Hugging Face's logo

Sehyo
/
Qwen3.5-397B-A17B-NVFP4

Safetensors
qwen3_5_moe
quantized
nvfp4
compressed-tensors
llm-compressor
Mixture of Experts
qwen3.5
Model card Files
xet
 Community
9
Sehyo commited on
Commit
70b659f
·
verified ·
1 Parent(s): 1a5ef2c

Add files using upload-large-folder tool

Browse files
Files changed (14) hide show
  1. .gitattributes +2 -0
  2. README.md +19 -0
  3. chat_template.jinja +154 -0
  4. config.json +789 -0
  5. generation_config.json +13 -0
  6. model-00001-of-00005.safetensors +3 -0
  7. model-00002-of-00005.safetensors +3 -0
  8. model-00003-of-00005.safetensors +3 -0
  9. model-00004-of-00005.safetensors +3 -0
  10. model-00005-of-00005.safetensors +3 -0
  11. model.safetensors.index.json +3 -0
  12. recipe.yaml +6 -0
  13. tokenizer.json +3 -0
  14. tokenizer_config.json +31 -0
.gitattributes CHANGED
@@ -33,3 +33,5 @@ saved_model/**/* 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
 
 
 
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 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: Qwen/Qwen3.5-397B-A17B
3
+ tags:
4
+ - quantized
5
+ - nvfp4
6
+ - compressed-tensors
7
+ - llm-compressor
8
+ - moe
9
+ - qwen3.5
10
+ quantized_by: Sehyo
11
+ ---
12
+
13
+ # Qwen3.5-397B-A17B-NVFP4
14
+
15
+ This is a quantized version of [Qwen/Qwen3.5-397B-A17B](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) using the **NVFP4** quantization scheme.
16
+
17
+ ## Creation
18
+
19
+ This model was created using [VLLM's LLM Compressor](https://github.com/vllm-project/llm-compressor) with Qwen3.5 MoE support added via [PR #2383](https://github.com/vllm-project/llm-compressor/pull/2383). The PR adds a custom `CalibrationQwen3MoeSparseMoeBlock` that routes calibration data to all experts during quantization, ensuring every expert receives proper calibration for accurate NVFP4 quantization.
chat_template.jinja ADDED
@@ -0,0 +1,154 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ {%- if tools and tools is iterable and tools is not mapping %}
46
+ {{- '<|im_start|>system\n' }}
47
+ {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
48
+ {%- for tool in tools %}
49
+ {{- "\n" }}
50
+ {{- tool | tojson }}
51
+ {%- endfor %}
52
+ {{- "\n</tools>" }}
53
+ {{- '\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>' }}
54
+ {%- if messages[0].role == 'system' %}
55
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
56
+ {%- if content %}
57
+ {{- '\n\n' + content }}
58
+ {%- endif %}
59
+ {%- endif %}
60
+ {{- '<|im_end|>\n' }}
61
+ {%- else %}
62
+ {%- if messages[0].role == 'system' %}
63
+ {%- set content = render_content(messages[0].content, false, true)|trim %}
64
+ {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
65
+ {%- endif %}
66
+ {%- endif %}
67
+ {%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
68
+ {%- for message in messages[::-1] %}
69
+ {%- set index = (messages|length - 1) - loop.index0 %}
70
+ {%- if ns.multi_step_tool and message.role == "user" %}
71
+ {%- set content = render_content(message.content, false)|trim %}
72
+ {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
73
+ {%- set ns.multi_step_tool = false %}
74
+ {%- set ns.last_query_index = index %}
75
+ {%- endif %}
76
+ {%- endif %}
77
+ {%- endfor %}
78
+ {%- if ns.multi_step_tool %}
79
+ {{- raise_exception('No user query found in messages.') }}
80
+ {%- endif %}
81
+ {%- for message in messages %}
82
+ {%- set content = render_content(message.content, true)|trim %}
83
+ {%- if message.role == "system" %}
84
+ {%- if not loop.first %}
85
+ {{- raise_exception('System message must be at the beginning.') }}
86
+ {%- endif %}
87
+ {%- elif message.role == "user" %}
88
+ {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
89
+ {%- elif message.role == "assistant" %}
90
+ {%- set reasoning_content = '' %}
91
+ {%- if message.reasoning_content is string %}
92
+ {%- set reasoning_content = message.reasoning_content %}
93
+ {%- else %}
94
+ {%- if '</think>' in content %}
95
+ {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
96
+ {%- set content = content.split('</think>')[-1].lstrip('\n') %}
97
+ {%- endif %}
98
+ {%- endif %}
99
+ {%- set reasoning_content = reasoning_content|trim %}
100
+ {%- if loop.index0 > ns.last_query_index %}
101
+ {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
102
+ {%- else %}
103
+ {{- '<|im_start|>' + message.role + '\n' + content }}
104
+ {%- endif %}
105
+ {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
106
+ {%- for tool_call in message.tool_calls %}
107
+ {%- if tool_call.function is defined %}
108
+ {%- set tool_call = tool_call.function %}
109
+ {%- endif %}
110
+ {%- if loop.first %}
111
+ {%- if content|trim %}
112
+ {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
113
+ {%- else %}
114
+ {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
115
+ {%- endif %}
116
+ {%- else %}
117
+ {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
118
+ {%- endif %}
119
+ {%- if tool_call.arguments is defined %}
120
+ {%- for args_name, args_value in tool_call.arguments|items %}
121
+ {{- '<parameter=' + args_name + '>\n' }}
122
+ {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
123
