Hugging Face's logo

zecanard
/
gemma-4-31B-it-Claude-Opus-Distilled-v2-MLX-4bit-nvfp4

Image-Text-to-Text
MLX
Safetensors
Transformers
English
gemma4
text-generation-inference
unsloth
reasoning
conversational
4-bit precision
Model card Files
xet
 Community
zecanard commited on
Commit
f272239
·
verified ·
1 Parent(s): fd22640

Add files using upload-large-folder tool

Browse files
Files changed (14) hide show
  1. .gitattributes +1 -0
  2. README.md +43 -0
  3. chat_template.jinja +356 -0
  4. config.json +1628 -0
  5. model-00001-of-00006.safetensors +2 -2
  6. model-00002-of-00006.safetensors +3 -0
  7. model-00003-of-00006.safetensors +3 -0
  8. model-00004-of-00006.safetensors +3 -0
  9. model-00005-of-00006.safetensors +3 -0
  10. model-00006-of-00006.safetensors +3 -0
  11. model.safetensors.index.json +0 -0
  12. processor_config.json +42 -0
  13. tokenizer.json +3 -0
  14. tokenizer_config.json +95 -0
.gitattributes CHANGED
@@ -33,3 +33,4 @@ 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
README.md ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: TeichAI/gemma-4-31B-it-Claude-Opus-Distill-v2
3
+ tags:
4
+ - text-generation-inference
5
+ - transformers
6
+ - unsloth
7
+ - gemma4
8
+ - reasoning
9
+ - mlx
10
+ license: apache-2.0
11
+ datasets:
12
+ - TeichAI/Claude-Opus-4.6-Reasoning-887x
13
+ - TeichAI/claude-4.5-opus-high-reasoning-250x
14
+ - Crownelius/Opus-4.6-Reasoning-2100x-formatted
15
+ pipeline_tag: image-text-to-text
16
+ library_name: mlx
17
+ ---
18
+ # 🦆 zecanard/gemma-4-31B-it-Claude-Opus-Distilled-v2-MLX-4bit-nvfp4
19
+
20
+ [This model](https://huggingface.co/zecanard/gemma-4-31B-it-Claude-Opus-Distilled-v2-MLX-4bit-nvfp4) was converted to MLX from [`TeichAI/gemma-4-31B-it-Claude-Opus-Distill-v2`](https://huggingface.co/TeichAI/gemma-4-31B-it-Claude-Opus-Distill-v2) using `mlx-vlm` version **0.4.4**.
21
+ Please refer to the [original model card](https://huggingface.co/TeichAI/gemma-4-31B-it-Claude-Opus-Distill-v2) for more details.
22
+
23
+ ## 🌟 Quality
24
+
25
+ Quantized vision language model with **7.377 bits per weight**.
26
+
27
+ `mlx_vlm.convert --hf-path TeichAI/gemma-4-31B-it-Claude-Opus-Distill-v2 --quantize --q-bits 4 --q-group-size 16 --q-mode nvfp4`
28
+
29
+ ## 🛠️ Customizations
30
+
31
+ This quant is aware of the current date, and also enables thinking (if available). You may disable this behavior by deleting the following line from the chat template:
32
+
33
+ `{%- set enable_thinking = true %}`
34
+
35
+ ## 🖥️ Use with `mlx`
36
+
37
+ ```bash
38
+ pip install -U mlx-vlm
39
+ ```
40
+
41
+ ```bash
42
+ mlx_vlm.generate --model zecanard/gemma-4-31B-it-Claude-Opus-Distilled-v2-MLX-4bit-nvfp4 --max-tokens 100 --temperature 0.0 --prompt "Describe this image." --image <path_to_image>
43
+ ```
chat_template.jinja ADDED
@@ -0,0 +1,356 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {#- Template customizations by @zecanard. #}
2
+ {%- set enable_thinking = true %}
3
+ {%- if strftime_now is defined %}
4
+ {%- set currentDate = strftime_now("%Y-%m-%d %G:%i:%s") %}
5
+ {%- elif not currentDate is defined %}
6
+ {%- set currentDate = "2026-04-13" %}
7
+ {%- endif %}
8
+ {{- "The current date is: " + currentDate + ".\n\n" }}
9
+ {#- Template customizations by @zecanard. #}
10
+ {%- macro format_parameters(properties, required) -%}
11
+ {%- set standard_keys = ['description', 'type', 'properties', 'required', 'nullable'] -%}
12
+ {%- set ns = namespace(found_first=false) -%}
13
+ {%- for key, value in properties | dictsort -%}
14
+ {%- set add_comma = false -%}
15
+ {%- if key not in standard_keys -%}
16
+ {%- if ns.found_first %},{% endif -%}
17
+ {%- set ns.found_first = true -%}
18
+ {{ key }}:{
19
+ {%- if value['description'] -%}
20
+ description:<|"|>{{ value['description'] }}<|"|>
21
+ {%- set add_comma = true -%}
22
+ {%- endif -%}
23
+ {%- if value['type'] | upper == 'STRING' -%}
24
+ {%- if value['enum'] -%}
25
+ {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
26
+ enum:{{ format_argument(value['enum']) }}
27
+ {%- endif -%}
28
+ {%- elif value['type'] | upper == 'ARRAY' -%}
29
+ {%- if value['items'] is mapping and value['items'] -%}
30
+ {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
31
+ items:{
32
+ {%- set ns_items = namespace(found_first=false) -%}
33
+ {%- for item_key, item_value in value['items'] | dictsort -%}
34
+ {%- if item_value is not none -%}
35
+ {%- if ns_items.found_first %},{% endif -%}
36
+ {%- set ns_items.found_first = true -%}
37
+ {%- if item_key == 'properties' -%}
38
+ properties:{
39
+ {%- if item_value is mapping -%}
40
+ {{- format_parameters(item_value, value['items']['required'] | default([])) -}}
