sychen52 commited on
Commit
fab0aec
·
0 Parent(s):

Super-squash branch 'main' using huggingface_hub

Browse files
.gitattributes ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
36
+ tokenizer.json filter=lfs diff=lfs merge=lfs -text
37
+ model.safetensors.index.json filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,214 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ pipeline_tag: image-text-to-text
3
+ base_model:
4
+ - Qwen/Qwen3.8-Flash-Next
5
+ license: other
6
+ license_name: nvidia-open-model-license
7
+ library_name: Model Optimizer
8
+ tags:
9
+ - nvidia
10
+ - ModelOpt
11
+ - Qwen3.8
12
+ - quantized
13
+ - FP4
14
+ - fp4
15
+ ---
16
+
17
+ # Model Overview
18
+
19
+ ## Description:
20
+ The NVIDIA Qwen3.8-Flash-Next NVFP4 model is the quantized version of Alibaba's Qwen3.8-Flash-Next model, which is an auto-regressive language model that uses an optimized transformer architecture. Qwen3.8-Flash-Next is a causal language model with a vision encoder, hybrid attention (Gated DeltaNet and Qwen Sparse Attention), Mixture-of-Experts, gated residual streams, and n-gram embeddings. For more information, please check [here](https://huggingface.co/Qwen/Qwen3.8-Flash-Next). The NVIDIA Qwen3.8-Flash-Next NVFP4 model is quantized with [Model Optimizer](https://github.com/NVIDIA/Model-Optimizer).
21
+
22
+ This model is ready for commercial or non-commercial use. <br>
23
+
24
+ ## Third-Party Community Consideration
25
+ This model is not owned or developed by NVIDIA. This model has been developed and built to a third-party's requirements for this application and use case; see link to Non-NVIDIA [(Qwen3.8-Flash-Next) Model Card](https://huggingface.co/Qwen/Qwen3.8-Flash-Next) from Qwen.
26
+
27
+ ### License/Terms of Use:
28
+ Governing Terms: Use of this model is governed by the [NVIDIA Open Model License](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/).
29
+
30
+ **ADDITIONAL INFORMATION** : [Qwen Community License 1.0](https://huggingface.co/Qwen/Qwen3.8-Flash-Next/blob/main/LICENSE). **Qwen3.8-Flash-Next** .
31
+
32
+ ### Deployment Geography:
33
+ Global <br>
34
+
35
+ ### Use Case: <br>
36
+ Developers looking to take off-the-shelf, pre-quantized models for deployment in AI Agent systems, chatbots, RAG systems, and other AI-powered applications. <br>
37
+
38
+ ### Release Date: <br>
39
+ Hugging Face 08/31/2026 via https://huggingface.co/nvidia/Qwen3.8-Flash-Next-NVFP4 <br>
40
+
41
+ ## Evaluation
42
+ The accuracy benchmark results are presented in the table below:
43
+ <table>
44
+ <tr>
45
+ <td><strong>Precision</strong>
46
+ </td>
47
+ <td><strong>GPQA Diamond</strong>
48
+ </td>
49
+ <td><strong>HLE</strong>
50
+ </td>
51
+ <td><strong>τ²-Bench Telecom</strong>
52
+ </td>
53
+ <td><strong>MMMU Pro</strong>
54
+ </td>
55
+ <td><strong>SciCode</strong>
56
+ </td>
57
+ <td><strong>AA-LCR</strong>
58
+ </td>
59
+ <td><strong>IFBench</strong>
60
+ </td>
61
+ <td><strong>Omniscience</strong>
62
+ </td>
63
+ <td><strong>Terminal-Bench 2.1</strong>
64
+ </td>
65
+ </tr>
66
+ <tr>
67
+ <td>FP8
68
+ </td>
69
+ <td><strong>92.0</strong>
70
+ </td>
71
+ <td><strong>34.7</strong>
72
+ </td>
73
+ <td><strong>90.8</strong>
74
+ </td>
75
+ <td><strong>77.1</strong>
76
+ </td>
77
+ <td><strong>16.3</strong>
78
+ </td>
79
+ <td><strong>71.9</strong>
80
+ </td>
81
+ <td><strong>80.5</strong>
82
+ </td>
83
+ <td><strong>28.1</strong>
84
+ </td>
85
+ <td><strong>83.3</strong>
86
+ </td>
87
+ </tr>
88
+ <tr>
89
+ <td>NVFP4
90
+ </td>
91
+ <td><strong>91.5</strong>
92
+ </td>
93
+ <td><strong>35.4</strong>
94
+ </td>
95
+ <td><strong>90.1</strong>
96
+ </td>
97
+ <td><strong>78.3</strong>
98
+ </td>
99
+ <td><strong>18.8</strong>
100
+ </td>
101
+ <td><strong>74.1</strong>
102
+ </td>
103
+ <td><strong>81.0</strong>
104
+ </td>
105
+ <td><strong>27.6</strong>
106
+ </td>
107
+ <td><strong>82.9</strong>
108
+ </td>
109
+ </tr>
110
+ </table>
111
+
112
+
113
+ > Baseline: [Qwen3.8-Flash-Next-FP8](https://huggingface.co/Qwen/Qwen3.8-Flash-Next-FP8). Benchmarked with temperature=1.0, top_p=0.95, max_new_tokens=131072, and reasoning_effort=xhigh.
114
+
115
+ ## References
116
+ NVIDIA Model Optimizer: https://github.com/NVIDIA/Model-Optimizer
117
+
118
+ ## Model Architecture:
119
+ **Architecture Type:** Transformers <br>
120
+ **Network Architecture:** Hybrid Attention (Gated DeltaNet and Qwen Sparse Attention) with Mixture-of-Experts (`Qwen4ExpForConditionalGeneration`) <br>
121
+ **Number of Model Parameters:** 125B in total and 6B activated, plus 51B n-gram embedding and 4B MTP <br>
122
+ **This model was developed based on [Qwen3.8-Flash-Next](https://huggingface.co/Qwen/Qwen3.8-Flash-Next)** <br>
123
+
124
+ ## Input:
125
+ **Input Type(s):** Text, Image, Video <br>
126
+ **Input Format(s):** String, Red, Green, Blue (RGB), Video (MP4/WebM) <br>
127
+ **Input Parameters:** One-Dimensional (1D), Two-Dimensional (2D), Three-Dimensional (3D) <br>
128
+ **Other Properties Related to Input:** Context length up to 262K natively and extensible up to 1,000,000 tokens. <br>
129
+
130
+ ## Output:
131
+ **Output Type(s):** Text <br>
132
+ **Output Format:** String <br>
133
+ **Output Parameters:** One-Dimensional (1D): Sequences <br>
134
+ **Other Properties Related to Output:** None <br>
135
+
136
+ 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>
137
+
138
+ ## Software Integration:
139
+ **Supported Runtime Engine(s):** <br>
140
+ * **vLLM** <br>
141
+
142
+ **Post-Training Quantization Toolkit:** <br>
143
+ * [NVIDIA Model Optimizer](https://github.com/NVIDIA/Model-Optimizer) v0.46.0 with [Transformers v5.16.0](https://github.com/huggingface/transformers/releases/tag/v5.16.0) or later for upstream Qwen4-Exp support <br>
144
+
145
+ **Supported Hardware Microarchitecture Compatibility:** <br>
146
+ * NVIDIA Blackwell B200 <br>
147
+ * NVIDIA Blackwell B300 <br>
148
+
149
+ **Preferred Operating System(s):** <br>
150
+ * Linux <br>
151
+
152
+ The integration of foundation and fine-tuned models into AI systems requires additional testing using use-case-specific data to ensure safe and effective deployment. Following the V-model methodology, iterative testing and validation at both unit and system levels are essential to mitigate risks, meet technical and functional requirements, and ensure compliance with safety and ethical standards before deployment.
153
+
154
+ ## Model Version(s):
155
+ The model version is NVFP4 1.0 and was produced with nvidia-modelopt **v0.46.0**. <br>
156
+
157
+ ## Training and Evaluation Datasets:
158
+
159
+ ## Calibration Dataset:
160
+ **Link:** [cnn_dailymail](https://huggingface.co/datasets/abisee/cnn_dailymail), [Nemotron-Post-Training-Dataset-v2](https://huggingface.co/datasets/nvidia/Nemotron-Post-Training-Dataset-v2) <br>
161
+ **Data Modality:** Text <br>
162
+ **Data Collection Method by dataset:** Automated. <br>
163
+ **Labeling Method by dataset:** Automated. <br>
164
+ **Properties:** The cnn_dailymail dataset is an English-language dataset containing just over 300k unique news articles as written by journalists at CNN and the Daily Mail. The Nemotron-Post-Training-Dataset-v2 is a post-training dataset curated by NVIDIA containing multi-turn conversations across diverse topics. <br>
165
+
166
+ ## Training Dataset:
167
+ **Data Modality:** Undisclosed <br>
168
+ **Data Collection Method by dataset:** Undisclosed <br>
169
+ **Labeling Method by dataset:** Undisclosed <br>
170
+ **Properties:** Undisclosed<br>
171
+ **Image Training Data Size:** Undisclosed<br>
172
+ **Text Training Data Size:** Undisclosed<br>
173
+ **Video Training Data Size:** Undisclosed<br>
174
+
175
+ ## Evaluation Dataset:
176
+ **Datasets:** GPQA Diamond, HLE, τ²-Bench Telecom, MMMU Pro, SciCode, AA-LCR, IFBench, Omniscience, Terminal-Bench 2.1 <br>
177
+ **Data Collection Method by dataset:** Hybrid: Automated, Manually-Collected <br>
178
+ **Labeling Method by dataset:** Hybrid: Manually-Labeled, Automated <br>
179
+ **Properties:** We evaluated the model on text-based reasoning, coding, agentic tool-use, and multimodal benchmarks: GPQA Diamond contains 448 graduate-level multiple-choice questions written by domain experts in biology, physics, and chemistry. HLE (Humanity's Last Exam) is an expert-level academic benchmark with 2158 text-only questions across mathematics, humanities and the natural sciences. τ²-Bench Telecom evaluates agentic tool-use and policy-adherence capabilities in dual-control telecom customer-service scenarios where the model interacts with a simulated user and external tools to resolve account issues. MMMU Pro is the more challenging version of the Massive Multi-discipline Multimodal Understanding benchmark, measuring college-level multimodal reasoning across diverse disciplines with expanded answer choices and a vision-only input setting. SciCode evaluates scientific coding capabilities. AA-LCR (Artificial Analysis Long Context Recall) evaluates a model's ability to accurately retrieve and recall information from long input contexts. IFBench evaluates instruction-following capabilities across diverse and structured task constraints. Omniscience evaluates broad factual knowledge and reasoning. Terminal-Bench 2.1 evaluates agentic coding performance in realistic terminal environments. <br>
180
+
181
+ ## Inference:
182
+ **Acceleration Engine:** **vLLM** <br>
183
+ **Test Hardware:** **NVIDIA Blackwell B200 and B300** <br>
184
+
185
+ ## Post Training Quantization
186
+ This is a mixed-precision checkpoint. The routed MoE expert linear layers in the main language model use W4A4 NVFP4 with MSE-calibrated weight scales. Attention layers, shared experts, and other main-model layers remain in BF16. The MTP routed experts use 128x128 block-scaled FP8, and the PLE n-gram embedding uses per-tensor FP8. The resulting checkpoint is approximately 2.7x smaller than the BF16 source checkpoint (about a 63% reduction in disk size).
187
+
188
+ The main language model's routed experts were quantized directly from [Qwen/Qwen3.8-Flash-Next](https://huggingface.co/Qwen/Qwen3.8-Flash-Next). The complete MTP module and PLE n-gram embedding match [Qwen/Qwen3.8-Flash-Next-FP8](https://huggingface.co/Qwen/Qwen3.8-Flash-Next-FP8) byte-for-byte. The FP8 MTP routed-expert tensors and PLE tensors were copied from that checkpoint; the remaining BF16 MTP tensors were already identical between the BF16 and FP8 sources and did not require replacement. No fine-tuning was performed.
189
+
190
+ ## Usage
191
+
192
+ The checkpoint combines NVFP4 routed experts with an FP8 PLE n-gram embedding. Use [vLLM commit `d4d703caf908786416585ceb1f369e2e0363358b`](https://github.com/vllm-project/vllm/commit/d4d703caf908786416585ceb1f369e2e0363358b) or any later commit to run this checkpoint.
193
+
194
+ Run the sample command below from a compatible vLLM environment or container:
195
+
196
+ ```sh
197
+ vllm serve nvidia/Qwen3.8-Flash-Next-NVFP4 \
198
+ --tensor-parallel-size 8 \
199
+ --quantization modelopt \
200
+ --max-model-len 262144 \
201
+ --reasoning-parser qwen3 \
202
+ --trust-remote-code
203
+ ```
204
+
205
+ ## Model Limitations:
206
+ The base model was trained on data that contains toxic language and societal biases originally crawled from the internet. Therefore, the model may amplify those biases and return toxic responses especially when prompted with toxic prompts. The model may generate answers that may be inaccurate, omit key information, or include irrelevant or redundant text producing socially unacceptable or undesirable text, even if the prompt itself does not include anything explicitly offensive.
207
+
208
+ ## Ethical Considerations
209
+
210
+ 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. 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.
211
+
212
+ Please make sure you have proper rights and permissions for all input image and video content; if image or video includes people, personal health information, or intellectual property, the image or video generated will not blur or maintain proportions of image subjects included.