+ {{- args_value }}
124
+ {{- '\n</parameter>\n' }}
125
+ {%- endfor %}
126
+ {%- endif %}
127
+ {{- '</function>\n</tool_call>' }}
128
+ {%- endfor %}
129
+ {%- endif %}
130
+ {{- '<|im_end|>\n' }}
131
+ {%- elif message.role == "tool" %}
132
+ {%- if loop.previtem and loop.previtem.role != "tool" %}
133
+ {{- '<|im_start|>user' }}
134
+ {%- endif %}
135
+ {{- '\n<tool_response>\n' }}
136
+ {{- content }}
137
+ {{- '\n</tool_response>' }}
138
+ {%- if not loop.last and loop.nextitem.role != "tool" %}
139
+ {{- '<|im_end|>\n' }}
140
+ {%- elif loop.last %}
141
+ {{- '<|im_end|>\n' }}
142
+ {%- endif %}
143
+ {%- else %}
144
+ {{- raise_exception('Unexpected message role.') }}
145
+ {%- endif %}
146
+ {%- endfor %}
147
+ {%- if add_generation_prompt %}
148
+ {{- '<|im_start|>assistant\n' }}
149
+ {%- if enable_thinking is defined and enable_thinking is false %}
150
+ {{- '<think>\n\n</think>\n\n' }}
151
+ {%- else %}
152
+ {{- '<think>\n' }}
153
+ {%- endif %}
154
+ {%- endif %}
config.json ADDED
@@ -0,0 +1,789 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Qwen3_5MoeForCausalLM"
4
+ ],
5
+ "model_type": "qwen3_5_moe",
6
+ "text_config": {
7
+ "attention_bias": false,
8
+ "attention_dropout": 0.0,
9
+ "attn_output_gate": true,
10
+ "bos_token_id": null,
11
+ "dtype": "bfloat16",
12
+ "eos_token_id": 248044,
13
+ "full_attention_interval": 4,
14
+ "head_dim": 256,
15
+ "hidden_act": "silu",
16
+ "hidden_size": 4096,
17
+ "initializer_range": 0.02,
18
+ "layer_types": [
19
+ "linear_attention",
20
+ "linear_attention",
21
+ "linear_attention",
22
+ "full_attention",
23
+ "linear_attention",
24
+ "linear_attention",
25
+ "linear_attention",
26
+ "full_attention",
27
+ "linear_attention",
28
+ "linear_attention",
29
+ "linear_attention",
30
+ "full_attention",
31
+ "linear_attention",
32
+ "linear_attention",
33
+ "linear_attention",
34
+ "full_attention",
35
+ "linear_attention",
36
+ "linear_attention",
37
+ "linear_attention",
38
+ "full_attention",
39
+ "linear_attention",
40
+ "linear_attention",
41
+ "linear_attention",
42
+ "full_attention",
43
+ "linear_attention",
44
+ "linear_attention",
45
+ "linear_attention",
46
+ "full_attention",
47
+ "linear_attention",
48
+ "linear_attention",
49
+ "linear_attention",
50
+ "full_attention",
51
+ "linear_attention",
52
+ "linear_attention",
53
+ "linear_attention",
54
+ "full_attention",
55
+ "linear_attention",
56
+ "linear_attention",
57
+ "linear_attention",
58
+ "full_attention",
59
+ "linear_attention",
60
+ "linear_attention",
61
+ "linear_attention",
62
+ "full_attention",
63
+ "linear_attention",
64
+ "linear_attention",
65
+ "linear_attention",
66
+ "full_attention",
67
+ "linear_attention",
68
+ "linear_attention",
69
+ "linear_attention",
70
+ "full_attention",
71
+ "linear_attention",
72
+ "linear_attention",
73
+ "linear_attention",
74
+ "full_attention",
75
+ "linear_attention",
76
+ "linear_attention",
77
+ "linear_attention",
78
+ "full_attention"
79
+ ],
80
+ "linear_conv_kernel_dim": 4,
81
+ "linear_key_head_dim": 128,
82
+ "linear_num_key_heads": 16,
83
+ "linear_num_value_heads": 64,
84
+ "linear_value_head_dim": 128,
85
+ "mamba_ssm_dtype": "float32",
86
+ "max_position_embeddings": 262144,
87
+ "mlp_only_layers": [],
88
+ "moe_intermediate_size": 1024,
89
+ "mtp_num_hidden_layers": 1,
90
+ "mtp_use_dedicated_embeddings": false,
91
+ "num_attention_heads": 32,
92
+ "num_experts": 512,
93
+ "num_experts_per_tok": 10,
94
+ "num_hidden_layers": 60,
95
+ "num_key_value_heads": 2,
96
+ "output_router_logits": false,
97
+ "pad_token_id": null,
98
+ "partial_rotary_factor": 0.25,
99
+ "quantization_config": {
100
+ "config_groups": {
101
+ "group_0": {
102
+ "format": "nvfp4-pack-quantized",
103
+ "input_activations": {
104
+ "actorder": null,
105
+ "block_structure": null,
106
+ "dynamic": "local",
107
+ "group_size": 16,
108
+ "num_bits": 4,
109
+ "observer": "static_minmax",
110
+ "observer_kwargs": {},
111
+ "scale_dtype": "torch.float8_e4m3fn",
112
+ "strategy": "tensor_group",
113
+ "symmetric": true,
114
+ "type": "float",
115
+ "zp_dtype": null
116
+ },
117
+ "output_activations": null,
118
+ "targets": [
119
+ "Linear"
120
+ ],
121
+ "weights": {
122
+ "actorder": null,
123
+ "block_structure": null,
124
+ "dynamic": false,
125
+ "group_size": 16,
126
+ "num_bits": 4,
127
+ "observer": "memoryless_minmax",
128
+ "observer_kwargs": {},
129
+ "scale_dtype": "torch.float8_e4m3fn",
130
+ "strategy": "tensor_group",
131
+ "symmetric": true,
132
+ "type": "float",
133
+ "zp_dtype": null
134
+ }
135
+ }
136
+ },
137
+ "format": "nvfp4-pack-quantized",
138
+ "global_compression_ratio": null,
139
+ "ignore": [
140
+ "model.layers.0.linear_attn.out_proj",
141
+ "model.layers.0.linear_attn.in_proj_qkv",
142
+ "model.layers.0.linear_attn.in_proj_z",
143
+ "model.layers.0.linear_attn.in_proj_b",
144
+ "model.layers.0.linear_attn.in_proj_a",
145
+ "model.layers.0.mlp.shared_expert_gate",
146
+ "model.layers.1.linear_attn.out_proj",
147
+ "model.layers.1.linear_attn.in_proj_qkv",
148
+ "model.layers.1.linear_attn.in_proj_z",
149
+ "model.layers.1.linear_attn.in_proj_b",
150
+ "model.layers.1.linear_attn.in_proj_a",
151
