1253,
46
+ 2.8251,
47
+ 1.916
48
+ ],
49
+ "num_res_blocks": 2,
50
+ "temperal_downsample": [
51
+ false,
52
+ true,
53
+ true
54
+ ],
55
+ "z_dim": 16
56
+ }
57
+
58
+
59
+
60
+
61
+
62
+
63
+
64
+
65
+
66
+
67
+
68
+
69
+
70
+
71
+
72
+
73
+
74
+
75
+
76
+
77
+
78
+
79
+
80
+
vae/tokenizer.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3c6eaf3c0e2b7db5d607b3f51fdd590aad844055102d369cdc64d6a6a9c8b631
3
+ size 507591212