213
+
214
+ Please report model quality, risk, security vulnerabilities or NVIDIA AI Concerns [here](https://www.nvidia.com/en-us/support/submit-security-vulnerability/).
chat_template.jinja ADDED
@@ -0,0 +1,170 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- set image_count = namespace(value=0) %}
2
+ {%- set video_count = namespace(value=0) %}
3
+ {%- macro render_content(content, do_vision_count, is_system_content=false) %}
4
+ {%- if content is string %}
5
+ {{- content }}
6
+ {%- elif content is iterable and content is not mapping %}
7
+ {%- for item in content %}
8
+ {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
9
+ {%- if is_system_content %}
10
+ {{- raise_exception('System message cannot contain images.') }}
11
+ {%- endif %}
12
+ {%- if do_vision_count %}
13
+ {%- set image_count.value = image_count.value + 1 %}
14
+ {%- endif %}
15
+ {%- if add_vision_id %}
16
+ {{- 'Picture ' ~ image_count.value ~ ': ' }}
17
+ {%- endif %}
18
+ {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
19
+ {%- elif 'video' in item or item.type == 'video' %}
20
+ {%- if is_system_content %}
21
+ {{- raise_exception('System message cannot contain videos.') }}
22
+ {%- endif %}
23
+ {%- if do_vision_count %}
24
+ {%- set video_count.value = video_count.value + 1 %}
25
+ {%- endif %}
26
+ {%- if add_vision_id %}
27
+ {{- 'Video ' ~ video_count.value ~ ': ' }}
28
+ {%- endif %}
29
+ {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
30
+ {%- elif 'text' in item %}
31
+ {{- item.text }}
32
+ {%- else %}
33
+ {{- raise_exception('Unexpected item type in content.') }}
34
+ {%- endif %}
35
+ {%- endfor %}
36
+ {%- elif content is none or content is undefined %}
37
+ {{- '' }}
38
+ {%- else %}
39
+ {{- raise_exception('Unexpected content type.') }}
40
+ {%- endif %}
41
+ {%- endmacro %}
42
+ {%- if not messages %}
43
+ {{- raise_exception('No messages provided.') }}
44
+ {%- endif %}
45
+ {%- set reasoning_instructions = '' %}
46
+ {%- if enable_thinking is undefined or enable_thinking is true %}
47
+ {%- set resolved_reasoning_effort = reasoning_effort|default('xhigh') %}
48
+ {%- if resolved_reasoning_effort not in ('xhigh', 'medium', 'low') %}
49
+ {{- raise_exception('Unexpected reasoning effort ' ~ reasoning_effort ~ '. Supported types are xhigh (default), medium, and low.') }}
50
+ {%- endif %}
51
+ {%- if resolved_reasoning_effort == 'xhigh' %}
52
+ {%- set reasoning_instructions = 'Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.' %}
53
+ {%- elif resolved_reasoning_effort == 'low' %}
54
+ {%- set reasoning_instructions = 'Reasoning effort is set to low. Keep your thinking brief and focused, moving directly to the conclusion without unnecessary elaboration.' %}
55
+ {%- endif %}
56
+ {%- endif %}
57
+ {%- if tools and tools is iterable and tools is not mapping %}
58
+ {{- '<|im_start|>system\n' }}
59
+ {%- if reasoning_instructions %}
60
+ {{- reasoning_instructions + '\n\n' }}
61
+ {%- endif %}
62
+ {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
63
+ {%- for tool in tools %}
64
+ {{- "\n" }}
65
+ {{- tool | tojson }}
66
+ {%- endfor %}
67
+ {{- "\n</tools>" }}
68
+ {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
69
+ {%- if messages[0].role == 'system' %}
70
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
71
+ {%- if content %}
72
+ {{- '\n\n' + content }}
73
+ {%- endif %}
74
+ {%- endif %}
75
+ {{- '<|im_end|>\n' }}
76
+ {%- else %}
77
+ {%- if messages[0].role == 'system' %}
78
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
79
+ {%- if content %}
80
+ {{- '<|im_start|>system\n' + (reasoning_instructions + '\n\n' if reasoning_instructions else '') + content + '<|im_end|>\n' }}
81
+ {%- elif reasoning_instructions %}
82
+ {{- '<|im_start|>system\n' + reasoning_instructions + '<|im_end|>\n' }}
83
+ {%- endif %}
84
+ {%- elif reasoning_instructions %}
85
+ {{- '<|im_start|>system\n' + reasoning_instructions + '<|im_end|>\n' }}
86
+ {%- endif %}
87
+ {%- endif %}
88
+ {%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
89
+ {%- for message in messages[::-1] %}
90
+ {%- set index = (messages|length - 1) - loop.index0 %}
91
+ {%- if ns.multi_step_tool and message.role == "user" %}
92
+ {%- set content = render_content(message.content, false)|trim %}
93
+ {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
94
+ {%- set ns.multi_step_tool = false %}
95
+ {%- set ns.last_query_index = index %}
96
+ {%- endif %}
97
+ {%- endif %}
98
+ {%- endfor %}
99
+ {%- if ns.multi_step_tool %}
100
+ {{- raise_exception('No user query found in messages.') }}
101
+ {%- endif %}
102
+ {%- for message in messages %}
103
+ {%- set content = render_content(message.content, true)|trim %}
104
+ {%- if message.role == "system" %}
105
+ {%- if not loop.first %}
106
+ {{- raise_exception('System message must be at the beginning.') }}
107
+ {%- endif %}
108
+ {%- elif message.role == "user" %}
109
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
110
+ {%- elif message.role == "assistant" %}
111
+ {%- set reasoning_content = '' %}
112
+ {%- if message.reasoning_content is string %}
113
+ {%- set reasoning_content = message.reasoning_content %}
114
+ {%- endif %}
115
+ {%- set reasoning_content = reasoning_content|trim %}
116
+ {%- if preserve_thinking is undefined or preserve_thinking is true or loop.index0 > ns.last_query_index %}
117
+ {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
118
+ {%- else %}
119
+ {{- '<|im_start|>' + message.role + '\n' + content }}
120
+ {%- endif %}
121
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
122
+ {%- for tool_call in message.tool_calls %}
123
+ {%- if tool_call.function is defined %}
124
+ {%- set tool_call = tool_call.function %}
125
+ {%- endif %}
126
+ {%- if loop.first %}
127
+ {%- if content|trim %}
128
+ {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
129
+ {%- else %}
130
+ {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
131
+ {%- endif %}
132
+ {%- else %}
133
+ {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
134
+ {%- endif %}
135
+ {%- if tool_call.arguments is defined and tool_call.arguments != '' %}
136
+ {%- for args_name, args_value in tool_call.arguments|items %}
137
+ {{- '<parameter=' + args_name + '>\n' }}
138
+ {%- set args_value = args_value | string if args_value is string else args_value | tojson | safe %}
139
+ {{- args_value }}
140
+ {{- '\n</parameter>\n' }}
141
+ {%- endfor %}
142
+ {%- endif %}
143
+ {{- '</function>\n</tool_call>' }}
144
+ {%- endfor %}
145
+ {%- endif %}
146
+ {{- '<|im_end|>\n' }}
147
+ {%- elif message.role == "tool" %}
148
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
149
+ {{- '<|im_start|>user' }}
150
+ {%- endif %}
151
+ {{- '\n<tool_response>\n' }}
152
+ {{- content }}
153
+ {{- '\n</tool_response>' }}
154
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
155
+ {{- '<|im_end|>\n' }}
156
+ {%- elif loop.last %}
157
+ {{- '<|im_end|>\n' }}
158
+ {%- endif %}
159
+ {%- else %}
160
+ {{- raise_exception('Unexpected message role.') }}
161
+ {%- endif %}
162
+ {%- endfor %}
163
+ {%- if add_generation_prompt %}
164
+ {{- '<|im_start|>assistant\n' }}
165
+ {%- if enable_thinking is defined and enable_thinking is false %}
166
+ {{- '<think>\n\n</think>\n\n' }}
167
+ {%- else %}
168
+ {{- '<think>\n' }}
169
+ {%- endif %}
170
+ {%- endif %}
config.json ADDED
@@ -0,0 +1,755 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen4ExpForConditionalGeneration"
4
+ ],
5
+ "dtype": "bfloat16",
6
+ "image_token_id": 248056,
7
+ "language_model_only": false,
8
+ "model_type": "qwen4_exp",
9
+ "text_config": {
10
+ "attention_bias": false,
11
+ "attention_dropout": 0.0,
12
+ "bos_token_id": 248044,
13
+ "dtype": "bfloat16",
14
+ "eos_token_id": 248044,
15
+ "full_attention_interval": 4,
16
+ "hc_count": 4,
17
+ "hc_lowrank": 320,
18
+ "head_dim": 256,
19
+ "heads_per_ngram": 8,
20
+ "hidden_act": "silu",
21
+ "hidden_size": 2560,
22
+ "indexer_budget": 2048,
23
+ "indexer_compress_ratio": 4,
24
+ "indexer_head_dim": 128,
25
+ "indexer_kv_heads": 1,
26
+ "indexer_n_heads": 4,
27
+ "initializer_range": 0.02,
28
+ "layer_types": [
29
+ "linear_attention",
30
+ "linear_attention",
31
+ "linear_attention",
32
+ "full_attention",
33
+ "linear_attention",
34
+ "linear_attention",
35
+ "linear_attention",
36
+ "full_attention",
37
+ "linear_attention",
38
+ "linear_attention",
39
+ "linear_attention",
40
+ "full_attention",
41
+ "linear_attention",
42
+ "linear_attention",
43
+ "linear_attention",
44
+ "full_attention",
45
+ "linear_attention",
46
+ "linear_attention",
47
+ "linear_attention",
48
+ "full_attention",
49
+ "linear_attention",
50
+ "linear_attention",
51
+ "linear_attention",
52
+ "full_attention",
53
+ "linear_attention",
54
+ "linear_attention",
55
+ "linear_attention",
56
+ "full_attention",
57
+ "linear_attention",
58
+ "linear_attention",
59
+ "linear_attention",
60
+ "full_attention",
61
+ "linear_attention",
62
+ "linear_attention",
63
+ "linear_attention",
64
+ "full_attention",
65
+ "linear_attention",
66
+ "linear_attention",
67
+ "linear_attention",
68
+ "full_attention",
69
+ "linear_attention",
70
+ "linear_attention",
71
+ "linear_attention",
72
+ "full_attention",
73
+ "linear_attention",
74
+ "linear_attention",
75
+ "linear_attention",
76
+ "full_attention"
77
+ ],
78
+ "linear_conv_kernel_dim": 4,
79
+ "linear_key_head_dim": 128,
80
+ "linear_num_key_heads": 16,
81
+ "linear_num_value_heads": 48,
82
+ "linear_value_head_dim": 128,
83
+ "make_ngram_vocab_size_divisible_by": 128,
84
+ "mamba_ssm_dtype": "float32",
85
+ "max_position_embeddings": 262144,
86
+ "model_type": "qwen4_exp_text",
87
+ "moe_intermediate_size": 640,
88
+ "mtp": {
89
+ "hybrid": true,
90
+ "layer_types": [
91
+ "full_attention"
92
+ ],
93
+ "mtp_use_hidden_state_from_layer": null,
94
+ "num_hidden_layers": 1,
95
+ "rope_theta": 10000000
96
+ },
97
+ "mtp_num_hidden_layers": 1,
98
+ "mtp_use_dedicated_embeddings": false,
99
+ "ngram_size": 3,
100
+ "ngram_vocab_size_base": 20000000,
101
+ "norm_topk_prob": true,
102
+ "num_attention_heads": 24,
103
+ "num_experts": 512,
104
+ "num_experts_per_tok": 10,
105
+ "num_hidden_layers": 48,
106
+ "num_key_value_heads": 2,
107
+ "number_of_conv_states": 3,
108
+ "output_gate_type": "sigmoid",
109
+ "output_router_logits": false,
110
+ "pad_token_id": null,
111
+ "partial_rotary_factor": 0.25,
112
+ "ple_conv_kernel_size": 4,
113
+ "ple_embed_dim": 2560,
114
+ "ple_layer_ids": [
115
+ 2
116
+ ],
117
+ "rms_norm_eps": 1e-06,
118
+ "rope_parameters": {
119
+ "mrope_interleaved": true,
120
+ "mrope_section": [
121
+ 11,
122
+ 11,
123
+ 10
124
+ ],
125
+ "partial_rotary_factor": 0.25,
126
+ "rope_theta": 10000000,
127
+ "rope_type": "default"
128
+ },
129
+ "router_aux_loss_coef": 0.001,
130
+ "seed": 1234,
131
+ "shared_expert_intermediate_size": 640,
132
+ "split_ngram_parts": 128,
133
+ "tie_word_embeddings": false,
134
+ "use_cache": true,
135
+ "vocab_size": 248320
136
+ },
137
+ "tie_word_embeddings": false,
138
+ "transformers_version": "5.16.0.dev0",
139
+ "video_token_id": 248057,
140
+ "vision_config": {
141
+ "deepstack_visual_indexes": [],
142
+ "depth": 27,
143
+ "dtype": "bfloat16",
144
+ "hidden_act": "gelu_pytorch_tanh",
145
+ "hidden_size": 1152,
146
+ "in_channels": 3,
147
+ "initializer_range": 0.02,
148
+ "intermediate_size": 4304,
149
+ "model_type": "qwen4_exp_vision",