+ "model.layers.1.mlp.shared_expert_gate",
152
+ "model.layers.2.linear_attn.out_proj",
153
+ "model.layers.2.linear_attn.in_proj_qkv",
154
+ "model.layers.2.linear_attn.in_proj_z",
155
+ "model.layers.2.linear_attn.in_proj_b",
156
+ "model.layers.2.linear_attn.in_proj_a",
157
+ "model.layers.2.mlp.shared_expert_gate",
158
+ "model.layers.3.mlp.shared_expert_gate",
159
+ "model.layers.4.linear_attn.out_proj",
160
+ "model.layers.4.linear_attn.in_proj_qkv",
161
+ "model.layers.4.linear_attn.in_proj_z",
162
+ "model.layers.4.linear_attn.in_proj_b",
163
+ "model.layers.4.linear_attn.in_proj_a",
164
+ "model.layers.4.mlp.shared_expert_gate",
165
+ "model.layers.5.linear_attn.out_proj",
166
+ "model.layers.5.linear_attn.in_proj_qkv",
167
+ "model.layers.5.linear_attn.in_proj_z",
168
+ "model.layers.5.linear_attn.in_proj_b",
169
+ "model.layers.5.linear_attn.in_proj_a",
170
+ "model.layers.5.mlp.shared_expert_gate",
171
+ "model.layers.6.linear_attn.out_proj",
172
+ "model.layers.6.linear_attn.in_proj_qkv",
173
+ "model.layers.6.linear_attn.in_proj_z",
174
+ "model.layers.6.linear_attn.in_proj_b",
175
+ "model.layers.6.linear_attn.in_proj_a",
176
+ "model.layers.6.mlp.shared_expert_gate",
177
+ "model.layers.7.mlp.shared_expert_gate",
178
+ "model.layers.8.linear_attn.out_proj",
179
+ "model.layers.8.linear_attn.in_proj_qkv",
180
+ "model.layers.8.linear_attn.in_proj_z",
181
+ "model.layers.8.linear_attn.in_proj_b",
182
+ "model.layers.8.linear_attn.in_proj_a",
183
+ "model.layers.8.mlp.shared_expert_gate",
184
+ "model.layers.9.linear_attn.out_proj",
185
+ "model.layers.9.linear_attn.in_proj_qkv",
186
+ "model.layers.9.linear_attn.in_proj_z",
187
+ "model.layers.9.linear_attn.in_proj_b",
188
+ "model.layers.9.linear_attn.in_proj_a",
189
+ "model.layers.9.mlp.shared_expert_gate",
190
+ "model.layers.10.linear_attn.out_proj",
191
+ "model.layers.10.linear_attn.in_proj_qkv",
192
+ "model.layers.10.linear_attn.in_proj_z",
193
+ "model.layers.10.linear_attn.in_proj_b",
194
+ "model.layers.10.linear_attn.in_proj_a",
195
+ "model.layers.10.mlp.shared_expert_gate",
196
+ "model.layers.11.mlp.shared_expert_gate",
197
+ "model.layers.12.linear_attn.out_proj",
198
+ "model.layers.12.linear_attn.in_proj_qkv",
199
+ "model.layers.12.linear_attn.in_proj_z",
200
+ "model.layers.12.linear_attn.in_proj_b",
201
+ "model.layers.12.linear_attn.in_proj_a",
202
+ "model.layers.12.mlp.shared_expert_gate",
203
+ "model.layers.13.linear_attn.out_proj",
204
+ "model.layers.13.linear_attn.in_proj_qkv",
205
+ "model.layers.13.linear_attn.in_proj_z",
206
+ "model.layers.13.linear_attn.in_proj_b",
207
+ "model.layers.13.linear_attn.in_proj_a",
208
+ "model.layers.13.mlp.shared_expert_gate",
209
+ "model.layers.14.linear_attn.out_proj",
210
+ "model.layers.14.linear_attn.in_proj_qkv",
211
+ "model.layers.14.linear_attn.in_proj_z",
212
+ "model.layers.14.linear_attn.in_proj_b",
213
+ "model.layers.14.linear_attn.in_proj_a",
214
+ "model.layers.14.mlp.shared_expert_gate",
215
+ "model.layers.15.mlp.shared_expert_gate",
216
+ "model.layers.16.linear_attn.out_proj",
217
+ "model.layers.16.linear_attn.in_proj_qkv",
218
+ "model.layers.16.linear_attn.in_proj_z",
219
+ "model.layers.16.linear_attn.in_proj_b",
220
+ "model.layers.16.linear_attn.in_proj_a",
221
+ "model.layers.16.mlp.shared_expert_gate",
222
+ "model.layers.17.linear_attn.out_proj",
223
+ "model.layers.17.linear_attn.in_proj_qkv",
224
+ "model.layers.17.linear_attn.in_proj_z",
225
+ "model.layers.17.linear_attn.in_proj_b",
226
+ "model.layers.17.linear_attn.in_proj_a",
227
+ "model.layers.17.mlp.shared_expert_gate",
228
+ "model.layers.18.linear_attn.out_proj",
229
+ "model.layers.18.linear_attn.in_proj_qkv",
230
+ "model.layers.18.linear_attn.in_proj_z",
231
+ "model.layers.18.linear_attn.in_proj_b",
232
+ "model.layers.18.linear_attn.in_proj_a",
233
+ "model.layers.18.mlp.shared_expert_gate",
234
+ "model.layers.19.mlp.shared_expert_gate",
235
+ "model.layers.20.linear_attn.out_proj",
236
+ "model.layers.20.linear_attn.in_proj_qkv",
237
+ "model.layers.20.linear_attn.in_proj_z",
238
+ "model.layers.20.linear_attn.in_proj_b",
239
+ "model.layers.20.linear_attn.in_proj_a",
240
+ "model.layers.20.mlp.shared_expert_gate",
241
+ "model.layers.21.linear_attn.out_proj",
242
+ "model.layers.21.linear_attn.in_proj_qkv",
243
+ "model.layers.21.linear_attn.in_proj_z",
244
+ "model.layers.21.linear_attn.in_proj_b",
245
+ "model.layers.21.linear_attn.in_proj_a",
246
+ "model.layers.21.mlp.shared_expert_gate",
247
+ "model.layers.22.linear_attn.out_proj",
248
+ "model.layers.22.linear_attn.in_proj_qkv",
249
+ "model.layers.22.linear_attn.in_proj_z",
250
+ "model.layers.22.linear_attn.in_proj_b",
251
+ "model.layers.22.linear_attn.in_proj_a",
252
+ "model.layers.22.mlp.shared_expert_gate",
253
+ "model.layers.23.mlp.shared_expert_gate",
254
+ "model.layers.24.linear_attn.out_proj",
255
+ "model.layers.24.linear_attn.in_proj_qkv",
256
+ "model.layers.24.linear_attn.in_proj_z",
257
+ "model.layers.24.linear_attn.in_proj_b",
258
+ "model.layers.24.linear_attn.in_proj_a",
259
+ "model.layers.24.mlp.shared_expert_gate",
260
+ "model.layers.25.linear_attn.out_proj",
261