41
+ {%- endif -%}
42
+ }
43
+ {%- elif item_key == 'required' -%}
44
+ required:[
45
+ {%- for req_item in item_value -%}
46
+ <|"|>{{- req_item -}}<|"|>
47
+ {%- if not loop.last %},{% endif -%}
48
+ {%- endfor -%}
49
+ ]
50
+ {%- elif item_key == 'type' -%}
51
+ {%- if item_value is string -%}
52
+ type:{{ format_argument(item_value | upper) }}
53
+ {%- else -%}
54
+ type:{{ format_argument(item_value | map('upper') | list) }}
55
+ {%- endif -%}
56
+ {%- else -%}
57
+ {{ item_key }}:{{ format_argument(item_value) }}
58
+ {%- endif -%}
59
+ {%- endif -%}
60
+ {%- endfor -%}
61
+ }
62
+ {%- endif -%}
63
+ {%- endif -%}
64
+ {%- if value['nullable'] %}
65
+ {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
66
+ nullable:true
67
+ {%- endif -%}
68
+ {%- if value['type'] | upper == 'OBJECT' -%}
69
+ {%- if value['properties'] is defined and value['properties'] is mapping -%}
70
+ {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
71
+ properties:{
72
+ {{- format_parameters(value['properties'], value['required'] | default([])) -}}
73
+ }
74
+ {%- elif value is mapping -%}
75
+ {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
76
+ properties:{
77
+ {{- format_parameters(value, value['required'] | default([])) -}}
78
+ }
79
+ {%- endif -%}
80
+ {%- if value['required'] -%}
81
+ {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
82
+ required:[
83
+ {%- for item in value['required'] | default([]) -%}
84
+ <|"|>{{- item -}}<|"|>
85
+ {%- if not loop.last %},{% endif -%}
86
+ {%- endfor -%}
87
+ ]
88
+ {%- endif -%}
89
+ {%- endif -%}
90
+ {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
91
+ type:<|"|>{{ value['type'] | upper }}<|"|>}
92
+ {%- endif -%}
93
+ {%- endfor -%}
94
+ {%- endmacro -%}
95
+ {%- macro format_function_declaration(tool_data) -%}
96
+ declaration:{{- tool_data['function']['name'] -}}{description:<|"|>{{- tool_data['function']['description'] -}}<|"|>
97
+ {%- set params = tool_data['function']['parameters'] -%}
98
+ {%- if params -%}
99
+ ,parameters:{
100
+ {%- if params['properties'] -%}
101
+ properties:{ {{- format_parameters(params['properties'], params['required']) -}} },
102
+ {%- endif -%}
103
+ {%- if params['required'] -%}
104
+ required:[
105
+ {%- for item in params['required'] -%}
106
+ <|"|>{{- item -}}<|"|>
107
+ {{- ',' if not loop.last -}}
108
+ {%- endfor -%}
109
+ ],
110
+ {%- endif -%}
111
+ {%- if params['type'] -%}
112
+ type:<|"|>{{- params['type'] | upper -}}<|"|>}
113
+ {%- endif -%}
114
+ {%- endif -%}
115
+ {%- if 'response' in tool_data['function'] -%}
116
+ {%- set response_declaration = tool_data['function']['response'] -%}
117
+ ,response:{
118
+ {%- if response_declaration['description'] -%}
119
+ description:<|"|>{{- response_declaration['description'] -}}<|"|>,
120
+ {%- endif -%}
121
+ {%- if response_declaration['type'] | upper == 'OBJECT' -%}
122
+ type:<|"|>{{- response_declaration['type'] | upper -}}<|"|>}
123
+ {%- endif -%}
124
+ {%- endif -%}
125
+ }
126
+ {%- endmacro -%}
127
+ {%- macro format_argument(argument, escape_keys=True) -%}
128
+ {%- if argument is string -%}
129
+ {{- '<|"|>' + argument + '<|"|>' -}}
130
+ {%- elif argument is boolean -%}
131
+ {{- 'true' if argument else 'false' -}}
132
+ {%- elif argument is mapping -%}
133
+ {{- '{' -}}
134
+ {%- set ns = namespace(found_first=false) -%}
135
+ {%- for key, value in argument | dictsort -%}
136
+ {%- if ns.found_first %},{% endif -%}
137
+ {%- set ns.found_first = true -%}
138
+ {%- if escape_keys -%}
139
+ {{- '<|"|>' + key + '<|"|>' -}}
140
+ {%- else -%}
141
+ {{- key -}}
142
+ {%- endif -%}
143
+ :{{- format_argument(value, escape_keys=escape_keys) -}}
144
+ {%- endfor -%}
145
+ {{- '}' -}}
146
+ {%- elif argument is sequence -%}
147
+ {{- '[' -}}
148
+ {%- for item in argument -%}
149
+ {{- format_argument(item, escape_keys=escape_keys) -}}
150
+ {%- if not loop.last %},{% endif -%}
151
+ {%- endfor -%}
152
+ {{- ']' -}}
153
+ {%- else -%}
154
+ {{- argument -}}
155
+ {%- endif -%}
156
+ {%- endmacro -%}
157
+ {%- macro strip_thinking(text) -%}
158
+ {%- set ns = namespace(result='') -%}
159
+ {%- for part in text.split('<channel|>') -%}
160
+ {%- if '<|channel>' in part -%}
161
+ {%- set ns.result = ns.result + part.split('<|channel>')[0] -%}
162
+ {%- else -%}
163
+ {%- set ns.result = ns.result + part -%}
164
+ {%- endif -%}
165
+ {%- endfor -%}
166
+ {{- ns.result | trim -}}
167
+ {%- endmacro -%}
168
+
169
+ {%- macro format_tool_response_block(tool_name, response) -%}
170
+ {{- '<|tool_response>' -}}
171
+ {%- if response is mapping -%}
172
+ {{- 'response:' + tool_name + '{' -}}