150
+ "num_heads": 16,
151
+ "num_position_embeddings": 2304,
152
+ "out_hidden_size": 2560,
153
+ "patch_size": 16,
154
+ "spatial_merge_size": 2,
155
+ "temporal_patch_size": 2
156
+ },
157
+ "vision_end_token_id": 248054,
158
+ "vision_start_token_id": 248053,
159
+ "quantization_config": {
160
+ "config_groups": {
161
+ "group_0": {
162
+ "input_activations": {
163
+ "dynamic": false,
164
+ "num_bits": 4,
165
+ "type": "float",
166
+ "group_size": 16
167
+ },
168
+ "weights": {
169
+ "dynamic": false,
170
+ "num_bits": 4,
171
+ "type": "float",
172
+ "group_size": 16
173
+ },
174
+ "targets": [
175
+ "model.language_model.layers.0.mlp.experts",
176
+ "model.language_model.layers.1.mlp.experts",
177
+ "model.language_model.layers.10.mlp.experts",
178
+ "model.language_model.layers.11.mlp.experts",
179
+ "model.language_model.layers.12.mlp.experts",
180
+ "model.language_model.layers.13.mlp.experts",
181
+ "model.language_model.layers.14.mlp.experts",
182
+ "model.language_model.layers.15.mlp.experts",
183
+ "model.language_model.layers.16.mlp.experts",
184
+ "model.language_model.layers.17.mlp.experts",
185
+ "model.language_model.layers.18.mlp.experts",
186
+ "model.language_model.layers.19.mlp.experts",
187
+ "model.language_model.layers.2.mlp.experts",
188
+ "model.language_model.layers.20.mlp.experts",
189
+ "model.language_model.layers.21.mlp.experts",
190
+ "model.language_model.layers.22.mlp.experts",
191
+ "model.language_model.layers.23.mlp.experts",
192
+ "model.language_model.layers.24.mlp.experts",
193
+ "model.language_model.layers.25.mlp.experts",
194
+ "model.language_model.layers.26.mlp.experts",
195
+ "model.language_model.layers.27.mlp.experts",
196
+ "model.language_model.layers.28.mlp.experts",
197
+ "model.language_model.layers.29.mlp.experts",
198
+ "model.language_model.layers.3.mlp.experts",
199
+ "model.language_model.layers.30.mlp.experts",
200
+ "model.language_model.layers.31.mlp.experts",
201
+ "model.language_model.layers.32.mlp.experts",
202
+ "model.language_model.layers.33.mlp.experts",
203
+ "model.language_model.layers.34.mlp.experts",
204
+ "model.language_model.layers.35.mlp.experts",
205
+ "model.language_model.layers.36.mlp.experts",
206
+ "model.language_model.layers.37.mlp.experts",
207
+ "model.language_model.layers.38.mlp.experts",
208
+ "model.language_model.layers.39.mlp.experts",
209
+ "model.language_model.layers.4.mlp.experts",
210
+ "model.language_model.layers.40.mlp.experts",
211
+ "model.language_model.layers.41.mlp.experts",
212
+ "model.language_model.layers.42.mlp.experts",
213
+ "model.language_model.layers.43.mlp.experts",
214
+ "model.language_model.layers.44.mlp.experts",
215
+ "model.language_model.layers.45.mlp.experts",
216
+ "model.language_model.layers.46.mlp.experts",
217
+ "model.language_model.layers.47.mlp.experts",
218
+ "model.language_model.layers.5.mlp.experts",
219
+ "model.language_model.layers.6.mlp.experts",
220
+ "model.language_model.layers.7.mlp.experts",
221
+ "model.language_model.layers.8.mlp.experts",
222
+ "model.language_model.layers.9.mlp.experts"
223
+ ]
224
+ },
225
+ "group_1": {
226
+ "input_activations": {
227
+ "dynamic": true,
228
+ "num_bits": 8,
229
+ "type": "float",
230
+ "group_size": 128
231
+ },
232
+ "weights": {
233
+ "dynamic": false,
234
+ "num_bits": 8,
235
+ "type": "float",
236
+ "group_size": 128
237
+ },
238
+ "targets": [
239
+ "mtp.layers.0.mlp.experts"
240
+ ]
241
+ },
242
+ "group_2": {
243
+ "weights": {
244
+ "dynamic": false,
245
+ "num_bits": 8,
246
+ "type": "float"
247
+ },
248
+ "targets": [
249
+ "model.language_model.layers.1.ple.ple_embedding.ngram_embedding"
250
+ ]
251
+ }
252
+ },
253
+ "ignore": [
254
+ "lm_head",
255
+ "model.language_model.embed_tokens",
256
+ "model.language_model.hyper_connection_mixer*",
257
+ "model.language_model.layers.0.attn_hyper_connection*",
258
+ "model.language_model.layers.0.linear_attn*",
259
+ "model.language_model.layers.0.mlp.gate",
260
+ "model.language_model.layers.0.mlp.shared_expert*",
261
+ "model.language_model.layers.0.mlp.shared_expert_gate",
262
+ "model.language_model.layers.0.mlp_hyper_connection*",
263
+ "model.language_model.layers.1.attn_hyper_connection*",
264
+ "model.language_model.layers.1.linear_attn*",
265
+ "model.language_model.layers.1.mlp.gate",
266
+ "model.language_model.layers.1.mlp.shared_expert*",
267
+ "model.language_model.layers.1.mlp.shared_expert_gate",
268
+ "model.language_model.layers.1.mlp_hyper_connection*",
269
+ "model.language_model.layers.10.attn_hyper_connection*",
270
+ "model.language_model.layers.10.linear_attn*",
271
+ "model.language_model.layers.10.mlp.gate",
272
+ "model.language_model.layers.10.mlp.shared_expert*",
273
+ "model.language_model.layers.10.mlp.shared_expert_gate",
274
+ "model.language_model.layers.10.mlp_hyper_connection*",
275
+ "model.language_model.layers.11.attn_hyper_connection*",
276
+ "model.language_model.layers.11.mlp.gate",
277
+ "model.language_model.layers.11.mlp.shared_expert*",
278
+ "model.language_model.layers.11.mlp.shared_expert_gate",
279
+ "model.language_model.layers.11.mlp_hyper_connection*",
280
+ "model.language_model.layers.11.self_attn*",
281
+ "model.language_model.layers.12.attn_hyper_connection*",
282
+ "model.language_model.layers.12.linear_attn*",
283
+ "model.language_model.layers.12.mlp.gate",
284
+ "model.language_model.layers.12.mlp.shared_expert*",
285
+ "model.language_model.layers.12.mlp.shared_expert_gate",
286
+ "model.language_model.layers.12.mlp_hyper_connection*",
287
+ "model.language_model.layers.13.attn_hyper_connection*",
288
+ "model.language_model.layers.13.linear_attn*",
289
+ "model.language_model.layers.13.mlp.gate",
290
+ "model.language_model.layers.13.mlp.shared_expert*",
291
+ "model.language_model.layers.13.mlp.shared_expert_gate",
292
+ "model.language_model.layers.13.mlp_hyper_connection*",
293
+ "model.language_model.layers.14.attn_hyper_connection*",
294
+ "model.language_model.layers.14.linear_attn*",
295
+ "model.language_model.layers.14.mlp.gate",
296
+ "model.language_model.layers.14.mlp.shared_expert*",
297
+ "model.language_model.layers.14.mlp.shared_expert_gate",
298
+ "model.language_model.layers.14.mlp_hyper_connection*",
299
+ "model.language_model.layers.15.attn_hyper_connection*",
300
+ "model.language_model.layers.15.mlp.gate",
301
+ "model.language_model.layers.15.mlp.shared_expert*",
302
+ "model.language_model.layers.15.mlp.shared_expert_gate",
303
+ "model.language_model.layers.15.mlp_hyper_connection*",
304
+ "model.language_model.layers.15.self_attn*",
305
+ "model.language_model.layers.16.attn_hyper_connection*",
306
+ "model.language_model.layers.16.linear_attn*",
307
+ "model.language_model.layers.16.mlp.gate",
308
+ "model.language_model.layers.16.mlp.shared_expert*",
309
+ "model.language_model.layers.16.mlp.shared_expert_gate",
310
+ "model.language_model.layers.16.mlp_hyper_connection*",
311
+ "model.language_model.layers.17.attn_hyper_connection*",
312
+ "model.language_model.layers.17.linear_attn*",
313
+ "model.language_model.layers.17.mlp.gate",
314
+ "model.language_model.layers.17.mlp.shared_expert*",
315
+ "model.language_model.layers.17.mlp.shared_expert_gate",
316
+ "model.language_model.layers.17.mlp_hyper_connection*",
317
+ "model.language_model.layers.18.attn_hyper_connection*",
318
+ "model.language_model.layers.18.linear_attn*",
319
+ "model.language_model.layers.18.mlp.gate",
320
+ "model.language_model.layers.18.mlp.shared_expert*",
321
+ "model.language_model.layers.18.mlp.shared_expert_gate",
322
+ "model.language_model.layers.18.mlp_hyper_connection*",
323
+ "model.language_model.layers.19.attn_hyper_connection*",
324
+ "model.language_model.layers.19.mlp.gate",
325
+ "model.language_model.layers.19.mlp.shared_expert*",
326
+ "model.language_model.layers.19.mlp.shared_expert_gate",
327
+ "model.language_model.layers.19.mlp_hyper_connection*",
328
+ "model.language_model.layers.19.self_attn*",
329
+ "model.language_model.layers.2.attn_hyper_connection*",
330
+ "model.language_model.layers.2.linear_attn*",
331
+ "model.language_model.layers.2.mlp.gate",
332
+ "model.language_model.layers.2.mlp.shared_expert*",
333
+ "model.language_model.layers.2.mlp.shared_expert_gate",
334
+ "model.language_model.layers.2.mlp_hyper_connection*",
335
+ "model.language_model.layers.20.attn_hyper_connection*",
336
+ "model.language_model.layers.20.linear_attn*",
337
+ "model.language_model.layers.20.mlp.gate",
338
+ "model.language_model.layers.20.mlp.shared_expert*",
339
+ "model.language_model.layers.20.mlp.shared_expert_gate",
340
+ "model.language_model.layers.20.mlp_hyper_connection*",
341
+ "model.language_model.layers.21.attn_hyper_connection*",
342
+ "model.language_model.layers.21.linear_attn*",
343
+ "model.language_model.layers.21.mlp.gate",
344
+ "model.language_model.layers.21.mlp.shared_expert*",
345
+ "model.language_model.layers.21.mlp.shared_expert_gate",
346
+ "model.language_model.layers.21.mlp_hyper_connection*",
347
+ "model.language_model.layers.22.attn_hyper_connection*",
348
+ "model.language_model.layers.22.linear_attn*",
349
+ "model.language_model.layers.22.mlp.gate",
350
+ "model.language_model.layers.22.mlp.shared_expert*",
351
+ "model.language_model.layers.22.mlp.shared_expert_gate",
352
+ "model.language_model.layers.22.mlp_hyper_connection*",
353
+ "model.language_model.layers.23.attn_hyper_connection*",
354
+ "model.language_model.layers.23.mlp.gate",
355
+ "model.language_model.layers.23.mlp.shared_expert*",
356
+ "model.language_model.layers.23.mlp.shared_expert_gate",
357
+ "model.language_model.layers.23.mlp_hyper_connection*",
358
+ "model.language_model.layers.23.self_attn*",
359
+ "model.language_model.layers.24.attn_hyper_connection*",
360
+ "model.language_model.layers.24.linear_attn*",
361
+ "model.language_model.layers.24.mlp.gate",
362
+ "model.language_model.layers.24.mlp.shared_expert*",
363
+ "model.language_model.layers.24.mlp.shared_expert_gate",
364
+ "model.language_model.layers.24.mlp_hyper_connection*",
365
+ "model.language_model.layers.25.attn_hyper_connection*",
366
+ "model.language_model.layers.25.linear_attn*",
367
+ "model.language_model.layers.25.mlp.gate",
368
+ "model.language_model.layers.25.mlp.shared_expert*",
369
+ "model.language_model.layers.25.mlp.shared_expert_gate",
370
+ "model.language_model.layers.25.mlp_hyper_connection*",
371
+ "model.language_model.layers.26.attn_hyper_connection*",
372
+ "model.language_model.layers.26.linear_attn*",
373
+ "model.language_model.layers.26.mlp.gate",
374
+ "model.language_model.layers.26.mlp.shared_expert*",
375
+ "model.language_model.layers.26.mlp.shared_expert_gate",
376
+ "model.language_model.layers.26.mlp_hyper_connection*",
377
+ "model.language_model.layers.27.attn_hyper_connection*",
378
+ "model.language_model.layers.27.mlp.gate",
379
+ "model.language_model.layers.27.mlp.shared_expert*",
380
+ "model.language_model.layers.27.mlp.shared_expert_gate",
381
+ "model.language_model.layers.27.mlp_hyper_connection*",
382
+ "model.language_model.layers.27.self_attn*",
383
+ "model.language_model.layers.28.attn_hyper_connection*",
384
+ "model.language_model.layers.28.linear_attn*",
385
+ "model.language_model.layers.28.mlp.gate",
386
+ "model.language_model.layers.28.mlp.shared_expert*",
387
+ "model.language_model.layers.28.mlp.shared_expert_gate",