+ "model.layers.25.linear_attn.in_proj_qkv",
262
+ "model.layers.25.linear_attn.in_proj_z",
263
+ "model.layers.25.linear_attn.in_proj_b",
264
+ "model.layers.25.linear_attn.in_proj_a",
265
+ "model.layers.25.mlp.shared_expert_gate",
266
+ "model.layers.26.linear_attn.out_proj",
267
+ "model.layers.26.linear_attn.in_proj_qkv",
268
+ "model.layers.26.linear_attn.in_proj_z",
269
+ "model.layers.26.linear_attn.in_proj_b",
270
+ "model.layers.26.linear_attn.in_proj_a",
271
+ "model.layers.26.mlp.shared_expert_gate",
272
+ "model.layers.27.mlp.shared_expert_gate",
273
+ "model.layers.28.linear_attn.out_proj",
274
+ "model.layers.28.linear_attn.in_proj_qkv",
275
+ "model.layers.28.linear_attn.in_proj_z",
276
+ "model.layers.28.linear_attn.in_proj_b",
277
+ "model.layers.28.linear_attn.in_proj_a",
278
+ "model.layers.28.mlp.shared_expert_gate",
279
+ "model.layers.29.linear_attn.out_proj",
280
+ "model.layers.29.linear_attn.in_proj_qkv",
281
+ "model.layers.29.linear_attn.in_proj_z",
282
+ "model.layers.29.linear_attn.in_proj_b",
283
+ "model.layers.29.linear_attn.in_proj_a",
284
+ "model.layers.29.mlp.shared_expert_gate",
285
+ "model.layers.30.linear_attn.out_proj",
286
+ "model.layers.30.linear_attn.in_proj_qkv",
287
+ "model.layers.30.linear_attn.in_proj_z",
288
+ "model.layers.30.linear_attn.in_proj_b",
289
+ "model.layers.30.linear_attn.in_proj_a",
290
+ "model.layers.30.mlp.shared_expert_gate",
291
+ "model.layers.31.mlp.shared_expert_gate",
292
+ "model.layers.32.linear_attn.out_proj",
293
+ "model.layers.32.linear_attn.in_proj_qkv",
294
+ "model.layers.32.linear_attn.in_proj_z",
295
+ "model.layers.32.linear_attn.in_proj_b",
296
+ "model.layers.32.linear_attn.in_proj_a",
297
+ "model.layers.32.mlp.shared_expert_gate",
298
+ "model.layers.33.linear_attn.out_proj",
299
+ "model.layers.33.linear_attn.in_proj_qkv",
300
+ "model.layers.33.linear_attn.in_proj_z",
301
+ "model.layers.33.linear_attn.in_proj_b",
302
+ "model.layers.33.linear_attn.in_proj_a",
303
+ "model.layers.33.mlp.shared_expert_gate",
304
+ "model.layers.34.linear_attn.out_proj",
305
+ "model.layers.34.linear_attn.in_proj_qkv",
306
+ "model.layers.34.linear_attn.in_proj_z",
307
+ "model.layers.34.linear_attn.in_proj_b",
308
+ "model.layers.34.linear_attn.in_proj_a",
309
+ "model.layers.34.mlp.shared_expert_gate",
310
+ "model.layers.35.mlp.shared_expert_gate",
311
+ "model.layers.36.linear_attn.out_proj",
312
+ "model.layers.36.linear_attn.in_proj_qkv",
313
+ "model.layers.36.linear_attn.in_proj_z",
314
+ "model.layers.36.linear_attn.in_proj_b",
315
+ "model.layers.36.linear_attn.in_proj_a",
316
+ "model.layers.36.mlp.shared_expert_gate",
317
+ "model.layers.37.linear_attn.out_proj",
318
+ "model.layers.37.linear_attn.in_proj_qkv",
319
+ "model.layers.37.linear_attn.in_proj_z",
320
+ "model.layers.37.linear_attn.in_proj_b",
321
+ "model.layers.37.linear_attn.in_proj_a",
322
+ "model.layers.37.mlp.shared_expert_gate",
323
+ "model.layers.38.linear_attn.out_proj",
324
+ "model.layers.38.linear_attn.in_proj_qkv",
325
+ "model.layers.38.linear_attn.in_proj_z",
326
+ "model.layers.38.linear_attn.in_proj_b",
327
+ "model.layers.38.linear_attn.in_proj_a",
328
+ "model.layers.38.mlp.shared_expert_gate",
329
+ "model.layers.39.mlp.shared_expert_gate",
330
+ "model.layers.40.linear_attn.out_proj",
331
+ "model.layers.40.linear_attn.in_proj_qkv",
332
+ "model.layers.40.linear_attn.in_proj_z",
333
+ "model.layers.40.linear_attn.in_proj_b",
334
+ "model.layers.40.linear_attn.in_proj_a",
335
+ "model.layers.40.mlp.shared_expert_gate",
336
+ "model.layers.41.linear_attn.out_proj",
337
+ "model.layers.41.linear_attn.in_proj_qkv",
338
+ "model.layers.41.linear_attn.in_proj_z",
339
+ "model.layers.41.linear_attn.in_proj_b",
340
+ "model.layers.41.linear_attn.in_proj_a",
341
+ "model.layers.41.mlp.shared_expert_gate",
342
+ "model.layers.42.linear_attn.out_proj",
343
+ "model.layers.42.linear_attn.in_proj_qkv",
344
+ "model.layers.42.linear_attn.in_proj_z",
345
+ "model.layers.42.linear_attn.in_proj_b",
346
+ "model.layers.42.linear_attn.in_proj_a",
347
+ "model.layers.42.mlp.shared_expert_gate",
348
+ "model.layers.43.mlp.shared_expert_gate",
349
+ "model.layers.44.linear_attn.out_proj",
350
+ "model.layers.44.linear_attn.in_proj_qkv",
351
+ "model.layers.44.linear_attn.in_proj_z",
352
+ "model.layers.44.linear_attn.in_proj_b",
353
+ "model.layers.44.linear_attn.in_proj_a",
354
+ "model.layers.44.mlp.shared_expert_gate",
355
+ "model.layers.45.linear_attn.out_proj",
356
+ "model.layers.45.linear_attn.in_proj_qkv",
357
+ "model.layers.45.linear_attn.in_proj_z",
358
+ "model.layers.45.linear_attn.in_proj_b",
359
+ "model.layers.45.linear_attn.in_proj_a",
360
+ "model.layers.45.mlp.shared_expert_gate",
361
+ "model.layers.46.linear_attn.out_proj",
362
+ "model.layers.46.linear_attn.in_proj_qkv",
363
+ "model.layers.46.linear_attn.in_proj_z",
364
+ "model.layers.46.linear_attn.in_proj_b",
365
+ "model.layers.46.linear_attn.in_proj_a",
366
+ "model.layers.46.mlp.shared_expert_gate",
367
+ "model.layers.47.mlp.shared_expert_gate",
368
+ "model.layers.48.linear_attn.out_proj",
369
+ "model.layers.48.linear_attn.in_proj_qkv",
370
+ "model.layers.48.linear_attn.in_proj_z",