173
+ {%- for key, value in response | dictsort -%}
174
+ {{- key -}}:{{- format_argument(value, escape_keys=False) -}}
175
+ {%- if not loop.last %},{% endif -%}
176
+ {%- endfor -%}
177
+ {{- '}' -}}
178
+ {%- else -%}
179
+ {{- 'response:' + tool_name + '{value:' + format_argument(response, escape_keys=False) + '}' -}}
180
+ {%- endif -%}
181
+ {{- '<tool_response|>' -}}
182
+ {%- endmacro -%}
183
+
184
+ {%- set ns = namespace(prev_message_type=None) -%}
185
+ {%- set loop_messages = messages -%}
186
+ {{- bos_token -}}
187
+ {#- Handle System/Tool Definitions Block -#}
188
+ {%- if enable_thinking | default(true) or tools or messages[0]['role'] in ['system', 'developer'] -%}
189
+ {{- '<|turn>system\n' -}}
190
+
191
+ {#- Inject Thinking token at the very top of the FIRST system turn -#}
192
+ {%- if enable_thinking | default(true) -%}
193
+ {{- '<|think|>\n' -}}
194
+ {%- set ns.prev_message_type = 'think' -%}
195
+ {%- endif -%}
196
+
197
+ {%- if messages[0]['role'] in ['system', 'developer'] -%}
198
+ {{- messages[0]['content'] | trim -}}
199
+ {%- set loop_messages = messages[1:] -%}
200
+ {%- endif -%}
201
+
202
+ {%- if tools -%}
203
+ {%- for tool in tools %}
204
+ {{- '<|tool>' -}}
205
+ {{- format_function_declaration(tool) | trim -}}
206
+ {{- '<tool|>' -}}
207
+ {%- endfor %}
208
+ {%- set ns.prev_message_type = 'tool' -%}
209
+ {%- endif -%}
210
+
211
+ {{- '<turn|>\n' -}}
212
+ {%- endif %}
213
+
214
+ {#- Pre-scan: find last user message index for reasoning guard -#}
215
+ {%- set ns_turn = namespace(last_user_idx=-1) -%}
216
+ {%- for i in range(loop_messages | length) -%}
217
+ {%- if loop_messages[i]['role'] == 'user' -%}
218
+ {%- set ns_turn.last_user_idx = i -%}
219
+ {%- endif -%}
220
+ {%- endfor -%}
221
+
222
+ {#- Loop through messages -#}
223
+ {%- for message in loop_messages -%}
224
+ {%- if message['role'] != 'tool' -%}
225
+ {%- set ns.prev_message_type = None -%}
226
+ {%- set role = 'model' if message['role'] == 'assistant' else message['role'] -%}
227
+ {#- Detect continuation: suppress duplicate <|turn>model when previous non-tool message was also assistant -#}
228
+ {%- set prev_nt = namespace(role=None, found=false) -%}
229
+ {%- if loop.index0 > 0 -%}
230
+ {%- for j in range(loop.index0 - 1, -1, -1) -%}
231
+ {%- if not prev_nt.found -%}
232
+ {%- if loop_messages[j]['role'] != 'tool' -%}
233
+ {%- set prev_nt.role = loop_messages[j]['role'] -%}
234
+ {%- set prev_nt.found = true -%}
235
+ {%- endif -%}
236
+ {%- endif -%}
237
+ {%- endfor -%}
238
+ {%- endif -%}
239
+ {%- set continue_same_model_turn = (role == 'model' and prev_nt.role == 'assistant') -%}
240
+ {%- if not continue_same_model_turn -%}
241
+ {{- '<|turn>' + role + '\n' }}
242
+ {%- endif -%}
243
+
244
+ {#- Render reasoning/reasoning_content as thinking channel -#}
245
+ {%- set thinking_text = message.get('reasoning') or message.get('reasoning_content') -%}
246
+ {%- if thinking_text and loop.index0 > ns_turn.last_user_idx and message.get('tool_calls') -%}
247
+ {{- '<|channel>thought\n' + thinking_text + '\n<channel|>' -}}
248
+ {%- endif -%}
249
+
250
+ {%- if message['tool_calls'] -%}
251
+ {%- for tool_call in message['tool_calls'] -%}
252
+ {%- set function = tool_call['function'] -%}
253
+ {{- '<|tool_call>call:' + function['name'] + '{' -}}
254
+ {%- if function['arguments'] is mapping -%}
255
+ {%- set ns_args = namespace(found_first=false) -%}
256
+ {%- for key, value in function['arguments'] | dictsort -%}
257
+ {%- if ns_args.found_first %},{% endif -%}
258
+ {%- set ns_args.found_first = true -%}
259
+ {{- key -}}:{{- format_argument(value, escape_keys=False) -}}
260
+ {%- endfor -%}
261
+ {%- elif function['arguments'] is string -%}
262
+ {{- function['arguments'] -}}
263
+ {%- endif -%}
264
+ {{- '}<tool_call|>' -}}
265
+ {%- endfor -%}
266
+ {%- set ns.prev_message_type = 'tool_call' -%}
267
+ {%- endif -%}
268
+
269
+ {%- set ns_tr_out = namespace(flag=false) -%}
270
+ {%- if message.get('tool_responses') -%}
271
+ {#- Legacy: tool_responses embedded on the assistant message (Google/Gemma native) -#}
272
+ {%- for tool_response in message['tool_responses'] -%}
273
+ {{- format_tool_response_block(tool_response['name'] | default('unknown'), tool_response['response']) -}}
274
+ {%- set ns_tr_out.flag = true -%}
275
+ {%- set ns.prev_message_type = 'tool_response' -%}
276
+ {%- endfor -%}
277
+ {%- elif message.get('tool_calls') -%}
278
+ {#- OpenAI Chat Completions: forward-scan consecutive role:tool messages -#}
279
+ {%- set ns_tool_scan = namespace(stopped=false) -%}
280
+ {%- for k in range(loop.index0 + 1, loop_messages | length) -%}