388
+ "model.language_model.layers.28.mlp_hyper_connection*",
389
+ "model.language_model.layers.29.attn_hyper_connection*",
390
+ "model.language_model.layers.29.linear_attn*",
391
+ "model.language_model.layers.29.mlp.gate",
392
+ "model.language_model.layers.29.mlp.shared_expert*",
393
+ "model.language_model.layers.29.mlp.shared_expert_gate",
394
+ "model.language_model.layers.29.mlp_hyper_connection*",
395
+ "model.language_model.layers.3.attn_hyper_connection*",
396
+ "model.language_model.layers.3.mlp.gate",
397
+ "model.language_model.layers.3.mlp.shared_expert*",
398
+ "model.language_model.layers.3.mlp.shared_expert_gate",
399
+ "model.language_model.layers.3.mlp_hyper_connection*",
400
+ "model.language_model.layers.3.self_attn*",
401
+ "model.language_model.layers.30.attn_hyper_connection*",
402
+ "model.language_model.layers.30.linear_attn*",
403
+ "model.language_model.layers.30.mlp.gate",
404
+ "model.language_model.layers.30.mlp.shared_expert*",
405
+ "model.language_model.layers.30.mlp.shared_expert_gate",
406
+ "model.language_model.layers.30.mlp_hyper_connection*",
407
+ "model.language_model.layers.31.attn_hyper_connection*",
408
+ "model.language_model.layers.31.mlp.gate",
409
+ "model.language_model.layers.31.mlp.shared_expert*",
410
+ "model.language_model.layers.31.mlp.shared_expert_gate",
411
+ "model.language_model.layers.31.mlp_hyper_connection*",
412
+ "model.language_model.layers.31.self_attn*",
413
+ "model.language_model.layers.32.attn_hyper_connection*",
414
+ "model.language_model.layers.32.linear_attn*",
415
+ "model.language_model.layers.32.mlp.gate",
416
+ "model.language_model.layers.32.mlp.shared_expert*",
417
+ "model.language_model.layers.32.mlp.shared_expert_gate",
418
+ "model.language_model.layers.32.mlp_hyper_connection*",
419
+ "model.language_model.layers.33.attn_hyper_connection*",
420
+ "model.language_model.layers.33.linear_attn*",
421
+ "model.language_model.layers.33.mlp.gate",
422
+ "model.language_model.layers.33.mlp.shared_expert*",
423
+ "model.language_model.layers.33.mlp.shared_expert_gate",
424
+ "model.language_model.layers.33.mlp_hyper_connection*",
425
+ "model.language_model.layers.34.attn_hyper_connection*",
426
+ "model.language_model.layers.34.linear_attn*",
427
+ "model.language_model.layers.34.mlp.gate",
428
+ "model.language_model.layers.34.mlp.shared_expert*",
429
+ "model.language_model.layers.34.mlp.shared_expert_gate",
430
+ "model.language_model.layers.34.mlp_hyper_connection*",
431
+ "model.language_model.layers.35.attn_hyper_connection*",
432
+ "model.language_model.layers.35.mlp.gate",
433
+ "model.language_model.layers.35.mlp.shared_expert*",
434
+ "model.language_model.layers.35.mlp.shared_expert_gate",
435
+ "model.language_model.layers.35.mlp_hyper_connection*",
436
+ "model.language_model.layers.35.self_attn*",
437
+ "model.language_model.layers.36.attn_hyper_connection*",
438
+ "model.language_model.layers.36.linear_attn*",
439
+ "model.language_model.layers.36.mlp.gate",
440
+ "model.language_model.layers.36.mlp.shared_expert*",
441
+ "model.language_model.layers.36.mlp.shared_expert_gate",
442
+ "model.language_model.layers.36.mlp_hyper_connection*",
443
+ "model.language_model.layers.37.attn_hyper_connection*",
444
+ "model.language_model.layers.37.linear_attn*",
445
+ "model.language_model.layers.37.mlp.gate",
446
+ "model.language_model.layers.37.mlp.shared_expert*",
447
+ "model.language_model.layers.37.mlp.shared_expert_gate",
448
+ "model.language_model.layers.37.mlp_hyper_connection*",
449
+ "model.language_model.layers.38.attn_hyper_connection*",
450
+ "model.language_model.layers.38.linear_attn*",
451
+ "model.language_model.layers.38.mlp.gate",
452
+ "model.language_model.layers.38.mlp.shared_expert*",
453
+ "model.language_model.layers.38.mlp.shared_expert_gate",
454
+ "model.language_model.layers.38.mlp_hyper_connection*",
455
+ "model.language_model.layers.39.attn_hyper_connection*",
456
+ "model.language_model.layers.39.mlp.gate",
457
+ "model.language_model.layers.39.mlp.shared_expert*",
458
+ "model.language_model.layers.39.mlp.shared_expert_gate",
459
+ "model.language_model.layers.39.mlp_hyper_connection*",
460
+ "model.language_model.layers.39.self_attn*",
461
+ "model.language_model.layers.4.attn_hyper_connection*",
462
+ "model.language_model.layers.4.linear_attn*",
463
+ "model.language_model.layers.4.mlp.gate",
464
+ "model.language_model.layers.4.mlp.shared_expert*",
465
+ "model.language_model.layers.4.mlp.shared_expert_gate",
466
+ "model.language_model.layers.4.mlp_hyper_connection*",
467
+ "model.language_model.layers.40.attn_hyper_connection*",
468
+ "model.language_model.layers.40.linear_attn*",
469
+ "model.language_model.layers.40.mlp.gate",
470
+ "model.language_model.layers.40.mlp.shared_expert*",
471
+ "model.language_model.layers.40.mlp.shared_expert_gate",
472
+ "model.language_model.layers.40.mlp_hyper_connection*",
473
+ "model.language_model.layers.41.attn_hyper_connection*",
474
+ "model.language_model.layers.41.linear_attn*",
475
+ "model.language_model.layers.41.mlp.gate",
476
+ "model.language_model.layers.41.mlp.shared_expert*",
477
+ "model.language_model.layers.41.mlp.shared_expert_gate",
478
+ "model.language_model.layers.41.mlp_hyper_connection*",
479
+ "model.language_model.layers.42.attn_hyper_connection*",
480
+ "model.language_model.layers.42.linear_attn*",
481
+ "model.language_model.layers.42.mlp.gate",
482
+ "model.language_model.layers.42.mlp.shared_expert*",
483
+ "model.language_model.layers.42.mlp.shared_expert_gate",
484
+ "model.language_model.layers.42.mlp_hyper_connection*",
485
+ "model.language_model.layers.43.attn_hyper_connection*",
486
+ "model.language_model.layers.43.mlp.gate",
487
+ "model.language_model.layers.43.mlp.shared_expert*",
488
+ "model.language_model.layers.43.mlp.shared_expert_gate",
489
+ "model.language_model.layers.43.mlp_hyper_connection*",
490
+ "model.language_model.layers.43.self_attn*",
491
+ "model.language_model.layers.44.attn_hyper_connection*",
492
+ "model.language_model.layers.44.linear_attn*",
493
+ "model.language_model.layers.44.mlp.gate",
494
+ "model.language_model.layers.44.mlp.shared_expert*",
495
+ "model.language_model.layers.44.mlp.shared_expert_gate",
496
+ "model.language_model.layers.44.mlp_hyper_connection*",
497
+ "model.language_model.layers.45.attn_hyper_connection*",
498
+ "model.language_model.layers.45.linear_attn*",
499
+ "model.language_model.layers.45.mlp.gate",
500
+ "model.language_model.layers.45.mlp.shared_expert*",
501
+ "model.language_model.layers.45.mlp.shared_expert_gate",
502
+ "model.language_model.layers.45.mlp_hyper_connection*",
503
+ "model.language_model.layers.46.attn_hyper_connection*",
504
+ "model.language_model.layers.46.linear_attn*",
505
+ "model.language_model.layers.46.mlp.gate",
506
+ "model.language_model.layers.46.mlp.shared_expert*",
507
+ "model.language_model.layers.46.mlp.shared_expert_gate",
508
+ "model.language_model.layers.46.mlp_hyper_connection*",
509
+ "model.language_model.layers.47.attn_hyper_connection*",
510
+ "model.language_model.layers.47.mlp.gate",
511
+ "model.language_model.layers.47.mlp.shared_expert*",
512
+ "model.language_model.layers.47.mlp.shared_expert_gate",
513
+ "model.language_model.layers.47.mlp_hyper_connection*",
514
+ "model.language_model.layers.47.self_attn*",
515
+ "model.language_model.layers.5.attn_hyper_connection*",
516
+ "model.language_model.layers.5.linear_attn*",
517
+ "model.language_model.layers.5.mlp.gate",
518
+ "model.language_model.layers.5.mlp.shared_expert*",
519
+ "model.language_model.layers.5.mlp.shared_expert_gate",
520
+ "model.language_model.layers.5.mlp_hyper_connection*",
521
+ "model.language_model.layers.6.attn_hyper_connection*",
522
+ "model.language_model.layers.6.linear_attn*",
523
+ "model.language_model.layers.6.mlp.gate",
524
+ "model.language_model.layers.6.mlp.shared_expert*",
525
+ "model.language_model.layers.6.mlp.shared_expert_gate",
526
+ "model.language_model.layers.6.mlp_hyper_connection*",
527
+ "model.language_model.layers.7.attn_hyper_connection*",
528
+ "model.language_model.layers.7.mlp.gate",
529
+ "model.language_model.layers.7.mlp.shared_expert*",
530
+ "model.language_model.layers.7.mlp.shared_expert_gate",
531
+ "model.language_model.layers.7.mlp_hyper_connection*",
532
+ "model.language_model.layers.7.self_attn*",
533
+ "model.language_model.layers.8.attn_hyper_connection*",
534
+ "model.language_model.layers.8.linear_attn*",
535
+ "model.language_model.layers.8.mlp.gate",
536
+ "model.language_model.layers.8.mlp.shared_expert*",
537
+ "model.language_model.layers.8.mlp.shared_expert_gate",
538
+ "model.language_model.layers.8.mlp_hyper_connection*",
539
+ "model.language_model.layers.9.attn_hyper_connection*",
540
+ "model.language_model.layers.9.linear_attn*",
541
+ "model.language_model.layers.9.mlp.gate",
542
+ "model.language_model.layers.9.mlp.shared_expert*",
543
+ "model.language_model.layers.9.mlp.shared_expert_gate",
544
+ "model.language_model.layers.9.mlp_hyper_connection*",
545
+ "model.visual*"
546
+ ],
547
+ "quant_algo": "MIXED_PRECISION",
548
+ "producer": {
549
+ "name": "modelopt",
550
+ "version": "0.46.0.dev281+g73d778422"
551
+ },
552
+ "quant_method": "modelopt",
553
+ "quantized_layers": {
554
+ "model.language_model.layers.0.mlp.experts": {
555
+ "quant_algo": "NVFP4",
556
+ "group_size": 16
557
+ },
558
+ "model.language_model.layers.1.mlp.experts": {
559
+ "quant_algo": "NVFP4",
560
+ "group_size": 16
561
+ },
562
+ "model.language_model.layers.1.ple.ple_embedding.ngram_embedding": {
563
+ "quant_algo": "FP8"
564
+ },
565
+ "model.language_model.layers.10.mlp.experts": {
566
+ "quant_algo": "NVFP4",
567
+ "group_size": 16
568
+ },
569
+ "model.language_model.layers.11.mlp.experts": {
570
+ "quant_algo": "NVFP4",
571
+ "group_size": 16
572
+ },
573
+ "model.language_model.layers.12.mlp.experts": {
574
+ "quant_algo": "NVFP4",
575
+ "group_size": 16
576
+ },
577
+ "model.language_model.layers.13.mlp.experts": {
578
+ "quant_algo": "NVFP4",
579
+ "group_size": 16
580
+ },
581
+ "model.language_model.layers.14.mlp.experts": {
582
+ "quant_algo": "NVFP4",
583
+ "group_size": 16
584
+ },
585
+ "model.language_model.layers.15.mlp.experts": {
586
+ "quant_algo": "NVFP4",
587
+ "group_size": 16
588
+ },
589
+ "model.language_model.layers.16.mlp.experts": {
590
+ "quant_algo": "NVFP4",
591
+ "group_size": 16
592
+ },
593
+ "model.language_model.layers.17.mlp.experts": {
594
+ "quant_algo": "NVFP4",
595
+ "group_size": 16
596
+ },
597
+ "model.language_model.layers.18.mlp.experts": {
598
+ "quant_algo": "NVFP4",
599
+ "group_size": 16
600
+ },
601
+ "model.language_model.layers.19.mlp.experts": {
602
+ "quant_algo": "NVFP4",
603
+ "group_size": 16
604
+ },
605
+ "model.language_model.layers.2.mlp.experts": {
606
+ "quant_algo": "NVFP4",
607
+ "group_size": 16
608
+ },
609
+ "model.language_model.layers.20.mlp.experts": {
610
+ "quant_algo": "NVFP4",
611
+ "group_size": 16
612
+ },
613
+ "model.language_model.layers.21.mlp.experts": {
614
+ "quant_algo": "NVFP4",
615
+ "group_size": 16
616
+ },
617
+ "model.language_model.layers.22.mlp.experts": {
618
+ "quant_algo": "NVFP4",
619
+ "group_size": 16
620
+ },
621
+ "model.language_model.layers.23.mlp.experts": {
622
+ "quant_algo": "NVFP4",
623
+ "group_size": 16