371
+ "model.layers.48.linear_attn.in_proj_b",
372
+ "model.layers.48.linear_attn.in_proj_a",
373
+ "model.layers.48.mlp.shared_expert_gate",
374
+ "model.layers.49.linear_attn.out_proj",
375
+ "model.layers.49.linear_attn.in_proj_qkv",
376
+ "model.layers.49.linear_attn.in_proj_z",
377
+ "model.layers.49.linear_attn.in_proj_b",
378
+ "model.layers.49.linear_attn.in_proj_a",
379
+ "model.layers.49.mlp.shared_expert_gate",
380
+ "model.layers.50.linear_attn.out_proj",
381
+ "model.layers.50.linear_attn.in_proj_qkv",
382
+ "model.layers.50.linear_attn.in_proj_z",
383
+ "model.layers.50.linear_attn.in_proj_b",
384
+ "model.layers.50.linear_attn.in_proj_a",
385
+ "model.layers.50.mlp.shared_expert_gate",
386
+ "model.layers.51.mlp.shared_expert_gate",
387
+ "model.layers.52.linear_attn.out_proj",
388
+ "model.layers.52.linear_attn.in_proj_qkv",
389
+ "model.layers.52.linear_attn.in_proj_z",
390
+ "model.layers.52.linear_attn.in_proj_b",
391
+ "model.layers.52.linear_attn.in_proj_a",
392
+ "model.layers.52.mlp.shared_expert_gate",
393
+ "model.layers.53.linear_attn.out_proj",
394
+ "model.layers.53.linear_attn.in_proj_qkv",
395
+ "model.layers.53.linear_attn.in_proj_z",
396
+ "model.layers.53.linear_attn.in_proj_b",
397
+ "model.layers.53.linear_attn.in_proj_a",
398
+ "model.layers.53.mlp.shared_expert_gate",
399
+ "model.layers.54.linear_attn.out_proj",
400
+ "model.layers.54.linear_attn.in_proj_qkv",
401
+ "model.layers.54.linear_attn.in_proj_z",
402
+ "model.layers.54.linear_attn.in_proj_b",
403
+ "model.layers.54.linear_attn.in_proj_a",
404
+ "model.layers.54.mlp.shared_expert_gate",
405
+ "model.layers.55.mlp.shared_expert_gate",
406
+ "model.layers.56.linear_attn.out_proj",
407
+ "model.layers.56.linear_attn.in_proj_qkv",
408
+ "model.layers.56.linear_attn.in_proj_z",
409
+ "model.layers.56.linear_attn.in_proj_b",
410
+ "model.layers.56.linear_attn.in_proj_a",
411
+ "model.layers.56.mlp.shared_expert_gate",
412
+ "model.layers.57.linear_attn.out_proj",
413
+ "model.layers.57.linear_attn.in_proj_qkv",
414
+ "model.layers.57.linear_attn.in_proj_z",
415
+ "model.layers.57.linear_attn.in_proj_b",
416
+ "model.layers.57.linear_attn.in_proj_a",
417
+ "model.layers.57.mlp.shared_expert_gate",
418
+ "model.layers.58.linear_attn.out_proj",
419
+ "model.layers.58.linear_attn.in_proj_qkv",
420
+ "model.layers.58.linear_attn.in_proj_z",
421
+ "model.layers.58.linear_attn.in_proj_b",
422
+ "model.layers.58.linear_attn.in_proj_a",
423
+ "model.layers.58.mlp.shared_expert_gate",
424
+ "model.layers.59.mlp.shared_expert_gate",
425
+ "lm_head"
426
+ ],
427
+ "kv_cache_scheme": null,
428
+ "quant_method": "compressed-tensors",
429
+ "quantization_status": "compressed",
430
+ "sparsity_config": {},
431
+ "transform_config": {},
432
+ "version": "0.13.1.a20260212"
433
+ },
434
+ "rms_norm_eps": 1e-06,
435
+ "rope_parameters": {
436
+ "mrope_interleaved": true,
437
+ "mrope_section": [
438
+ 11,
439
+ 11,
440
+ 10
441
+ ],
442
+ "partial_rotary_factor": 0.25,
443
+ "rope_theta": 10000000,
444
+ "rope_type": "default"
445
+ },
446
+ "router_aux_loss_coef": 0.001,
447
+ "shared_expert_intermediate_size": 1024,
448
+ "tie_word_embeddings": false,
449
+ "transformers_version": "5.2.0.dev0",
450
+ "use_cache": true,
451
+ "vocab_size": 248320,
452
+ "model_type": "qwen3_5_moe_text"
453
+ },
454
+ "quantization_config": {
455
+ "config_groups": {
456
+ "group_0": {
457
+ "format": "nvfp4-pack-quantized",
458
+ "input_activations": {
459
+ "actorder": null,
460
+ "block_structure": null,
461
+ "dynamic": "local",
462
+ "group_size": 16,
463
+ "num_bits": 4,
464
+ "observer": "static_minmax",
465
+ "observer_kwargs": {},
466
+ "scale_dtype": "torch.float8_e4m3fn",
467
+ "strategy": "tensor_group",
468
+ "symmetric": true,
469
+ "type": "float",
470
+ "zp_dtype": null
471
+ },
472
+ "output_activations": null,
473
+ "targets": [
474
+ "Linear"
475
+ ],
476
+ "weights": {
477
+ "actorder": null,
478
+ "block_structure": null,
479
+ "dynamic": false,
480
+ "group_size": 16,
481
+ "num_bits": 4,
482
+ "observer": "memoryless_minmax",
483
+ "observer_kwargs": {},
484
+ "scale_dtype": "torch.float8_e4m3fn",
485
+ "strategy": "tensor_group",
486
+ "symmetric": true,
487
+ "type": "float",
488
+ "zp_dtype": null
489
+ }
490
+ }
491
+ },
492
+ "format": "nvfp4-pack-quantized",
493
+ "global_compression_ratio": null,
494
+ "ignore": [
495
+ "model.layers.0.linear_attn.out_proj",
496
+ "model.layers.0.linear_attn.in_proj_qkv",
497
+ "model.layers.0.linear_attn.in_proj_z",
498
+ "model.layers.0.linear_attn.in_proj_b",
499
+ "model.layers.0.linear_attn.in_proj_a",
500
+ "model.layers.0.mlp.shared_expert_gate",
501
+ "model.layers.1.linear_attn.out_proj",
502
+ "model.layers.1.linear_attn.in_proj_qkv",
503
+ "model.layers.1.linear_attn.in_proj_z",
504
+ "model.layers.1.linear_attn.in_proj_b",
505
+ "model.layers.1.linear_attn.in_proj_a",
506
+ "model.layers.1.mlp.shared_expert_gate",
507
+ "model.layers.2.linear_attn.out_proj",
508
+ "model.layers.2.linear_attn.in_proj_qkv",
509
+ "model.layers.2.linear_attn.in_proj_z",
510
+ "model.layers.2.linear_attn.in_proj_b",
511
+ "model.layers.2.linear_attn.in_proj_a",
512