281
+ {%- if ns_tool_scan.stopped -%}
282
+ {%- elif loop_messages[k]['role'] != 'tool' -%}
283
+ {%- set ns_tool_scan.stopped = true -%}
284
+ {%- else -%}
285
+ {%- set follow = loop_messages[k] -%}
286
+ {#- Resolve tool_call_id to function name -#}
287
+ {%- set ns_tname = namespace(name=follow.get('name') | default('unknown')) -%}
288
+ {%- for tc in message['tool_calls'] -%}
289
+ {%- if tc.get('id') == follow.get('tool_call_id') -%}
290
+ {%- set ns_tname.name = tc['function']['name'] -%}
291
+ {%- endif -%}
292
+ {%- endfor -%}
293
+ {#- Handle content as string or content-parts array -#}
294
+ {%- set tool_body = follow.get('content') -%}
295
+ {%- if tool_body is string -%}
296
+ {{- format_tool_response_block(ns_tname.name, tool_body) -}}
297
+ {%- elif tool_body is sequence and tool_body is not string -%}
298
+ {%- set ns_txt = namespace(s='') -%}
299
+ {%- for part in tool_body -%}
300
+ {%- if part.get('type') == 'text' -%}
301
+ {%- set ns_txt.s = ns_txt.s + (part.get('text') | default('')) -%}
302
+ {%- endif -%}
303
+ {%- endfor -%}
304
+ {{- format_tool_response_block(ns_tname.name, ns_txt.s) -}}
305
+ {%- else -%}
306
+ {{- format_tool_response_block(ns_tname.name, tool_body) -}}
307
+ {%- endif -%}
308
+ {%- set ns_tr_out.flag = true -%}
309
+ {%- set ns.prev_message_type = 'tool_response' -%}
310
+ {%- endif -%}
311
+ {%- endfor -%}
312
+ {%- endif -%}
313
+
314
+ {%- if message['content'] is string -%}
315
+ {%- if role == 'model' -%}
316
+ {{- strip_thinking(message['content']) -}}
317
+ {%- else -%}
318
+ {{- message['content'] | trim -}}
319
+ {%- endif -%}
320
+ {%- elif message['content'] is sequence -%}
321
+ {%- for item in message['content'] -%}
322
+ {%- if item['type'] == 'text' -%}
323
+ {%- if role == 'model' -%}
324
+ {{- strip_thinking(item['text']) -}}
325
+ {%- else -%}
326
+ {{- item['text'] | trim -}}
327
+ {%- endif -%}
328
+ {%- elif item['type'] == 'image' -%}
329
+ {{- '<|image|>' -}}
330
+ {%- set ns.prev_message_type = 'image' -%}
331
+ {%- elif item['type'] == 'audio' -%}
332
+ {{- '<|audio|>' -}}
333
+ {%- set ns.prev_message_type = 'audio' -%}
334
+ {%- elif item['type'] == 'video' -%}
335
+ {{- '<|video|>' -}}
336
+ {%- set ns.prev_message_type = 'video' -%}
337
+ {%- endif -%}
338
+ {%- endfor -%}
339
+ {%- endif -%}
340
+
341
+ {%- if ns.prev_message_type == 'tool_call' and not ns_tr_out.flag -%}
342
+ {{- '<|tool_response>' -}}
343
+ {%- elif not (ns_tr_out.flag and not message.get('content')) -%}
344
+ {{- '<turn|>\n' -}}
345
+ {%- endif -%}
346
+ {%- endif -%}
347
+ {%- endfor -%}
348
+
349
+ {%- if add_generation_prompt -%}
350
+ {%- if ns.prev_message_type != 'tool_response' and ns.prev_message_type != 'tool_call' -%}
351
+ {{- '<|turn>model\n' -}}
352
+ {%- if not enable_thinking | default(true) -%}
353
+ {{- '<|channel>thought\n<channel|>' -}}
354
+ {%- endif -%}
355
+ {%- endif -%}
356
+ {%- endif -%}
config.json ADDED
@@ -0,0 +1,1628 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Gemma4ForConditionalGeneration"
4
+ ],
5
+ "audio_config": null,
6
+ "audio_token_id": 258881,
7
+ "boa_token_id": 256000,
8
+ "boi_token_id": 255999,
9
+ "bos_token_id": 2,
10
+ "eoa_token_id": 258883,
11
+ "eoa_token_index": 258883,
12
+ "eoi_token_id": 258882,
13
+ "eos_token_id": 106,
14
+ "image_token_id": 258880,
15
+ "initializer_range": 0.02,
16
+ "model_name": "unsloth/gemma-4-31B-it",
17
+ "model_type": "gemma4",
18
+ "pad_token_id": 0,
19
+ "quantization": {
20
+ "group_size": 16,
21
+ "bits": 4,
22
+ "mode": "nvfp4",
23
+ "language_model.model.layers.0.mlp.gate_proj": {
24
+ "group_size": 64,
25
+ "bits": 8
26
+ },
27
+ "language_model.model.layers.0.mlp.down_proj": {
28
+ "group_size": 64,
29
+ "bits": 8
30
+ },
31
+ "language_model.model.layers.0.mlp.up_proj": {
32
+ "group_size": 64,
33
+ "bits": 8
34
+ },
35
+ "language_model.model.layers.1.mlp.gate_proj": {
36
+ "group_size": 64,
37
+ "bits": 8
38
+ },
39
+ "language_model.model.layers.1.mlp.down_proj": {
40
+ "group_size": 64,
41
+ "bits": 8
42
+ },
43
+ "language_model.model.layers.1.mlp.up_proj": {
44
+ "group_size": 64,
45
+ "bits": 8
46
+ },
47
+ "language_model.model.layers.2.mlp.gate_proj": {
48
+ "group_size": 64,
49
+ "bits": 8
50
+ },
51
+ "language_model.model.layers.2.mlp.down_proj": {
52
+ "group_size": 64,
53
+ "bits": 8
54
+ },
55
+ "language_model.model.layers.2.mlp.up_proj": {
56
+ "group_size": 64,
57
+ "bits": 8
58
+ },
59
+ "language_model.model.layers.3.mlp.gate_proj": {
60
+ "group_size": 64,
61
+ "bits": 8
62
+ },
63
+ "language_model.model.layers.3.mlp.down_proj": {
64
+ "group_size": 64,
65
+ "bits": 8
66
+ },
67