624
+ },
625
+ "model.language_model.layers.24.mlp.experts": {
626
+ "quant_algo": "NVFP4",
627
+ "group_size": 16
628
+ },
629
+ "model.language_model.layers.25.mlp.experts": {
630
+ "quant_algo": "NVFP4",
631
+ "group_size": 16
632
+ },
633
+ "model.language_model.layers.26.mlp.experts": {
634
+ "quant_algo": "NVFP4",
635
+ "group_size": 16
636
+ },
637
+ "model.language_model.layers.27.mlp.experts": {
638
+ "quant_algo": "NVFP4",
639
+ "group_size": 16
640
+ },
641
+ "model.language_model.layers.28.mlp.experts": {
642
+ "quant_algo": "NVFP4",
643
+ "group_size": 16
644
+ },
645
+ "model.language_model.layers.29.mlp.experts": {
646
+ "quant_algo": "NVFP4",
647
+ "group_size": 16
648
+ },
649
+ "model.language_model.layers.3.mlp.experts": {
650
+ "quant_algo": "NVFP4",
651
+ "group_size": 16
652
+ },
653
+ "model.language_model.layers.30.mlp.experts": {
654
+ "quant_algo": "NVFP4",
655
+ "group_size": 16
656
+ },
657
+ "model.language_model.layers.31.mlp.experts": {
658
+ "quant_algo": "NVFP4",
659
+ "group_size": 16
660
+ },
661
+ "model.language_model.layers.32.mlp.experts": {
662
+ "quant_algo": "NVFP4",
663
+ "group_size": 16
664
+ },
665
+ "model.language_model.layers.33.mlp.experts": {
666
+ "quant_algo": "NVFP4",
667
+ "group_size": 16
668
+ },
669
+ "model.language_model.layers.34.mlp.experts": {
670
+ "quant_algo": "NVFP4",
671
+ "group_size": 16
672
+ },
673
+ "model.language_model.layers.35.mlp.experts": {
674
+ "quant_algo": "NVFP4",
675
+ "group_size": 16
676
+ },
677
+ "model.language_model.layers.36.mlp.experts": {
678
+ "quant_algo": "NVFP4",
679
+ "group_size": 16
680
+ },
681
+ "model.language_model.layers.37.mlp.experts": {
682
+ "quant_algo": "NVFP4",
683
+ "group_size": 16
684
+ },
685
+ "model.language_model.layers.38.mlp.experts": {
686
+ "quant_algo": "NVFP4",
687
+ "group_size": 16
688
+ },
689
+ "model.language_model.layers.39.mlp.experts": {
690
+ "quant_algo": "NVFP4",
691
+ "group_size": 16
692
+ },
693
+ "model.language_model.layers.4.mlp.experts": {
694
+ "quant_algo": "NVFP4",
695
+ "group_size": 16
696
+ },
697
+ "model.language_model.layers.40.mlp.experts": {
698
+ "quant_algo": "NVFP4",
699
+ "group_size": 16
700
+ },
701
+ "model.language_model.layers.41.mlp.experts": {
702
+ "quant_algo": "NVFP4",
703
+ "group_size": 16
704
+ },
705
+ "model.language_model.layers.42.mlp.experts": {
706
+ "quant_algo": "NVFP4",
707
+ "group_size": 16
708
+ },
709
+ "model.language_model.layers.43.mlp.experts": {
710
+ "quant_algo": "NVFP4",
711
+ "group_size": 16
712
+ },
713
+ "model.language_model.layers.44.mlp.experts": {
714
+ "quant_algo": "NVFP4",
715
+ "group_size": 16
716
+ },
717
+ "model.language_model.layers.45.mlp.experts": {
718
+ "quant_algo": "NVFP4",
719
+ "group_size": 16
720
+ },
721
+ "model.language_model.layers.46.mlp.experts": {
722
+ "quant_algo": "NVFP4",
723
+ "group_size": 16
724
+ },
725
+ "model.language_model.layers.47.mlp.experts": {
726
+ "quant_algo": "NVFP4",
727
+ "group_size": 16
728
+ },
729
+ "model.language_model.layers.5.mlp.experts": {
730
+ "quant_algo": "NVFP4",
731
+ "group_size": 16
732
+ },
733
+ "model.language_model.layers.6.mlp.experts": {
734
+ "quant_algo": "NVFP4",
735
+ "group_size": 16
736
+ },
737
+ "model.language_model.layers.7.mlp.experts": {
738
+ "quant_algo": "NVFP4",
739
+ "group_size": 16
740
+ },
741
+ "model.language_model.layers.8.mlp.experts": {
742
+ "quant_algo": "NVFP4",
743
+ "group_size": 16
744
+ },
745
+ "model.language_model.layers.9.mlp.experts": {
746
+ "quant_algo": "NVFP4",
747
+ "group_size": 16
748
+ },
749
+ "mtp.layers.0.mlp.experts": {
750
+ "quant_algo": "FP8_BLOCK_SCALES",
751
+ "group_size": 128
752
+ }
753
+ }
754
+ }
755
+ }
generation_config.json ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 248044,
3
+ "do_sample": true,
4
+ "eos_token_id": [
5
+ 248046,
6
+ 248044
7
+ ],
8
+ "pad_token_id": 248044,
9
+ "temperature": 1.0,
10
+ "top_k": 20,
11
+ "top_p": 0.95
12
+ }
hf_quant_config.json ADDED
@@ -0,0 +1,506 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "producer": {
3
+ "name": "modelopt",
4
+ "version": "0.46.0.dev281+g73d778422"
5
+ },
6
+ "quantization": {
7
+ "quant_algo": "MIXED_PRECISION",
8
+ "kv_cache_quant_algo": null,
9
+ "group_size": 16,
10
+ "exclude_modules": [
11
+ "lm_head",
12
+ "model.language_model.embed_tokens",
13
+ "model.language_model.hyper_connection_mixer*",
14
+ "model.language_model.layers.0.attn_hyper_connection*",
15
+ "model.language_model.layers.0.linear_attn*",
16
+ "model.language_model.layers.0.mlp.gate",
17
+ "model.language_model.layers.0.mlp.shared_expert*",
18
+ "model.language_model.layers.0.mlp.shared_expert_gate",
19
+ "model.language_model.layers.0.mlp_hyper_connection*",
20
+ "model.language_model.layers.1.attn_hyper_connection*",
21
+ "model.language_model.layers.1.linear_attn*",
22
+ "model.language_model.layers.1.mlp.gate",
23
+ "model.language_model.layers.1.mlp.shared_expert*",
24
+ "model.language_model.layers.1.mlp.shared_expert_gate",
25
+ "model.language_model.layers.1.mlp_hyper_connection*",
26
+ "model.language_model.layers.10.attn_hyper_connection*",
27
+ "model.language_model.layers.10.linear_attn*",
28
+ "model.language_model.layers.10.mlp.gate",
29
+ "model.language_model.layers.10.mlp.shared_expert*",
30
+ "model.language_model.layers.10.mlp.shared_expert_gate",
31
+ "model.language_model.layers.10.mlp_hyper_connection*",
32
+ "model.language_model.layers.11.attn_hyper_connection*",
33
+ "model.language_model.layers.11.mlp.gate",
34
+ "model.language_model.layers.11.mlp.shared_expert*",
35
+ "model.language_model.layers.11.mlp.shared_expert_gate",
36
+ "model.language_model.layers.11.mlp_hyper_connection*",
37
+ "model.language_model.layers.11.self_attn*",
38
+ "model.language_model.layers.12.attn_hyper_connection*",
39
+ "model.language_model.layers.12.linear_attn*",
40
+ "model.language_model.layers.12.mlp.gate",
41
+ "model.language_model.layers.12.mlp.shared_expert*",
42
+ "model.language_model.layers.12.mlp.shared_expert_gate",
43
+ "model.language_model.layers.12.mlp_hyper_connection*",
44
+ "model.language_model.layers.13.attn_hyper_connection*",
45
+ "model.language_model.layers.13.linear_attn*",
46
+ "model.language_model.layers.13.mlp.gate",
47
+ "model.language_model.layers.13.mlp.shared_expert*",
48
+ "model.language_model.layers.13.mlp.shared_expert_gate",
49
+ "model.language_model.layers.13.mlp_hyper_connection*",
50
+ "model.language_model.layers.14.attn_hyper_connection*",
51
+ "model.language_model.layers.14.linear_attn*",
52
+ "model.language_model.layers.14.mlp.gate",
53
+ "model.language_model.layers.14.mlp.shared_expert*",
54
+ "model.language_model.layers.14.mlp.shared_expert_gate",
55
+ "model.language_model.layers.14.mlp_hyper_connection*",
56
+ "model.language_model.layers.15.attn_hyper_connection*",
57
+ "model.language_model.layers.15.mlp.gate",
58
+ "model.language_model.layers.15.mlp.shared_expert*",
59
+ "model.language_model.layers.15.mlp.shared_expert_gate",
60
+ "model.language_model.layers.15.mlp_hyper_connection*",
61
+ "model.language_model.layers.15.self_attn*",
62
+ "model.language_model.layers.16.attn_hyper_connection*",
63
+ "model.language_model.layers.16.linear_attn*",
64
+ "model.language_model.layers.16.mlp.gate",
65
+ "model.language_model.layers.16.mlp.shared_expert*",
66
+ "model.language_model.layers.16.mlp.shared_expert_gate",
67
+ "model.language_model.layers.16.mlp_hyper_connection*",
68
+ "model.language_model.layers.17.attn_hyper_connection*",
69
+ "model.language_model.layers.17.linear_attn*",
70
+ "model.language_model.layers.17.mlp.gate",
71
+ "model.language_model.layers.17.mlp.shared_expert*",
72
+ "model.language_model.layers.17.mlp.shared_expert_gate",
73
+ "model.language_model.layers.17.mlp_hyper_connection*",
74
+ "model.language_model.layers.18.attn_hyper_connection*",
75
+ "model.language_model.layers.18.linear_attn*",
76
+ "model.language_model.layers.18.mlp.gate",
77
+ "model.language_model.layers.18.mlp.shared_expert*",
78
+ "model.language_model.layers.18.mlp.shared_expert_gate",
79
+ "model.language_model.layers.18.mlp_hyper_connection*",
80
+ "model.language_model.layers.19.attn_hyper_connection*",
81
+ "model.language_model.layers.19.mlp.gate",
82
+ "model.language_model.layers.19.mlp.shared_expert*",
83
+ "model.language_model.layers.19.mlp.shared_expert_gate",
84
+ "model.language_model.layers.19.mlp_hyper_connection*",
85
+ "model.language_model.layers.19.self_attn*",
86
+ "model.language_model.layers.2.attn_hyper_connection*",
87
+ "model.language_model.layers.2.linear_attn*",
88
+ "model.language_model.layers.2.mlp.gate",
89
+ "model.language_model.layers.2.mlp.shared_expert*",
90
+ "model.language_model.layers.2.mlp.shared_expert_gate",
91
+ "model.language_model.layers.2.mlp_hyper_connection*",
92
+ "model.language_model.layers.20.attn_hyper_connection*",
93
+ "model.language_model.layers.20.linear_attn*",
94
+ "model.language_model.layers.20.mlp.gate",
95
+ "model.language_model.layers.20.mlp.shared_expert*",
96
+ "model.language_model.layers.20.mlp.shared_expert_gate",
97
+ "model.language_model.layers.20.mlp_hyper_connection*",
98
+ "model.language_model.layers.21.attn_hyper_connection*",
99
+ "model.language_model.layers.21.linear_attn*",
100
+ "model.language_model.layers.21.mlp.gate",
101
+ "model.language_model.layers.21.mlp.shared_expert*",
102
+ "model.language_model.layers.21.mlp.shared_expert_gate",
103
+ "model.language_model.layers.21.mlp_hyper_connection*",
104
+ "model.language_model.layers.22.attn_hyper_connection*",
105
+ "model.language_model.layers.22.linear_attn*",
106
+ "model.language_model.layers.22.mlp.gate",
107
+ "model.language_model.layers.22.mlp.shared_expert*",
108
+ "model.language_model.layers.22.mlp.shared_expert_gate",
109
+ "model.language_model.layers.22.mlp_hyper_connection*",
110
+ "model.language_model.layers.23.attn_hyper_connection*",
111
+ "model.language_model.layers.23.mlp.gate",
112
+ "model.language_model.layers.23.mlp.shared_expert*",
113
+ "model.language_model.layers.23.mlp.shared_expert_gate",
114
+ "model.language_model.layers.23.mlp_hyper_connection*",
115
+ "model.language_model.layers.23.self_attn*",
116
+ "model.language_model.layers.24.attn_hyper_connection*",
117
+ "model.language_model.layers.24.linear_attn*",
118
+ "model.language_model.layers.24.mlp.gate",
119
+ "model.language_model.layers.24.mlp.shared_expert*",
120
+ "model.language_model.layers.24.mlp.shared_expert_gate",
121
+ "model.language_model.layers.24.mlp_hyper_connection*",
122
+ "model.language_model.layers.25.attn_hyper_connection*",
123
+ "model.language_model.layers.25.linear_attn*",
124
+ "model.language_model.layers.25.mlp.gate",
125
+ "model.language_model.layers.25.mlp.shared_expert*",
126
+ "model.language_model.layers.25.mlp.shared_expert_gate",
127
+ "model.language_model.layers.25.mlp_hyper_connection*",
128
+ "model.language_model.layers.26.attn_hyper_connection*",
129
+ "model.language_model.layers.26.linear_attn*",
130
+ "model.language_model.layers.26.mlp.gate",
131
+ "model.language_model.layers.26.mlp.shared_expert*",
132
+ "model.language_model.layers.26.mlp.shared_expert_gate",
133
+ "model.language_model.layers.26.mlp_hyper_connection*",
134
+ "model.language_model.layers.27.attn_hyper_connection*",
135
+ "model.language_model.layers.27.mlp.gate",
136