+ "model.layers.2.mlp.shared_expert_gate",
513
+ "model.layers.3.mlp.shared_expert_gate",
514
+ "model.layers.4.linear_attn.out_proj",
515
+ "model.layers.4.linear_attn.in_proj_qkv",
516
+ "model.layers.4.linear_attn.in_proj_z",
517
+ "model.layers.4.linear_attn.in_proj_b",
518
+ "model.layers.4.linear_attn.in_proj_a",
519
+ "model.layers.4.mlp.shared_expert_gate",
520
+ "model.layers.5.linear_attn.out_proj",
521
+ "model.layers.5.linear_attn.in_proj_qkv",
522
+ "model.layers.5.linear_attn.in_proj_z",
523
+ "model.layers.5.linear_attn.in_proj_b",
524
+ "model.layers.5.linear_attn.in_proj_a",
525
+ "model.layers.5.mlp.shared_expert_gate",
526
+ "model.layers.6.linear_attn.out_proj",
527
+ "model.layers.6.linear_attn.in_proj_qkv",
528
+ "model.layers.6.linear_attn.in_proj_z",
529
+ "model.layers.6.linear_attn.in_proj_b",
530
+ "model.layers.6.linear_attn.in_proj_a",
531
+ "model.layers.6.mlp.shared_expert_gate",
532
+ "model.layers.7.mlp.shared_expert_gate",
533
+ "model.layers.8.linear_attn.out_proj",
534
+ "model.layers.8.linear_attn.in_proj_qkv",
535
+ "model.layers.8.linear_attn.in_proj_z",
536
+ "model.layers.8.linear_attn.in_proj_b",
537
+ "model.layers.8.linear_attn.in_proj_a",
538
+ "model.layers.8.mlp.shared_expert_gate",
539
+ "model.layers.9.linear_attn.out_proj",
540
+ "model.layers.9.linear_attn.in_proj_qkv",
541
+ "model.layers.9.linear_attn.in_proj_z",
542
+ "model.layers.9.linear_attn.in_proj_b",
543
+ "model.layers.9.linear_attn.in_proj_a",
544
+ "model.layers.9.mlp.shared_expert_gate",
545
+ "model.layers.10.linear_attn.out_proj",
546
+ "model.layers.10.linear_attn.in_proj_qkv",
547
+ "model.layers.10.linear_attn.in_proj_z",
548
+ "model.layers.10.linear_attn.in_proj_b",
549
+ "model.layers.10.linear_attn.in_proj_a",
550
+ "model.layers.10.mlp.shared_expert_gate",
551
+ "model.layers.11.mlp.shared_expert_gate",
552
+ "model.layers.12.linear_attn.out_proj",
553
+ "model.layers.12.linear_attn.in_proj_qkv",
554
+ "model.layers.12.linear_attn.in_proj_z",
555
+ "model.layers.12.linear_attn.in_proj_b",
556
+ "model.layers.12.linear_attn.in_proj_a",
557
+ "model.layers.12.mlp.shared_expert_gate",
558
+ "model.layers.13.linear_attn.out_proj",
559
+ "model.layers.13.linear_attn.in_proj_qkv",
560
+ "model.layers.13.linear_attn.in_proj_z",
561
+ "model.layers.13.linear_attn.in_proj_b",
562
+ "model.layers.13.linear_attn.in_proj_a",
563
+ "model.layers.13.mlp.shared_expert_gate",
564
+ "model.layers.14.linear_attn.out_proj",
565
+ "model.layers.14.linear_attn.in_proj_qkv",
566
+ "model.layers.14.linear_attn.in_proj_z",
567
+ "model.layers.14.linear_attn.in_proj_b",
568
+ "model.layers.14.linear_attn.in_proj_a",
569
+ "model.layers.14.mlp.shared_expert_gate",
570
+ "model.layers.15.mlp.shared_expert_gate",
571
+ "model.layers.16.linear_attn.out_proj",
572
+ "model.layers.16.linear_attn.in_proj_qkv",
573
+ "model.layers.16.linear_attn.in_proj_z",
574
+ "model.layers.16.linear_attn.in_proj_b",
575
+ "model.layers.16.linear_attn.in_proj_a",
576
+ "model.layers.16.mlp.shared_expert_gate",
577
+ "model.layers.17.linear_attn.out_proj",
578
+ "model.layers.17.linear_attn.in_proj_qkv",
579
+ "model.layers.17.linear_attn.in_proj_z",
580
+ "model.layers.17.linear_attn.in_proj_b",
581
+ "model.layers.17.linear_attn.in_proj_a",
582
+ "model.layers.17.mlp.shared_expert_gate",
583
+ "model.layers.18.linear_attn.out_proj",
584
+ "model.layers.18.linear_attn.in_proj_qkv",
585
+ "model.layers.18.linear_attn.in_proj_z",
586
+ "model.layers.18.linear_attn.in_proj_b",
587
+ "model.layers.18.linear_attn.in_proj_a",
588
+ "model.layers.18.mlp.shared_expert_gate",
589
+ "model.layers.19.mlp.shared_expert_gate",
590
+ "model.layers.20.linear_attn.out_proj",
591
+ "model.layers.20.linear_attn.in_proj_qkv",
592
+ "model.layers.20.linear_attn.in_proj_z",
593
+ "model.layers.20.linear_attn.in_proj_b",
594
+ "model.layers.20.linear_attn.in_proj_a",
595
+ "model.layers.20.mlp.shared_expert_gate",
596
+ "model.layers.21.linear_attn.out_proj",
597
+ "model.layers.21.linear_attn.in_proj_qkv",
598
+ "model.layers.21.linear_attn.in_proj_z",
599
+ "model.layers.21.linear_attn.in_proj_b",
600
+ "model.layers.21.linear_attn.in_proj_a",
601
+ "model.layers.21.mlp.shared_expert_gate",
602
+ "model.layers.22.linear_attn.out_proj",
603
+ "model.layers.22.linear_attn.in_proj_qkv",
604
+ "model.layers.22.linear_attn.in_proj_z",
605
+ "model.layers.22.linear_attn.in_proj_b",
606
+ "model.layers.22.linear_attn.in_proj_a",
607
+ "model.layers.22.mlp.shared_expert_gate",
608
+ "model.layers.23.mlp.shared_expert_gate",
609
+ "model.layers.24.linear_attn.out_proj",
610
+ "model.layers.24.linear_attn.in_proj_qkv",
611
+ "model.layers.24.linear_attn.in_proj_z",
612
+ "model.layers.24.linear_attn.in_proj_b",
613
+ "model.layers.24.linear_attn.in_proj_a",
614
+ "model.layers.24.mlp.shared_expert_gate",
615
+ "model.layers.25.linear_attn.out_proj",
616
+ "model.layers.25.linear_attn.in_proj_qkv",
617
+ "model.layers.25.linear_attn.in_proj_z",
618
+ "model.layers.25.linear_attn.in_proj_b",
619
+ "model.layers.25.linear_attn.in_proj_a",
620
+ "model.layers.25.mlp.shared_expert_gate",
621
+ "model.layers.26.linear_attn.out_proj",
622