+ "language_model.model.layers.3.mlp.up_proj": {
68
+ "group_size": 64,
69
+ "bits": 8
70
+ },
71
+ "language_model.model.layers.4.mlp.gate_proj": {
72
+ "group_size": 64,
73
+ "bits": 8
74
+ },
75
+ "language_model.model.layers.4.mlp.down_proj": {
76
+ "group_size": 64,
77
+ "bits": 8
78
+ },
79
+ "language_model.model.layers.4.mlp.up_proj": {
80
+ "group_size": 64,
81
+ "bits": 8
82
+ },
83
+ "language_model.model.layers.5.mlp.gate_proj": {
84
+ "group_size": 64,
85
+ "bits": 8
86
+ },
87
+ "language_model.model.layers.5.mlp.down_proj": {
88
+ "group_size": 64,
89
+ "bits": 8
90
+ },
91
+ "language_model.model.layers.5.mlp.up_proj": {
92
+ "group_size": 64,
93
+ "bits": 8
94
+ },
95
+ "language_model.model.layers.6.mlp.gate_proj": {
96
+ "group_size": 64,
97
+ "bits": 8
98
+ },
99
+ "language_model.model.layers.6.mlp.down_proj": {
100
+ "group_size": 64,
101
+ "bits": 8
102
+ },
103
+ "language_model.model.layers.6.mlp.up_proj": {
104
+ "group_size": 64,
105
+ "bits": 8
106
+ },
107
+ "language_model.model.layers.7.mlp.gate_proj": {
108
+ "group_size": 64,
109
+ "bits": 8
110
+ },
111
+ "language_model.model.layers.7.mlp.down_proj": {
112
+ "group_size": 64,
113
+ "bits": 8
114
+ },
115
+ "language_model.model.layers.7.mlp.up_proj": {
116
+ "group_size": 64,
117
+ "bits": 8
118
+ },
119
+ "language_model.model.layers.8.mlp.gate_proj": {
120
+ "group_size": 64,
121
+ "bits": 8
122
+ },
123
+ "language_model.model.layers.8.mlp.down_proj": {
124
+ "group_size": 64,
125
+ "bits": 8
126
+ },
127
+ "language_model.model.layers.8.mlp.up_proj": {
128
+ "group_size": 64,
129
+ "bits": 8
130
+ },
131
+ "language_model.model.layers.9.mlp.gate_proj": {
132
+ "group_size": 64,
133
+ "bits": 8
134
+ },
135
+ "language_model.model.layers.9.mlp.down_proj": {
136
+ "group_size": 64,
137
+ "bits": 8
138
+ },
139
+ "language_model.model.layers.9.mlp.up_proj": {
140
+ "group_size": 64,
141
+ "bits": 8
142
+ },
143
+ "language_model.model.layers.10.mlp.gate_proj": {
144
+ "group_size": 64,
145
+ "bits": 8
146
+ },
147
+ "language_model.model.layers.10.mlp.down_proj": {
148
+ "group_size": 64,
149
+ "bits": 8
150
+ },
151
+ "language_model.model.layers.10.mlp.up_proj": {
152
+ "group_size": 64,
153
+ "bits": 8
154
+ },
155
+ "language_model.model.layers.11.mlp.gate_proj": {
156
+ "group_size": 64,
157
+ "bits": 8
158
+ },
159
+ "language_model.model.layers.11.mlp.down_proj": {
160
+ "group_size": 64,
161
+ "bits": 8
162
+ },
163
+ "language_model.model.layers.11.mlp.up_proj": {
164
+ "group_size": 64,
165
+ "bits": 8
166
+ },
167
+ "language_model.model.layers.12.mlp.gate_proj": {
168
+ "group_size": 64,
169
+ "bits": 8
170
+ },
171
+ "language_model.model.layers.12.mlp.down_proj": {
172
+ "group_size": 64,
173
+ "bits": 8
174
+ },
175
+ "language_model.model.layers.12.mlp.up_proj": {
176
+ "group_size": 64,
177
+ "bits": 8
178
+ },
179
+ "language_model.model.layers.13.mlp.gate_proj": {
180
+ "group_size": 64,
181
+ "bits": 8
182
+ },
183
+ "language_model.model.layers.13.mlp.down_proj": {
184
+ "group_size": 64,
185
+ "bits": 8
186
+ },
187
+ "language_model.model.layers.13.mlp.up_proj": {
188
+ "group_size": 64,
189
+ "bits": 8
190
+ },
191
+ "language_model.model.layers.14.mlp.gate_proj": {
192
+ "group_size": 64,
193
+ "bits": 8
194
+ },
195
+ "language_model.model.layers.14.mlp.down_proj": {
196
+ "group_size": 64,
197
+ "bits": 8
198
+ },
199
+ "language_model.model.layers.14.mlp.up_proj": {
200
+ "group_size": 64,
201
+ "bits": 8
202
+ },
203
+ "language_model.model.layers.15.mlp.gate_proj": {
204
+ "group_size": 64,
205
+ "bits": 8
206
+ },
207
+ "language_model.model.layers.15.mlp.down_proj": {
208
+ "group_size": 64,
209
+ "bits": 8
210
+ },
211
+ "language_model.model.layers.15.mlp.up_proj": {
212
+ "group_size": 64,
213
+ "bits": 8
214
+ },
215
+ "language_model.model.layers.16.mlp.gate_proj": {
216
+ "group_size": 64,
217
+ "bits": 8
218
+ },
219
+ "language_model.model.layers.16.mlp.down_proj": {
220
+ "group_size": 64,
221
+ "bits": 8
222
+ },
223
+ "language_model.model.layers.16.mlp.up_proj": {
224
+ "group_size": 64,
225
+ "bits": 8
226
+ },
227
+ "language_model.model.layers.17.mlp.gate_proj": {
228
+ "group_size": 64,
229
+ "bits": 8
230
+ },
231
+ "language_model.model.layers.17.mlp.down_proj": {
232
+ "group_size": 64,
233
+ "bits": 8
234
+ },
235
+ "language_model.model.layers.17.mlp.up_proj": {
236
+ "group_size": 64,
237
+ "bits": 8
238
+ },
239
+ "language_model.model.layers.18.mlp.gate_proj": {
240
+ "group_size": 64,
241
+ "bits": 8
242
+ },
243
+ "language_model.model.layers.18.mlp.down_proj": {
244
+ "group_size": 64,
245
+ "bits": 8
246
+ },
247