+ "model.language_model.layers.27.mlp.shared_expert*",
137
+ "model.language_model.layers.27.mlp.shared_expert_gate",
138
+ "model.language_model.layers.27.mlp_hyper_connection*",
139
+ "model.language_model.layers.27.self_attn*",
140
+ "model.language_model.layers.28.attn_hyper_connection*",
141
+ "model.language_model.layers.28.linear_attn*",
142
+ "model.language_model.layers.28.mlp.gate",
143
+ "model.language_model.layers.28.mlp.shared_expert*",
144
+ "model.language_model.layers.28.mlp.shared_expert_gate",
145
+ "model.language_model.layers.28.mlp_hyper_connection*",
146
+ "model.language_model.layers.29.attn_hyper_connection*",
147
+ "model.language_model.layers.29.linear_attn*",
148
+ "model.language_model.layers.29.mlp.gate",
149
+ "model.language_model.layers.29.mlp.shared_expert*",
150
+ "model.language_model.layers.29.mlp.shared_expert_gate",
151
+ "model.language_model.layers.29.mlp_hyper_connection*",
152
+ "model.language_model.layers.3.attn_hyper_connection*",
153
+ "model.language_model.layers.3.mlp.gate",
154
+ "model.language_model.layers.3.mlp.shared_expert*",
155
+ "model.language_model.layers.3.mlp.shared_expert_gate",
156
+ "model.language_model.layers.3.mlp_hyper_connection*",
157
+ "model.language_model.layers.3.self_attn*",
158
+ "model.language_model.layers.30.attn_hyper_connection*",
159
+ "model.language_model.layers.30.linear_attn*",
160
+ "model.language_model.layers.30.mlp.gate",
161
+ "model.language_model.layers.30.mlp.shared_expert*",
162
+ "model.language_model.layers.30.mlp.shared_expert_gate",
163
+ "model.language_model.layers.30.mlp_hyper_connection*",
164
+ "model.language_model.layers.31.attn_hyper_connection*",
165
+ "model.language_model.layers.31.mlp.gate",
166
+ "model.language_model.layers.31.mlp.shared_expert*",
167
+ "model.language_model.layers.31.mlp.shared_expert_gate",
168
+ "model.language_model.layers.31.mlp_hyper_connection*",
169
+ "model.language_model.layers.31.self_attn*",
170
+ "model.language_model.layers.32.attn_hyper_connection*",
171
+ "model.language_model.layers.32.linear_attn*",
172
+ "model.language_model.layers.32.mlp.gate",
173
+ "model.language_model.layers.32.mlp.shared_expert*",
174
+ "model.language_model.layers.32.mlp.shared_expert_gate",
175
+ "model.language_model.layers.32.mlp_hyper_connection*",
176
+ "model.language_model.layers.33.attn_hyper_connection*",
177
+ "model.language_model.layers.33.linear_attn*",
178
+ "model.language_model.layers.33.mlp.gate",
179
+ "model.language_model.layers.33.mlp.shared_expert*",
180
+ "model.language_model.layers.33.mlp.shared_expert_gate",
181
+ "model.language_model.layers.33.mlp_hyper_connection*",
182
+ "model.language_model.layers.34.attn_hyper_connection*",
183
+ "model.language_model.layers.34.linear_attn*",
184
+ "model.language_model.layers.34.mlp.gate",
185
+ "model.language_model.layers.34.mlp.shared_expert*",
186
+ "model.language_model.layers.34.mlp.shared_expert_gate",
187
+ "model.language_model.layers.34.mlp_hyper_connection*",
188
+ "model.language_model.layers.35.attn_hyper_connection*",
189
+ "model.language_model.layers.35.mlp.gate",
190
+ "model.language_model.layers.35.mlp.shared_expert*",
191
+ "model.language_model.layers.35.mlp.shared_expert_gate",
192
+ "model.language_model.layers.35.mlp_hyper_connection*",
193
+ "model.language_model.layers.35.self_attn*",
194
+ "model.language_model.layers.36.attn_hyper_connection*",
195
+ "model.language_model.layers.36.linear_attn*",
196
+ "model.language_model.layers.36.mlp.gate",
197
+ "model.language_model.layers.36.mlp.shared_expert*",
198
+ "model.language_model.layers.36.mlp.shared_expert_gate",
199
+ "model.language_model.layers.36.mlp_hyper_connection*",
200
+ "model.language_model.layers.37.attn_hyper_connection*",
201
+ "model.language_model.layers.37.linear_attn*",
202
+ "model.language_model.layers.37.mlp.gate",
203
+ "model.language_model.layers.37.mlp.shared_expert*",
204
+ "model.language_model.layers.37.mlp.shared_expert_gate",
205
+ "model.language_model.layers.37.mlp_hyper_connection*",
206
+ "model.language_model.layers.38.attn_hyper_connection*",
207
+ "model.language_model.layers.38.linear_attn*",
208
+ "model.language_model.layers.38.mlp.gate",
209
+ "model.language_model.layers.38.mlp.shared_expert*",
210
+ "model.language_model.layers.38.mlp.shared_expert_gate",
211
+ "model.language_model.layers.38.mlp_hyper_connection*",
212
+ "model.language_model.layers.39.attn_hyper_connection*",
213
+ "model.language_model.layers.39.mlp.gate",
214
+ "model.language_model.layers.39.mlp.shared_expert*",
215
+ "model.language_model.layers.39.mlp.shared_expert_gate",
216
+ "model.language_model.layers.39.mlp_hyper_connection*",
217
+ "model.language_model.layers.39.self_attn*",
218
+ "model.language_model.layers.4.attn_hyper_connection*",
219
+ "model.language_model.layers.4.linear_attn*",
220
+ "model.language_model.layers.4.mlp.gate",
221
+ "model.language_model.layers.4.mlp.shared_expert*",
222
+ "model.language_model.layers.4.mlp.shared_expert_gate",
223
+ "model.language_model.layers.4.mlp_hyper_connection*",
224
+ "model.language_model.layers.40.attn_hyper_connection*",
225
+ "model.language_model.layers.40.linear_attn*",
226
+ "model.language_model.layers.40.mlp.gate",
227
+ "model.language_model.layers.40.mlp.shared_expert*",
228
+ "model.language_model.layers.40.mlp.shared_expert_gate",
229
+ "model.language_model.layers.40.mlp_hyper_connection*",
230
+ "model.language_model.layers.41.attn_hyper_connection*",
231
+ "model.language_model.layers.41.linear_attn*",
232
+ "model.language_model.layers.41.mlp.gate",
233
+ "model.language_model.layers.41.mlp.shared_expert*",
234
+ "model.language_model.layers.41.mlp.shared_expert_gate",
235
+ "model.language_model.layers.41.mlp_hyper_connection*",
236
+ "model.language_model.layers.42.attn_hyper_connection*",
237
+ "model.language_model.layers.42.linear_attn*",
238
+ "model.language_model.layers.42.mlp.gate",
239
+ "model.language_model.layers.42.mlp.shared_expert*",
240
+ "model.language_model.layers.42.mlp.shared_expert_gate",
241
+ "model.language_model.layers.42.mlp_hyper_connection*",
242
+ "model.language_model.layers.43.attn_hyper_connection*",
243
+ "model.language_model.layers.43.mlp.gate",
244
+ "model.language_model.layers.43.mlp.shared_expert*",
245
+ "model.language_model.layers.43.mlp.shared_expert_gate",
246
+ "model.language_model.layers.43.mlp_hyper_connection*",
247
+ "model.language_model.layers.43.self_attn*",
248
+ "model.language_model.layers.44.attn_hyper_connection*",
249
+ "model.language_model.layers.44.linear_attn*",
250
+ "model.language_model.layers.44.mlp.gate",
251
+ "model.language_model.layers.44.mlp.shared_expert*",
252
+ "model.language_model.layers.44.mlp.shared_expert_gate",
253
+ "model.language_model.layers.44.mlp_hyper_connection*",
254
+ "model.language_model.layers.45.attn_hyper_connection*",
255
+ "model.language_model.layers.45.linear_attn*",
256
+ "model.language_model.layers.45.mlp.gate",
257
+ "model.language_model.layers.45.mlp.shared_expert*",
258
+ "model.language_model.layers.45.mlp.shared_expert_gate",
259
+ "model.language_model.layers.45.mlp_hyper_connection*",
260
+ "model.language_model.layers.46.attn_hyper_connection*",
261
+ "model.language_model.layers.46.linear_attn*",
262
+ "model.language_model.layers.46.mlp.gate",
263
+ "model.language_model.layers.46.mlp.shared_expert*",
264
+ "model.language_model.layers.46.mlp.shared_expert_gate",
265
+ "model.language_model.layers.46.mlp_hyper_connection*",
266
+ "model.language_model.layers.47.attn_hyper_connection*",
267
+ "model.language_model.layers.47.mlp.gate",
268
+ "model.language_model.layers.47.mlp.shared_expert*",
269
+ "model.language_model.layers.47.mlp.shared_expert_gate",
270
+ "model.language_model.layers.47.mlp_hyper_connection*",
271
+ "model.language_model.layers.47.self_attn*",
272
+ "model.language_model.layers.5.attn_hyper_connection*",
273
+ "model.language_model.layers.5.linear_attn*",
274
+ "model.language_model.layers.5.mlp.gate",
275
+ "model.language_model.layers.5.mlp.shared_expert*",
276
+ "model.language_model.layers.5.mlp.shared_expert_gate",
277
+ "model.language_model.layers.5.mlp_hyper_connection*",
278
+ "model.language_model.layers.6.attn_hyper_connection*",
279
+ "model.language_model.layers.6.linear_attn*",
280
+ "model.language_model.layers.6.mlp.gate",
281
+ "model.language_model.layers.6.mlp.shared_expert*",
282
+ "model.language_model.layers.6.mlp.shared_expert_gate",
283
+ "model.language_model.layers.6.mlp_hyper_connection*",
284
+ "model.language_model.layers.7.attn_hyper_connection*",
285
+ "model.language_model.layers.7.mlp.gate",
286
+ "model.language_model.layers.7.mlp.shared_expert*",
287
+ "model.language_model.layers.7.mlp.shared_expert_gate",
288
+ "model.language_model.layers.7.mlp_hyper_connection*",
289
+ "model.language_model.layers.7.self_attn*",
290
+ "model.language_model.layers.8.attn_hyper_connection*",
291
+ "model.language_model.layers.8.linear_attn*",
292
+ "model.language_model.layers.8.mlp.gate",
293
+ "model.language_model.layers.8.mlp.shared_expert*",
294
+ "model.language_model.layers.8.mlp.shared_expert_gate",
295
+ "model.language_model.layers.8.mlp_hyper_connection*",
296
+ "model.language_model.layers.9.attn_hyper_connection*",
297
+ "model.language_model.layers.9.linear_attn*",
298
+ "model.language_model.layers.9.mlp.gate",
299
+ "model.language_model.layers.9.mlp.shared_expert*",
300
+ "model.language_model.layers.9.mlp.shared_expert_gate",
301
+ "model.language_model.layers.9.mlp_hyper_connection*",
302
+ "model.visual*"
303
+ ],
304
+ "quantized_layers": {
305
+ "model.language_model.layers.0.mlp.experts": {
306
+ "quant_algo": "NVFP4",
307
+ "group_size": 16
308
+ },
309
+ "model.language_model.layers.1.mlp.experts": {
310
+ "quant_algo": "NVFP4",
311
+ "group_size": 16
312
+ },
313
+ "model.language_model.layers.1.ple.ple_embedding.ngram_embedding": {
314
+ "quant_algo": "FP8"
315
+ },
316
+ "model.language_model.layers.10.mlp.experts": {
317
+ "quant_algo": "NVFP4",
318
+ "group_size": 16
319
+ },
320
+ "model.language_model.layers.11.mlp.experts": {
321
+ "quant_algo": "NVFP4",
322
+ "group_size": 16
323
+ },
324
+ "model.language_model.layers.12.mlp.experts": {
325
+ "quant_algo": "NVFP4",
326
+ "group_size": 16
327
+ },
328
+ "model.language_model.layers.13.mlp.experts": {
329
+ "quant_algo": "NVFP4",
330
+ "group_size": 16
331
+ },
332
+ "model.language_model.layers.14.mlp.experts": {
333
+ "quant_algo": "NVFP4",
334
+ "group_size": 16
335
+ },
336
+ "model.language_model.layers.15.mlp.experts": {
337
+ "quant_algo": "NVFP4",
338
+ "group_size": 16
339
+ },
340
+ "model.language_model.layers.16.mlp.experts": {
341
+ "quant_algo": "NVFP4",
342
+ "group_size": 16
343
+ },
344
+ "model.language_model.layers.17.mlp.experts": {
345
+ "quant_algo": "NVFP4",
346
+ "group_size": 16
347
+ },
348
+ "model.language_model.layers.18.mlp.experts": {
349
+ "quant_algo": "NVFP4",
350
+ "group_size": 16
351
+ },
352
+ "model.language_model.layers.19.mlp.experts": {
353
+ "quant_algo": "NVFP4",
354
+ "group_size": 16
355
+ },
356
+ "model.language_model.layers.2.mlp.experts": {
357
+ "quant_algo": "NVFP4",
358
+ "group_size": 16
359
+ },
360
+ "model.language_model.layers.20.mlp.experts": {
361
+ "quant_algo": "NVFP4",
362
+ "group_size": 16
363
+ },
364