+ "model.layers.26.linear_attn.in_proj_qkv",
623
+ "model.layers.26.linear_attn.in_proj_z",
624
+ "model.layers.26.linear_attn.in_proj_b",
625
+ "model.layers.26.linear_attn.in_proj_a",
626
+ "model.layers.26.mlp.shared_expert_gate",
627
+ "model.layers.27.mlp.shared_expert_gate",
628
+ "model.layers.28.linear_attn.out_proj",
629
+ "model.layers.28.linear_attn.in_proj_qkv",
630
+ "model.layers.28.linear_attn.in_proj_z",
631
+ "model.layers.28.linear_attn.in_proj_b",
632
+ "model.layers.28.linear_attn.in_proj_a",
633
+ "model.layers.28.mlp.shared_expert_gate",
634
+ "model.layers.29.linear_attn.out_proj",
635
+ "model.layers.29.linear_attn.in_proj_qkv",
636
+ "model.layers.29.linear_attn.in_proj_z",
637
+ "model.layers.29.linear_attn.in_proj_b",
638
+ "model.layers.29.linear_attn.in_proj_a",
639
+ "model.layers.29.mlp.shared_expert_gate",
640
+ "model.layers.30.linear_attn.out_proj",
641
+ "model.layers.30.linear_attn.in_proj_qkv",
642
+ "model.layers.30.linear_attn.in_proj_z",
643
+ "model.layers.30.linear_attn.in_proj_b",
644
+ "model.layers.30.linear_attn.in_proj_a",
645
+ "model.layers.30.mlp.shared_expert_gate",
646
+ "model.layers.31.mlp.shared_expert_gate",
647
+ "model.layers.32.linear_attn.out_proj",
648
+ "model.layers.32.linear_attn.in_proj_qkv",
649
+ "model.layers.32.linear_attn.in_proj_z",
650
+ "model.layers.32.linear_attn.in_proj_b",
651
+ "model.layers.32.linear_attn.in_proj_a",
652
+ "model.layers.32.mlp.shared_expert_gate",
653
+ "model.layers.33.linear_attn.out_proj",
654
+ "model.layers.33.linear_attn.in_proj_qkv",
655
+ "model.layers.33.linear_attn.in_proj_z",
656
+ "model.layers.33.linear_attn.in_proj_b",
657
+ "model.layers.33.linear_attn.in_proj_a",
658
+ "model.layers.33.mlp.shared_expert_gate",
659
+ "model.layers.34.linear_attn.out_proj",
660
+ "model.layers.34.linear_attn.in_proj_qkv",
661
+ "model.layers.34.linear_attn.in_proj_z",
662
+ "model.layers.34.linear_attn.in_proj_b",
663
+ "model.layers.34.linear_attn.in_proj_a",
664
+ "model.layers.34.mlp.shared_expert_gate",
665
+ "model.layers.35.mlp.shared_expert_gate",
666
+ "model.layers.36.linear_attn.out_proj",
667
+ "model.layers.36.linear_attn.in_proj_qkv",
668
+ "model.layers.36.linear_attn.in_proj_z",
669
+ "model.layers.36.linear_attn.in_proj_b",
670
+ "model.layers.36.linear_attn.in_proj_a",
671
+ "model.layers.36.mlp.shared_expert_gate",
672
+ "model.layers.37.linear_attn.out_proj",
673
+ "model.layers.37.linear_attn.in_proj_qkv",
674
+ "model.layers.37.linear_attn.in_proj_z",
675
+ "model.layers.37.linear_attn.in_proj_b",
676
+ "model.layers.37.linear_attn.in_proj_a",
677
+ "model.layers.37.mlp.shared_expert_gate",
678
+ "model.layers.38.linear_attn.out_proj",
679
+ "model.layers.38.linear_attn.in_proj_qkv",
680
+ "model.layers.38.linear_attn.in_proj_z",
681
+ "model.layers.38.linear_attn.in_proj_b",
682
+ "model.layers.38.linear_attn.in_proj_a",
683
+ "model.layers.38.mlp.shared_expert_gate",
684
+ "model.layers.39.mlp.shared_expert_gate",
685
+ "model.layers.40.linear_attn.out_proj",
686
+ "model.layers.40.linear_attn.in_proj_qkv",
687
+ "model.layers.40.linear_attn.in_proj_z",
688
+ "model.layers.40.linear_attn.in_proj_b",
689
+ "model.layers.40.linear_attn.in_proj_a",
690
+ "model.layers.40.mlp.shared_expert_gate",
691
+ "model.layers.41.linear_attn.out_proj",
692
+ "model.layers.41.linear_attn.in_proj_qkv",
693
+ "model.layers.41.linear_attn.in_proj_z",
694
+ "model.layers.41.linear_attn.in_proj_b",
695
+ "model.layers.41.linear_attn.in_proj_a",
696
+ "model.layers.41.mlp.shared_expert_gate",
697
+ "model.layers.42.linear_attn.out_proj",
698
+ "model.layers.42.linear_attn.in_proj_qkv",
699
+ "model.layers.42.linear_attn.in_proj_z",
700
+ "model.layers.42.linear_attn.in_proj_b",
701
+ "model.layers.42.linear_attn.in_proj_a",
702
+ "model.layers.42.mlp.shared_expert_gate",
703
+ "model.layers.43.mlp.shared_expert_gate",
704
+ "model.layers.44.linear_attn.out_proj",
705
+ "model.layers.44.linear_attn.in_proj_qkv",
706
+ "model.layers.44.linear_attn.in_proj_z",
707
+ "model.layers.44.linear_attn.in_proj_b",
708
+ "model.layers.44.linear_attn.in_proj_a",
709
+ "model.layers.44.mlp.shared_expert_gate",
710
+ "model.layers.45.linear_attn.out_proj",
711
+ "model.layers.45.linear_attn.in_proj_qkv",
712
+ "model.layers.45.linear_attn.in_proj_z",
713
+ "model.layers.45.linear_attn.in_proj_b",
714
+ "model.layers.45.linear_attn.in_proj_a",
715
+ "model.layers.45.mlp.shared_expert_gate",
716
+ "model.layers.46.linear_attn.out_proj",
717
+ "model.layers.46.linear_attn.in_proj_qkv",
718
+ "model.layers.46.linear_attn.in_proj_z",
719
+ "model.layers.46.linear_attn.in_proj_b",
720
+ "model.layers.46.linear_attn.in_proj_a",
721
+ "model.layers.46.mlp.shared_expert_gate",
722
+ "model.layers.47.mlp.shared_expert_gate",
723
+ "model.layers.48.linear_attn.out_proj",
724
+ "model.layers.48.linear_attn.in_proj_qkv",
725
+ "model.layers.48.linear_attn.in_proj_z",
726
+ "model.layers.48.linear_attn.in_proj_b",
727
+ "model.layers.48.linear_attn.in_proj_a",
728
+ "model.layers.48.mlp.shared_expert_gate",
729
+ "model.layers.49.linear_attn.out_proj",
730
+ "model.layers.49.linear_attn.in_proj_qkv",
731