+ "language_model.model.layers.18.mlp.up_proj": {
248
+ "group_size": 64,
249
+ "bits": 8
250
+ },
251
+ "language_model.model.layers.19.mlp.gate_proj": {
252
+ "group_size": 64,
253
+ "bits": 8
254
+ },
255
+ "language_model.model.layers.19.mlp.down_proj": {
256
+ "group_size": 64,
257
+ "bits": 8
258
+ },
259
+ "language_model.model.layers.19.mlp.up_proj": {
260
+ "group_size": 64,
261
+ "bits": 8
262
+ },
263
+ "language_model.model.layers.20.mlp.gate_proj": {
264
+ "group_size": 64,
265
+ "bits": 8
266
+ },
267
+ "language_model.model.layers.20.mlp.down_proj": {
268
+ "group_size": 64,
269
+ "bits": 8
270
+ },
271
+ "language_model.model.layers.20.mlp.up_proj": {
272
+ "group_size": 64,
273
+ "bits": 8
274
+ },
275
+ "language_model.model.layers.21.mlp.gate_proj": {
276
+ "group_size": 64,
277
+ "bits": 8
278
+ },
279
+ "language_model.model.layers.21.mlp.down_proj": {
280
+ "group_size": 64,
281
+ "bits": 8
282
+ },
283
+ "language_model.model.layers.21.mlp.up_proj": {
284
+ "group_size": 64,
285
+ "bits": 8
286
+ },
287
+ "language_model.model.layers.22.mlp.gate_proj": {
288
+ "group_size": 64,
289
+ "bits": 8
290
+ },
291
+ "language_model.model.layers.22.mlp.down_proj": {
292
+ "group_size": 64,
293
+ "bits": 8
294
+ },
295
+ "language_model.model.layers.22.mlp.up_proj": {
296
+ "group_size": 64,
297
+ "bits": 8
298
+ },
299
+ "language_model.model.layers.23.mlp.gate_proj": {
300
+ "group_size": 64,
301
+ "bits": 8
302
+ },
303
+ "language_model.model.layers.23.mlp.down_proj": {
304
+ "group_size": 64,
305
+ "bits": 8
306
+ },
307
+ "language_model.model.layers.23.mlp.up_proj": {
308
+ "group_size": 64,
309
+ "bits": 8
310
+ },
311
+ "language_model.model.layers.24.mlp.gate_proj": {
312
+ "group_size": 64,
313
+ "bits": 8
314
+ },
315
+ "language_model.model.layers.24.mlp.down_proj": {
316
+ "group_size": 64,
317
+ "bits": 8
318
+ },
319
+ "language_model.model.layers.24.mlp.up_proj": {
320
+ "group_size": 64,
321
+ "bits": 8
322
+ },
323
+ "language_model.model.layers.25.mlp.gate_proj": {
324
+ "group_size": 64,
325
+ "bits": 8
326
+ },
327
+ "language_model.model.layers.25.mlp.down_proj": {
328
+ "group_size": 64,
329
+ "bits": 8
330
+ },
331
+ "language_model.model.layers.25.mlp.up_proj": {
332
+ "group_size": 64,
333
+ "bits": 8
334
+ },
335
+ "language_model.model.layers.26.mlp.gate_proj": {
336
+ "group_size": 64,
337
+ "bits": 8
338
+ },
339
+ "language_model.model.layers.26.mlp.down_proj": {
340
+ "group_size": 64,
341
+ "bits": 8
342
+ },
343
+ "language_model.model.layers.26.mlp.up_proj": {
344
+ "group_size": 64,
345
+ "bits": 8
346
+ },
347
+ "language_model.model.layers.27.mlp.gate_proj": {
348
+ "group_size": 64,
349
+ "bits": 8
350
+ },
351
+ "language_model.model.layers.27.mlp.down_proj": {
352
+ "group_size": 64,
353
+ "bits": 8
354
+ },
355
+ "language_model.model.layers.27.mlp.up_proj": {
356
+ "group_size": 64,
357
+ "bits": 8
358
+ },
359
+ "language_model.model.layers.28.mlp.gate_proj": {
360
+ "group_size": 64,
361
+ "bits": 8
362
+ },
363
+ "language_model.model.layers.28.mlp.down_proj": {
364
+ "group_size": 64,
365
+ "bits": 8
366
+ },
367
+ "language_model.model.layers.28.mlp.up_proj": {
368
+ "group_size": 64,
369
+ "bits": 8
370
+ },
371
+ "language_model.model.layers.29.mlp.gate_proj": {
372
+ "group_size": 64,
373
+ "bits": 8
374
+ },
375
+ "language_model.model.layers.29.mlp.down_proj": {
376
+ "group_size": 64,
377
+ "bits": 8
378
+ },
379
+ "language_model.model.layers.29.mlp.up_proj": {
380
+ "group_size": 64,
381
+ "bits": 8
382
+ },
383
+ "language_model.model.layers.30.mlp.gate_proj": {
384
+ "group_size": 64,
385
+ "bits": 8
386
+ },
387
+ "language_model.model.layers.30.mlp.down_proj": {
388
+ "group_size": 64,
389
+ "bits": 8
390
+ },
391
+ "language_model.model.layers.30.mlp.up_proj": {
392
+ "group_size": 64,
393
+ "bits": 8
394
+ },
395
+ "language_model.model.layers.31.mlp.gate_proj": {
396
+ "group_size": 64,
397
+ "bits": 8
398
+ },
399
+ "language_model.model.layers.31.mlp.down_proj": {
400
+ "group_size": 64,
401
+ "bits": 8
402
+ },
403
+ "language_model.model.layers.31.mlp.up_proj": {
404
+ "group_size": 64,
405
+ "bits": 8
406
+ },
407
+ "language_model.model.layers.32.mlp.gate_proj": {
408
+ "group_size": 64,
409
+ "bits": 8
410
+ },
411
+ "language_model.model.layers.32.mlp.down_proj": {
412
+ "group_size": 64,
413
+ "bits": 8
414
+ },
415
+ "language_model.model.layers.32.mlp.up_proj": {
416
+ "group_size": 64,
417
+ "bits": 8
418
+ },
419
+ "language_model.model.layers.33.mlp.gate_proj": {
420
+ "group_size": 64,
421
+ "bits": 8
422
+ },
423
+ "language_model.model.layers.33.mlp.down_proj": {
424
+ "group_size": 64,
425