+ "model.language_model.layers.21.mlp.experts": {
365
+ "quant_algo": "NVFP4",
366
+ "group_size": 16
367
+ },
368
+ "model.language_model.layers.22.mlp.experts": {
369
+ "quant_algo": "NVFP4",
370
+ "group_size": 16
371
+ },
372
+ "model.language_model.layers.23.mlp.experts": {
373
+ "quant_algo": "NVFP4",
374
+ "group_size": 16
375
+ },
376
+ "model.language_model.layers.24.mlp.experts": {
377
+ "quant_algo": "NVFP4",
378
+ "group_size": 16
379
+ },
380
+ "model.language_model.layers.25.mlp.experts": {
381
+ "quant_algo": "NVFP4",
382
+ "group_size": 16
383
+ },
384
+ "model.language_model.layers.26.mlp.experts": {
385
+ "quant_algo": "NVFP4",
386
+ "group_size": 16
387
+ },
388
+ "model.language_model.layers.27.mlp.experts": {
389
+ "quant_algo": "NVFP4",
390
+ "group_size": 16
391
+ },
392
+ "model.language_model.layers.28.mlp.experts": {
393
+ "quant_algo": "NVFP4",
394
+ "group_size": 16
395
+ },
396
+ "model.language_model.layers.29.mlp.experts": {
397
+ "quant_algo": "NVFP4",
398
+ "group_size": 16
399
+ },
400
+ "model.language_model.layers.3.mlp.experts": {
401
+ "quant_algo": "NVFP4",
402
+ "group_size": 16
403
+ },
404
+ "model.language_model.layers.30.mlp.experts": {
405
+ "quant_algo": "NVFP4",
406
+ "group_size": 16
407
+ },
408
+ "model.language_model.layers.31.mlp.experts": {
409
+ "quant_algo": "NVFP4",
410
+ "group_size": 16
411
+ },
412
+ "model.language_model.layers.32.mlp.experts": {
413
+ "quant_algo": "NVFP4",
414
+ "group_size": 16
415
+ },
416
+ "model.language_model.layers.33.mlp.experts": {
417
+ "quant_algo": "NVFP4",
418
+ "group_size": 16
419
+ },
420
+ "model.language_model.layers.34.mlp.experts": {
421
+ "quant_algo": "NVFP4",
422
+ "group_size": 16
423
+ },
424
+ "model.language_model.layers.35.mlp.experts": {
425
+ "quant_algo": "NVFP4",
426
+ "group_size": 16
427
+ },
428
+ "model.language_model.layers.36.mlp.experts": {
429
+ "quant_algo": "NVFP4",
430
+ "group_size": 16
431
+ },
432
+ "model.language_model.layers.37.mlp.experts": {
433
+ "quant_algo": "NVFP4",
434
+ "group_size": 16
435
+ },
436
+ "model.language_model.layers.38.mlp.experts": {
437
+ "quant_algo": "NVFP4",
438
+ "group_size": 16
439
+ },
440
+ "model.language_model.layers.39.mlp.experts": {
441
+ "quant_algo": "NVFP4",
442
+ "group_size": 16
443
+ },
444
+ "model.language_model.layers.4.mlp.experts": {
445
+ "quant_algo": "NVFP4",
446
+ "group_size": 16
447
+ },
448
+ "model.language_model.layers.40.mlp.experts": {
449
+ "quant_algo": "NVFP4",
450
+ "group_size": 16
451
+ },
452
+ "model.language_model.layers.41.mlp.experts": {
453
+ "quant_algo": "NVFP4",
454
+ "group_size": 16
455
+ },
456
+ "model.language_model.layers.42.mlp.experts": {
457
+ "quant_algo": "NVFP4",
458
+ "group_size": 16
459
+ },
460
+ "model.language_model.layers.43.mlp.experts": {
461
+ "quant_algo": "NVFP4",
462
+ "group_size": 16
463
+ },
464
+ "model.language_model.layers.44.mlp.experts": {
465
+ "quant_algo": "NVFP4",
466
+ "group_size": 16
467
+ },
468
+ "model.language_model.layers.45.mlp.experts": {
469
+ "quant_algo": "NVFP4",
470
+ "group_size": 16
471
+ },
472
+ "model.language_model.layers.46.mlp.experts": {
473
+ "quant_algo": "NVFP4",
474
+ "group_size": 16
475
+ },
476
+ "model.language_model.layers.47.mlp.experts": {
477
+ "quant_algo": "NVFP4",
478
+ "group_size": 16
479
+ },
480
+ "model.language_model.layers.5.mlp.experts": {
481
+ "quant_algo": "NVFP4",
482
+ "group_size": 16
483
+ },
484
+ "model.language_model.layers.6.mlp.experts": {
485
+ "quant_algo": "NVFP4",
486
+ "group_size": 16
487
+ },
488
+ "model.language_model.layers.7.mlp.experts": {
489
+ "quant_algo": "NVFP4",
490
+ "group_size": 16
491
+ },
492
+ "model.language_model.layers.8.mlp.experts": {
493
+ "quant_algo": "NVFP4",
494
+ "group_size": 16
495
+ },
496
+ "model.language_model.layers.9.mlp.experts": {
497
+ "quant_algo": "NVFP4",
498
+ "group_size": 16
499
+ },
500
+ "mtp.layers.0.mlp.experts": {
501
+ "quant_algo": "FP8_BLOCK_SCALES",
502
+ "group_size": 128
503
+ }
504
+ }
505
+ }
506
+ }
merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
model-00001-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:63fde954be6f08b49b876f4f70a0ad0bcfee71aff7b1faa33779b6b32feca2a2
3
+ size 3115991696
model-00002-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4dafaef62a908e49e0d92c7c2a3fa99f4d9651fdeeb0ca09f09a9b40095af4e0
3
+ size 10005510304
model-00003-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3218ddc129258e91a721a8e81329ca8f00588332ec721bd8d2cf7d20e324bf47
3
+ size 10005364728
model-00004-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:55e2bdf6a3a1f6e65270787f65c63b2a2a56b308b54eb1fb318dfabc9b48b141
3
+ size 10006060112
model-00005-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d3c169d3694bfba846455d01e48f047d770f78d071c5a4d22f129389411d686e
3
+ size 10005362392
model-00006-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:38f65a9d11090428e139cc60d0587e8e5433ea2ce79820883c583d7cfcc53130
3
+ size 10005375328
model-00007-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8cf3fa05c04cb2e060963b69bb01f7a7b9f43c58435fbe380836a0457a19cd3e
3
+ size 10005956544
model-00008-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:53d1f80746aa0cc0a7c2f4836587002a8a4dd72c827432cd2e43abae3c3899e4
3
+ size 10005375200
model-00009-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eaff6a87ece6fbfd6ad0fbaa0a5cf9ff542f8076acba38e62410b1ed55e12cc3
3
+ size 5679113808
model-00010-of-00010.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0d49b0cf7c15bf9b5bd1e17403b36727315860a8f2a2e0f5a53655547084f640
3
+ size 128587536
model-fp8-mtp-ple.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3525520c8602d850003eb1960aec0b64291dae33b83f8b65dd639a451df78823
3
+ size 53717551730
model.safetensors.index.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:660414e8300728ba80062e6a3c81cb76a65e2ab136ac9e9ceb450ba0c51c3c0d
3
+ size 31275518
preprocessor_config.json ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "size": {
3
+ "longest_edge": 16777216,
4
+ "shortest_edge": 65536
5
+ },
6
+ "patch_size": 16,
7
+ "temporal_patch_size": 2,
8
+ "merge_size": 2,
9
+ "image_mean": [
10
+ 0.5,
11
+ 0.5,
12
+ 0.5
13
+ ],
14
+ "image_std": [
15
+ 0.5,
16
+ 0.5,
17
+ 0.5
18
+ ],
19
+ "processor_class": "Qwen3VLProcessor",
20
+ "image_processor_type": "Qwen2VLImageProcessorFast"
21
+ }
processor_config.json ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "image_processor": {
3
+ "do_convert_rgb": true,
4
+ "do_normalize": true,
5
+ "do_rescale": true,
6
+ "do_resize": true,
7
+ "image_mean": [
8
+ 0.5,
9
+ 0.5,
10
+ 0.5
11
+ ],
12
+ "image_processor_type": "Qwen2VLImageProcessor",
13
+ "image_std": [
14
+ 0.5,
15
+ 0.5,
16
+ 0.5
17
+ ],
18
+ "merge_size": 2,
19
+ "patch_size": 16,
20
+ "resample": 3,
21
+ "rescale_factor": 0.00392156862745098,
22
+ "size": {
23
+ "longest_edge": 16777216,
24
+ "shortest_edge": 65536
25
+ },
26
+ "temporal_patch_size": 2
27
+ },
28
+ "processor_class": "Qwen3VLProcessor",
29
+ "video_processor": {
30
+ "do_convert_rgb": true,
31
+ "do_normalize": true,
32
+ "do_rescale": true,
33
+ "do_resize": true,
34
+ "do_sample_frames": true,
35
+ "fps": 2,
36
+ "image_mean": [
37
+ 0.5,
38
+ 0.5,
39
+ 0.5
40
+ ],
41
+ "image_std": [
42
+ 0.5,
43
+ 0.5,
44
+ 0.5
45
+ ],
46
+ "max_frames": 768,
47
+ "merge_size": 2,
48
+ "min_frames": 4,
49
+ "patch_size": 16,
50
+ "resample": 3,
51
+ "rescale_factor": 0.00392156862745098,
52
+ "return_metadata": false,
53
+ "size": {
54
+ "longest_edge": 25165824,
55
+ "shortest_edge": 4096
56
+ },
57
+ "temporal_patch_size": 2,
58
+ "video_processor_type": "Qwen3VLVideoProcessor"
59
+ }
60
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0997f410c57a1f4e53b09e4be8f4a172d90edd9564368fb0847030937229b9f3
3
+ size 12809320
tokenizer_config.json ADDED
@@ -0,0 +1,305 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "added_tokens_decoder": {
4
+ "248044": {
5
+ "content": "<|endoftext|>",
6
+ "lstrip": false,
7
+ "normalized": false,
8
+ "rstrip": false,
9
+ "single_word": false,
10
+ "special": true
11
+ },
12
+ "248045": {
13
+ "content": "<|im_start|>",
14
+ "lstrip": false,
15
+ "normalized": false,
16
+ "rstrip": false,
17
+ "single_word": false,
18
+ "special": true
19
+ },
20
+ "248046": {
21
+ "content": "<|im_end|>",
22
+ "lstrip": false,
23
+ "normalized": false,
24
+ "rstrip": false,
25
+ "single_word": false,
26
+ "special": true
27
+ },
28
+ "248047": {
29
+ "content": "<|object_ref_start|>",
30
+ "lstrip": false,
31
+ "normalized": false,
32
+ "rstrip": false,
33
+ "single_word": false,
34
+ "special": true
35
+ },
36
+ "248048": {
37
+ "content": "<|object_ref_end|>",
38
+ "lstrip": false,
39
+ "normalized": false,
40
+ "rstrip": false,
41
+ "single_word": false,
42
+ "special": true
43
+ },
44
+ "248049": {
45
+ "content": "<|box_start|>",
46
+ "lstrip": false,
47
+ "normalized": false,
48
+ "rstrip": false,
49
+ "single_word": false,
50
+ "special": true
51
+ },
52
+ "248050": {
53
+ "content": "<|box_end|>",
54
+ "lstrip": false,
55
+ "normalized": false,
56
+ "rstrip": false,
57
+ "single_word": false,
58
+ "special": true
59
+ },
60
+ "248051": {
61
+ "content": "<|quad_start|>",
62
+ "lstrip": false,
63
+ "normalized": false,
64
+ "rstrip": false,
65
+ "single_word": false,
66
+ "special": true
67
+ },
68
+ "248052": {
69
+ "content": "<|quad_end|>",
70
+ "lstrip": false,
71
+ "normalized": false,
72
+ "rstrip": false,
73
+ "single_word": false,
74
+ "special": true
75
+ },
76
+ "248053": {
77
+ "content": "<|vision_start|>",
78
+ "lstrip": false,
79
+ "normalized": false,
80
+ "rstrip": false,
81
+ "single_word": false,
82
+ "special": true
83
+ },
84
+ "248054": {
85
+ "content": "<|vision_end|>",
86
+ "lstrip": false,
87
+ "normalized": false,
88
+ "rstrip": false,
89
+ "single_word": false,
90
+ "special": true
91
+ },
92
+ "248055": {
93
+ "content": "<|vision_pad|>",
94
+ "lstrip": false,
95
+ "normalized": false,
96
+ "rstrip": false,
97
+ "single_word": false,
98
+ "special": true
99
+ },
100
+ "248056": {
101
+ "content": "<|image_pad|>",
102
+ "lstrip": false,
103
+ "normalized": false,
104
+ "rstrip": false,
105
+ "single_word": false,
106
+ "special": true
107
+ },
108
+ "248057": {
109
+ "content": "<|video_pad|>",
110
+ "lstrip": false,
111
+ "normalized": false,
112
+ "rstrip": false,
113
+ "single_word": false,
114
+ "special": true
115
+ },
116
+ "248058": {
117
+ "content": "<tool_call>",
118
+ "lstrip": false,
119
+ "normalized": false,
120
+ "rstrip": false,
121
+ "single_word": false,
122
+ "special": false
123
+ },
124
+ "248059": {
125
+ "content": "</tool_call>",
126
+ "lstrip": false,
127
+ "normalized": false,
128
+ "rstrip": false,
129
+ "single_word": false,
130
+ "special": false
131
+ },
132
+ "248060": {
133
+ "content": "<|fim_prefix|>",
134
+ "lstrip": false,
135
+ "normalized": false,
136
+ "rstrip": false,
137
+ "single_word": false,
138
+ "special": false
139
+ },
140
+ "248061": {
141
+ "content": "<|fim_middle|>",
142
+ "lstrip": false,
143
+ "normalized": false,
144
+ "rstrip": false,
145
+ "single_word": false,
146
+ "special": false
147
+ },
148
+ "248062": {
149
+ "content": "<|fim_suffix|>",
150
+ "lstrip": false,
151
+ "normalized": false,
152
+ "rstrip": false,
153
+ "single_word": false,
154
+ "special": false
155
+ },
156
+ "248063": {
157
+ "content": "<|fim_pad|>",
158