+ "model.layers.49.linear_attn.in_proj_z",
732
+ "model.layers.49.linear_attn.in_proj_b",
733
+ "model.layers.49.linear_attn.in_proj_a",
734
+ "model.layers.49.mlp.shared_expert_gate",
735
+ "model.layers.50.linear_attn.out_proj",
736
+ "model.layers.50.linear_attn.in_proj_qkv",
737
+ "model.layers.50.linear_attn.in_proj_z",
738
+ "model.layers.50.linear_attn.in_proj_b",
739
+ "model.layers.50.linear_attn.in_proj_a",
740
+ "model.layers.50.mlp.shared_expert_gate",
741
+ "model.layers.51.mlp.shared_expert_gate",
742
+ "model.layers.52.linear_attn.out_proj",
743
+ "model.layers.52.linear_attn.in_proj_qkv",
744
+ "model.layers.52.linear_attn.in_proj_z",
745
+ "model.layers.52.linear_attn.in_proj_b",
746
+ "model.layers.52.linear_attn.in_proj_a",
747
+ "model.layers.52.mlp.shared_expert_gate",
748
+ "model.layers.53.linear_attn.out_proj",
749
+ "model.layers.53.linear_attn.in_proj_qkv",
750
+ "model.layers.53.linear_attn.in_proj_z",
751
+ "model.layers.53.linear_attn.in_proj_b",
752
+ "model.layers.53.linear_attn.in_proj_a",
753
+ "model.layers.53.mlp.shared_expert_gate",
754
+ "model.layers.54.linear_attn.out_proj",
755
+ "model.layers.54.linear_attn.in_proj_qkv",
756
+ "model.layers.54.linear_attn.in_proj_z",
757
+ "model.layers.54.linear_attn.in_proj_b",
758
+ "model.layers.54.linear_attn.in_proj_a",
759
+ "model.layers.54.mlp.shared_expert_gate",
760
+ "model.layers.55.mlp.shared_expert_gate",
761
+ "model.layers.56.linear_attn.out_proj",
762
+ "model.layers.56.linear_attn.in_proj_qkv",
763
+ "model.layers.56.linear_attn.in_proj_z",
764
+ "model.layers.56.linear_attn.in_proj_b",
765
+ "model.layers.56.linear_attn.in_proj_a",
766
+ "model.layers.56.mlp.shared_expert_gate",
767
+ "model.layers.57.linear_attn.out_proj",
768
+ "model.layers.57.linear_attn.in_proj_qkv",
769
+ "model.layers.57.linear_attn.in_proj_z",
770
+ "model.layers.57.linear_attn.in_proj_b",
771
+ "model.layers.57.linear_attn.in_proj_a",
772
+ "model.layers.57.mlp.shared_expert_gate",
773
+ "model.layers.58.linear_attn.out_proj",
774
+ "model.layers.58.linear_attn.in_proj_qkv",
775
+ "model.layers.58.linear_attn.in_proj_z",
776
+ "model.layers.58.linear_attn.in_proj_b",
777
+ "model.layers.58.linear_attn.in_proj_a",
778
+ "model.layers.58.mlp.shared_expert_gate",
779
+ "model.layers.59.mlp.shared_expert_gate",
780
+ "lm_head"
781
+ ],
782
+ "kv_cache_scheme": null,
783
+ "quant_method": "compressed-tensors",
784
+ "quantization_status": "compressed",
785
+ "sparsity_config": {},
786
+ "transform_config": {},
787
+ "version": "0.13.1.a20260212"
788
+ }
789
+ }
generation_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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": 0.6,
10
+ "top_k": 20,
11
+ "top_p": 0.95,
12
+ "transformers_version": "5.2.0.dev0"
13
+ }
model-00001-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c774a786c4c8486a0580947177d8646b0dcf2bfe2631609b072a4b7d62b717b7
3
+ size 50003529056
model-00002-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a192cbf00aa6e3a97a0bbe314d3e4b9f3f0d33f421c84b29be6953fee7e81f04
3
+ size 49973816648
model-00003-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:90eb6ba88baf2ddfd4173e79c02fc16090438c52556aa9d4061ab83164cfcd06
3
+ size 50010025384
model-00004-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3a84b647cba5fe1a4d5875ec29177c89cfc8d5ed6432699017e0496d0b08612a
3
+ size 50011309608
model-00005-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:032ab8240167076ba901a5b2b5d5fdeeecb9f205854ea8f953051abd93f51a59
3
+ size 33739493320
model.safetensors.index.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0ace404746ed05dcd3e98498b01c94c05a22b79ed3e6e3f7dba9934ff3d83ec1
3
+ size 42738125
recipe.yaml ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ default_stage:
2
+ default_modifiers:
3
+ QuantizationModifier:
4
+ targets: [Linear]
5
+ ignore: [lm_head, 're:.*mlp.gate$', 're:.*mlp.shared_expert_gate$', 're:.*linear_attn.*']
6
+ scheme: NVFP4
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:87a7830d63fcf43bf241c3c5242e96e62dd3fdc29224ca26fed8ea333db72de4
3
+ size 19989343
tokenizer_config.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "audio_bos_token": "<|audio_start|>",
4
+ "audio_eos_token": "<|audio_end|>",
5
+ "audio_token": "<|audio_pad|>",
6
+ "backend": "tokenizers",
7
+ "bos_token": null,
8
+ "clean_up_tokenization_spaces": false,
9
+ "eos_token": "<|im_end|>",
10
+ "errors": "replace",
11
+ "image_token": "<|image_pad|>",
12
+ "is_local": false,
13
+ "model_max_length": 262144,
14
+ "model_specific_special_tokens": {
15
+ "audio_bos_token": "<|audio_start|>",
16
+ "audio_eos_token": "<|audio_end|>",
17
+ "audio_token": "<|audio_pad|>",
18
+ "image_token": "<|image_pad|>",
19
+ "video_token": "<|video_pad|>",
20
+ "vision_bos_token": "<|vision_start|>",
21
+ "vision_eos_token": "<|vision_end|>"
22
+ },
23
+ "pad_token": "<|endoftext|>",
24
+ "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+",
25
+ "split_special_tokens": false,
26
+ "tokenizer_class": "PreTrainedTokenizerFast",
27
+ "unk_token": null,
28
+ "video_token": "<|video_pad|>",
29
+ "vision_bos_token": "<|vision_start|>",
30
+ "vision_eos_token": "<|vision_end|>"
31
+ }