+ "bits": 8
426
+ },
427
+ "language_model.model.layers.33.mlp.up_proj": {
428
+ "group_size": 64,
429
+ "bits": 8
430
+ },
431
+ "language_model.model.layers.34.mlp.gate_proj": {
432
+ "group_size": 64,
433
+ "bits": 8
434
+ },
435
+ "language_model.model.layers.34.mlp.down_proj": {
436
+ "group_size": 64,
437
+ "bits": 8
438
+ },
439
+ "language_model.model.layers.34.mlp.up_proj": {
440
+ "group_size": 64,
441
+ "bits": 8
442
+ },
443
+ "language_model.model.layers.35.mlp.gate_proj": {
444
+ "group_size": 64,
445
+ "bits": 8
446
+ },
447
+ "language_model.model.layers.35.mlp.down_proj": {
448
+ "group_size": 64,
449
+ "bits": 8
450
+ },
451
+ "language_model.model.layers.35.mlp.up_proj": {
452
+ "group_size": 64,
453
+ "bits": 8
454
+ },
455
+ "language_model.model.layers.36.mlp.gate_proj": {
456
+ "group_size": 64,
457
+ "bits": 8
458
+ },
459
+ "language_model.model.layers.36.mlp.down_proj": {
460
+ "group_size": 64,
461
+ "bits": 8
462
+ },
463
+ "language_model.model.layers.36.mlp.up_proj": {
464
+ "group_size": 64,
465
+ "bits": 8
466
+ },
467
+ "language_model.model.layers.37.mlp.gate_proj": {
468
+ "group_size": 64,
469
+ "bits": 8
470
+ },
471
+ "language_model.model.layers.37.mlp.down_proj": {
472
+ "group_size": 64,
473
+ "bits": 8
474
+ },
475
+ "language_model.model.layers.37.mlp.up_proj": {
476
+ "group_size": 64,
477
+ "bits": 8
478
+ },
479
+ "language_model.model.layers.38.mlp.gate_proj": {
480
+ "group_size": 64,
481
+ "bits": 8
482
+ },
483
+ "language_model.model.layers.38.mlp.down_proj": {
484
+ "group_size": 64,
485
+ "bits": 8
486
+ },
487
+ "language_model.model.layers.38.mlp.up_proj": {
488
+ "group_size": 64,
489
+ "bits": 8
490
+ },
491
+ "language_model.model.layers.39.mlp.gate_proj": {
492
+ "group_size": 64,
493
+ "bits": 8
494
+ },
495
+ "language_model.model.layers.39.mlp.down_proj": {
496
+ "group_size": 64,
497
+ "bits": 8
498
+ },
499
+ "language_model.model.layers.39.mlp.up_proj": {
500
+ "group_size": 64,
501
+ "bits": 8
502
+ },
503
+ "language_model.model.layers.40.mlp.gate_proj": {
504
+ "group_size": 64,
505
+ "bits": 8
506
+ },
507
+ "language_model.model.layers.40.mlp.down_proj": {
508
+ "group_size": 64,
509
+ "bits": 8
510
+ },
511
+ "language_model.model.layers.40.mlp.up_proj": {
512
+ "group_size": 64,
513
+ "bits": 8
514
+ },
515
+ "language_model.model.layers.41.mlp.gate_proj": {
516
+ "group_size": 64,
517
+ "bits": 8
518
+ },
519
+ "language_model.model.layers.41.mlp.down_proj": {
520
+ "group_size": 64,
521
+ "bits": 8
522
+ },
523
+ "language_model.model.layers.41.mlp.up_proj": {
524
+ "group_size": 64,
525
+ "bits": 8
526
+ },
527
+ "language_model.model.layers.42.mlp.gate_proj": {
528
+ "group_size": 64,
529
+ "bits": 8
530
+ },
531
+ "language_model.model.layers.42.mlp.down_proj": {
532
+ "group_size": 64,
533
+ "bits": 8
534
+ },
535
+ "language_model.model.layers.42.mlp.up_proj": {
536
+ "group_size": 64,
537
+ "bits": 8
538
+ },
539
+ "language_model.model.layers.43.mlp.gate_proj": {
540
+ "group_size": 64,
541
+ "bits": 8
542
+ },
543
+ "language_model.model.layers.43.mlp.down_proj": {
544
+ "group_size": 64,
545
+ "bits": 8
546
+ },
547
+ "language_model.model.layers.43.mlp.up_proj": {
548
+ "group_size": 64,
549
+ "bits": 8
550
+ },
551
+ "language_model.model.layers.44.mlp.gate_proj": {
552
+ "group_size": 64,
553
+ "bits": 8
554
+ },
555
+ "language_model.model.layers.44.mlp.down_proj": {
556
+ "group_size": 64,
557
+ "bits": 8
558
+ },
559
+ "language_model.model.layers.44.mlp.up_proj": {
560
+ "group_size": 64,
561
+ "bits": 8
562
+ },
563
+ "language_model.model.layers.45.mlp.gate_proj": {
564
+ "group_size": 64,
565
+ "bits": 8
566
+ },
567
+ "language_model.model.layers.45.mlp.down_proj": {
568
+ "group_size": 64,
569
+ "bits": 8
570
+ },
571
+ "language_model.model.layers.45.mlp.up_proj": {
572
+ "group_size": 64,
573
+ "bits": 8
574
+ },
575
+ "language_model.model.layers.46.mlp.gate_proj": {
576
+ "group_size": 64,
577
+ "bits": 8
578
+ },
579
+ "language_model.model.layers.46.mlp.down_proj": {
580
+ "group_size": 64,
581
+ "bits": 8
582
+ },
583
+ "language_model.model.layers.46.mlp.up_proj": {
584
+ "group_size": 64,
585
+ "bits": 8
586
+ },
587
+ "language_model.model.layers.47.mlp.gate_proj": {
588
+ "group_size": 64,
589
+ "bits": 8
590
+ },
591
+ "language_model.model.layers.47.mlp.down_proj": {
592
+ "group_size": 64,
593
+ "bits": 8
594
+ },
595
+ "language_model.model.layers.47.mlp.up_proj": {
596
+ "group_size": 64,
597
+ "bits": 8
598
+ },
599
+ "language_model.model.layers.48.mlp.gate_proj": {
600
+ "group_size": 64,
601
+ "bits": 8
602
+ },
603
+ "language_model.model.layers.48.mlp.down_proj": {