+ "lstrip": false,
159
+ "normalized": false,
160
+ "rstrip": false,
161
+ "single_word": false,
162
+ "special": false
163
+ },
164
+ "248064": {
165
+ "content": "<|repo_name|>",
166
+ "lstrip": false,
167
+ "normalized": false,
168
+ "rstrip": false,
169
+ "single_word": false,
170
+ "special": false
171
+ },
172
+ "248065": {
173
+ "content": "<|file_sep|>",
174
+ "lstrip": false,
175
+ "normalized": false,
176
+ "rstrip": false,
177
+ "single_word": false,
178
+ "special": false
179
+ },
180
+ "248066": {
181
+ "content": "<tool_response>",
182
+ "lstrip": false,
183
+ "normalized": false,
184
+ "rstrip": false,
185
+ "single_word": false,
186
+ "special": false
187
+ },
188
+ "248067": {
189
+ "content": "</tool_response>",
190
+ "lstrip": false,
191
+ "normalized": false,
192
+ "rstrip": false,
193
+ "single_word": false,
194
+ "special": false
195
+ },
196
+ "248068": {
197
+ "content": "<think>",
198
+ "lstrip": false,
199
+ "normalized": false,
200
+ "rstrip": false,
201
+ "single_word": false,
202
+ "special": false
203
+ },
204
+ "248069": {
205
+ "content": "</think>",
206
+ "lstrip": false,
207
+ "normalized": false,
208
+ "rstrip": false,
209
+ "single_word": false,
210
+ "special": false
211
+ },
212
+ "248070": {
213
+ "content": "<|audio_start|>",
214
+ "lstrip": false,
215
+ "normalized": false,
216
+ "rstrip": false,
217
+ "single_word": false,
218
+ "special": true
219
+ },
220
+ "248071": {
221
+ "content": "<|audio_end|>",
222
+ "lstrip": false,
223
+ "normalized": false,
224
+ "rstrip": false,
225
+ "single_word": false,
226
+ "special": true
227
+ },
228
+ "248072": {
229
+ "content": "<tts_pad>",
230
+ "lstrip": false,
231
+ "normalized": false,
232
+ "rstrip": false,
233
+ "single_word": false,
234
+ "special": true
235
+ },
236
+ "248073": {
237
+ "content": "<tts_text_bos>",
238
+ "lstrip": false,
239
+ "normalized": false,
240
+ "rstrip": false,
241
+ "single_word": false,
242
+ "special": true
243
+ },
244
+ "248074": {
245
+ "content": "<tts_text_eod>",
246
+ "lstrip": false,
247
+ "normalized": false,
248
+ "rstrip": false,
249
+ "single_word": false,
250
+ "special": true
251
+ },
252
+ "248075": {
253
+ "content": "<tts_text_bos_single>",
254
+ "lstrip": false,
255
+ "normalized": false,
256
+ "rstrip": false,
257
+ "single_word": false,
258
+ "special": true
259
+ },
260
+ "248076": {
261
+ "content": "<|audio_pad|>",
262
+ "lstrip": false,
263
+ "normalized": false,
264
+ "rstrip": false,
265
+ "single_word": false,
266
+ "special": true
267
+ }
268
+ },
269
+ "additional_special_tokens": [
270
+ "<|im_start|>",
271
+ "<|im_end|>",
272
+ "<|object_ref_start|>",
273
+ "<|object_ref_end|>",
274
+ "<|box_start|>",
275
+ "<|box_end|>",
276
+ "<|quad_start|>",
277
+ "<|quad_end|>",
278
+ "<|vision_start|>",
279
+ "<|vision_end|>",
280
+ "<|vision_pad|>",
281
+ "<|image_pad|>",
282
+ "<|video_pad|>"
283
+ ],
284
+ "bos_token": null,
285
+ "chat_template": "{%- set image_count = namespace(value=0) %}\n{%- set video_count = namespace(value=0) %}\n{%- macro render_content(content, do_vision_count, is_system_content=false) %}\n {%- if content is string %}\n {{- content }}\n {%- elif content is iterable and content is not mapping %}\n {%- for item in content %}\n {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}\n {%- if is_system_content %}\n {{- raise_exception('System message cannot contain images.') }}\n {%- endif %}\n {%- if do_vision_count %}\n {%- set image_count.value = image_count.value + 1 %}\n {%- endif %}\n {%- if add_vision_id %}\n {{- 'Picture ' ~ image_count.value ~ ': ' }}\n {%- endif %}\n {{- '<|vision_start|><|image_pad|><|vision_end|>' }}\n {%- elif 'video' in item or item.type == 'video' %}\n {%- if is_system_content %}\n {{- raise_exception('System message cannot contain videos.') }}\n {%- endif %}\n {%- if do_vision_count %}\n {%- set video_count.value = video_count.value + 1 %}\n {%- endif %}\n {%- if add_vision_id %}\n {{- 'Video ' ~ video_count.value ~ ': ' }}\n {%- endif %}\n {{- '<|vision_start|><|video_pad|><|vision_end|>' }}\n {%- elif 'text' in item %}\n {{- item.text }}\n {%- else %}\n {{- raise_exception('Unexpected item type in content.') }}\n {%- endif %}\n {%- endfor %}\n {%- elif content is none or content is undefined %}\n {{- '' }}\n {%- else %}\n {{- raise_exception('Unexpected content type.') }}\n {%- endif %}\n{%- endmacro %}\n{%- if not messages %}\n {{- raise_exception('No messages provided.') }}\n{%- endif %}\n{%- set reasoning_instructions = '' %}\n{%- if enable_thinking is undefined or enable_thinking is true %}\n {%- set resolved_reasoning_effort = reasoning_effort|default('xhigh') %}\n {%- if resolved_reasoning_effort not in ('xhigh', 'medium', 'low') %}\n {{- raise_exception('Unexpected reasoning effort ' ~ reasoning_effort ~ '. Supported types are xhigh (default), medium, and low.') }}\n {%- endif %}\n {%- if resolved_reasoning_effort == 'xhigh' %}\n {%- set reasoning_instructions = 'Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.' %}\n {%- elif resolved_reasoning_effort == 'low' %}\n {%- set reasoning_instructions = 'Reasoning effort is set to low. Keep your thinking brief and focused, moving directly to the conclusion without unnecessary elaboration.' %}\n {%- endif %}\n{%- endif %}\n{%- if tools and tools is iterable and tools is not mapping %}\n {{- '<|im_start|>system\\n' }}\n {%- if reasoning_instructions %}\n {{- reasoning_instructions + '\\n\\n' }}\n {%- endif %}\n {{- \"# Tools\\n\\nYou have access to the following functions:\\n\\n<tools>\" }}\n {%- for tool in tools %}\n {{- \"\\n\" }}\n {{- tool | tojson }}\n {%- endfor %}\n {{- \"\\n</tools>\" }}\n {{- '\\n\\nIf you choose to call a function ONLY reply in the following format with NO suffix:\\n\\n<tool_call>\\n<function=example_function_name>\\n<parameter=example_parameter_1>\\nvalue_1\\n</parameter>\\n<parameter=example_parameter_2>\\nThis is the value for the second parameter\\nthat can span\\nmultiple lines\\n</parameter>\\n</function>\\n</tool_call>\\n\\n<IMPORTANT>\\nReminder:\\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\\n- Required parameters MUST be specified\\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\\n</IMPORTANT>' }}\n {%- if messages[0].role == 'system' %}\n {%- set content = render_content(messages[0].content, false, true)|trim %}\n {%- if content %}\n {{- '\\n\\n' + content }}\n {%- endif %}\n {%- endif %}\n {{- '<|im_end|>\\n' }}\n{%- else %}\n {%- if messages[0].role == 'system' %}\n {%- set content = render_content(messages[0].content, false, true)|trim %}\n {%- if content %}\n {{- '<|im_start|>system\\n' + (reasoning_instructions + '\\n\\n' if reasoning_instructions else '') + content + '<|im_end|>\\n' }}\n {%- elif reasoning_instructions %}\n {{- '<|im_start|>system\\n' + reasoning_instructions + '<|im_end|>\\n' }}\n {%- endif %}\n {%- elif reasoning_instructions %}\n {{- '<|im_start|>system\\n' + reasoning_instructions + '<|im_end|>\\n' }}\n {%- endif %}\n{%- endif %}\n{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}\n{%- for message in messages[::-1] %}\n {%- set index = (messages|length - 1) - loop.index0 %}\n {%- if ns.multi_step_tool and message.role == \"user\" %}\n {%- set content = render_content(message.content, false)|trim %}\n {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}\n {%- set ns.multi_step_tool = false %}\n {%- set ns.last_query_index = index %}\n {%- endif %}\n {%- endif %}\n{%- endfor %}\n{%- if ns.multi_step_tool %}\n {{- raise_exception('No user query found in messages.') }}\n{%- endif %}\n{%- for message in messages %}\n {%- set content = render_content(message.content, true)|trim %}\n {%- if message.role == \"system\" %}\n {%- if not loop.first %}\n {{- raise_exception('System message must be at the beginning.') }}\n {%- endif %}\n {%- elif message.role == \"user\" %}\n {{- '<|im_start|>' + message.role + '\\n' + content + '<|im_end|>' + '\\n' }}\n {%- elif message.role == \"assistant\" %}\n {%- set reasoning_content = '' %}\n {%- if message.reasoning_content is string %}\n {%- set reasoning_content = message.reasoning_content %}\n {%- endif %}\n {%- set reasoning_content = reasoning_content|trim %}\n {%- if preserve_thinking is undefined or preserve_thinking is true or loop.index0 > ns.last_query_index %}\n {{- '<|im_start|>' + message.role + '\\n<think>\\n' + reasoning_content + '\\n</think>\\n\\n' + content }}\n {%- else %}\n {{- '<|im_start|>' + message.role + '\\n' + content }}\n {%- endif %}\n {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}\n {%- for tool_call in message.tool_calls %}\n {%- if tool_call.function is defined %}\n {%- set tool_call = tool_call.function %}\n {%- endif %}\n {%- if loop.first %}\n {%- if content|trim %}\n {{- '\\n\\n<tool_call>\\n<function=' + tool_call.name + '>\\n' }}\n {%- else %}\n {{- '<tool_call>\\n<function=' + tool_call.name + '>\\n' }}\n {%- endif %}\n {%- else %}\n {{- '\\n<tool_call>\\n<function=' + tool_call.name + '>\\n' }}\n {%- endif %}\n {%- if tool_call.arguments is defined and tool_call.arguments != '' %}\n {%- for args_name, args_value in tool_call.arguments|items %}\n {{- '<parameter=' + args_name + '>\\n' }}\n {%- set args_value = args_value | string if args_value is string else args_value | tojson | safe %}\n {{- args_value }}\n {{- '\\n</parameter>\\n' }}\n {%- endfor %}\n {%- endif %}\n {{- '</function>\\n</tool_call>' }}\n {%- endfor %}\n {%- endif %}\n {{- '<|im_end|>\\n' }}\n {%- elif message.role == \"tool\" %}\n {%- if loop.previtem and loop.previtem.role != \"tool\" %}\n {{- '<|im_start|>user' }}\n {%- endif %}\n {{- '\\n<tool_response>\\n' }}\n {{- content }}\n {{- '\\n</tool_response>' }}\n {%- if not loop.last and loop.nextitem.role != \"tool\" %}\n {{- '<|im_end|>\\n' }}\n {%- elif loop.last %}\n {{- '<|im_end|>\\n' }}\n {%- endif %}\n {%- else %}\n {{- raise_exception('Unexpected message role.') }}\n {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n {{- '<|im_start|>assistant\\n' }}\n {%- if enable_thinking is defined and enable_thinking is false %}\n {{- '<think>\\n\\n</think>\\n\\n' }}\n {%- else %}\n {{- '<think>\\n' }}\n {%- endif %}\n{%- endif %}",
286
+ "clean_up_tokenization_spaces": false,
287
+ "eos_token": "<|im_end|>",
288
+ "errors": "replace",
289
+ "model_max_length": 262144,
290
+ "pad_token": "<|endoftext|>",
291
+ "split_special_tokens": false,
292
+ "tokenizer_class": "Qwen2Tokenizer",
293
+ "unk_token": null,
294
+ "add_bos_token": false,
295
+ "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
296
+ "extra_special_tokens": {
297
+ "audio_bos_token": "<|audio_start|>",
298
+ "audio_eos_token": "<|audio_end|>",
299
+ "audio_token": "<|audio_pad|>",
300
+ "image_token": "<|image_pad|>",
301
+ "video_token": "<|video_pad|>",
302
+ "vision_bos_token": "<|vision_start|>",
303
+ "vision_eos_token": "<|vision_end|>"
304
+ }
305
+ }
video_preprocessor_config.json ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "size": {
3
+ "longest_edge": 25165824,
4
+ "shortest_edge": 4096
5
+ },
6
+ "patch_size": 16,
7
+ "temporal_patch_size": 2,
8
+ "merge_size": 2,
9
+ "image_mean": [
10
+ 0.5,
11
+ 0.5,
12
+ 0.5
13
+ ],
14
+ "image_std": [
15
+ 0.5,
16
+ 0.5,
17
+ 0.5
18
+ ],
19
+ "processor_class": "Qwen3VLProcessor",
20
+ "video_processor_type": "Qwen3VLVideoProcessor"
21
+ }
vocab.json ADDED
The diff for this file is too large to render. See raw diff