604
+ "group_size": 64,
605
+ "bits": 8
606
+ },
607
+ "language_model.model.layers.48.mlp.up_proj": {
608
+ "group_size": 64,
609
+ "bits": 8
610
+ },
611
+ "language_model.model.layers.49.mlp.gate_proj": {
612
+ "group_size": 64,
613
+ "bits": 8
614
+ },
615
+ "language_model.model.layers.49.mlp.down_proj": {
616
+ "group_size": 64,
617
+ "bits": 8
618
+ },
619
+ "language_model.model.layers.49.mlp.up_proj": {
620
+ "group_size": 64,
621
+ "bits": 8
622
+ },
623
+ "language_model.model.layers.50.mlp.gate_proj": {
624
+ "group_size": 64,
625
+ "bits": 8
626
+ },
627
+ "language_model.model.layers.50.mlp.down_proj": {
628
+ "group_size": 64,
629
+ "bits": 8
630
+ },
631
+ "language_model.model.layers.50.mlp.up_proj": {
632
+ "group_size": 64,
633
+ "bits": 8
634
+ },
635
+ "language_model.model.layers.51.mlp.gate_proj": {
636
+ "group_size": 64,
637
+ "bits": 8
638
+ },
639
+ "language_model.model.layers.51.mlp.down_proj": {
640
+ "group_size": 64,
641
+ "bits": 8
642
+ },
643
+ "language_model.model.layers.51.mlp.up_proj": {
644
+ "group_size": 64,
645
+ "bits": 8
646
+ },
647
+ "language_model.model.layers.52.mlp.gate_proj": {
648
+ "group_size": 64,
649
+ "bits": 8
650
+ },
651
+ "language_model.model.layers.52.mlp.down_proj": {
652
+ "group_size": 64,
653
+ "bits": 8
654
+ },
655
+ "language_model.model.layers.52.mlp.up_proj": {
656
+ "group_size": 64,
657
+ "bits": 8
658
+ },
659
+ "language_model.model.layers.53.mlp.gate_proj": {
660
+ "group_size": 64,
661
+ "bits": 8
662
+ },
663
+ "language_model.model.layers.53.mlp.down_proj": {
664
+ "group_size": 64,
665
+ "bits": 8
666
+ },
667
+ "language_model.model.layers.53.mlp.up_proj": {
668
+ "group_size": 64,
669
+ "bits": 8
670
+ },
671
+ "language_model.model.layers.54.mlp.gate_proj": {
672
+ "group_size": 64,
673
+ "bits": 8
674
+ },
675
+ "language_model.model.layers.54.mlp.down_proj": {
676
+ "group_size": 64,
677
+ "bits": 8
678
+ },
679
+ "language_model.model.layers.54.mlp.up_proj": {
680
+ "group_size": 64,
681
+ "bits": 8
682
+ },
683
+ "language_model.model.layers.55.mlp.gate_proj": {
684
+ "group_size": 64,
685
+ "bits": 8
686
+ },
687
+ "language_model.model.layers.55.mlp.down_proj": {
688
+ "group_size": 64,
689
+ "bits": 8
690
+ },
691
+ "language_model.model.layers.55.mlp.up_proj": {
692
+ "group_size": 64,
693
+ "bits": 8
694
+ },
695
+ "language_model.model.layers.56.mlp.gate_proj": {
696
+ "group_size": 64,
697
+ "bits": 8
698
+ },
699
+ "language_model.model.layers.56.mlp.down_proj": {
700
+ "group_size": 64,
701
+ "bits": 8
702
+ },
703
+ "language_model.model.layers.56.mlp.up_proj": {
704
+ "group_size": 64,
705
+ "bits": 8
706
+ },
707
+ "language_model.model.layers.57.mlp.gate_proj": {
708
+ "group_size": 64,
709
+ "bits": 8
710
+ },
711
+ "language_model.model.layers.57.mlp.down_proj": {
712
+ "group_size": 64,
713
+ "bits": 8
714
+ },
715
+ "language_model.model.layers.57.mlp.up_proj": {
716
+ "group_size": 64,
717
+ "bits": 8
718
+ },
719
+ "language_model.model.layers.58.mlp.gate_proj": {
720
+ "group_size": 64,
721
+ "bits": 8
722
+ },
723
+ "language_model.model.layers.58.mlp.down_proj": {
724
+ "group_size": 64,
725
+ "bits": 8
726
+ },
727
+ "language_model.model.layers.58.mlp.up_proj": {
728
+ "group_size": 64,
729
+ "bits": 8
730
+ },
731
+ "language_model.model.layers.59.mlp.gate_proj": {
732
+ "group_size": 64,
733
+ "bits": 8
734
+ },
735
+ "language_model.model.layers.59.mlp.down_proj": {
736
+ "group_size": 64,
737
+ "bits": 8
738
+ },
739
+ "language_model.model.layers.59.mlp.up_proj": {
740
+ "group_size": 64,
741
+ "bits": 8
742
+ }
743
+ },
744
+ "quantization_config": {
745
+ "group_size": 16,
746
+ "bits": 4,
747
+ "mode": "nvfp4",
748
+ "language_model.model.layers.0.mlp.gate_proj": {
749
+ "group_size": 64,
750
+ "bits": 8
751
+ },
752
+ "language_model.model.layers.0.mlp.down_proj": {
753
+ "group_size": 64,
754
+ "bits": 8
755
+ },
756
+ "language_model.model.layers.0.mlp.up_proj": {
757
+ "group_size": 64,
758
+ "bits": 8
759
+ },
760
+ "language_model.model.layers.1.mlp.gate_proj": {
761
+ "group_size": 64,
762
+ "bits": 8
763
+ },
764
+ "language_model.model.layers.1.mlp.down_proj": {
765
+ "group_size": 64,
766
+ "bits": 8
767
+ },
768
+ "language_model.model.layers.1.mlp.up_proj": {
769
+ "group_size": 64,
770
+ "bits": 8
771
+ },
772
+ "language_model.model.layers.2.mlp.gate_proj": {
773
+ "group_size": 64,
774
+ "bits": 8
775
+ },
776
+ "language_model.model.layers.2.mlp.down_proj": {
777
+ "group_size": 64,
778
+ "bits": 8
779
+ },
780
+ "language_model.model.layers.2.mlp.up_proj": {
781
+ "group_size": 64,
782
+ "bits": 8
783
+ },
784
+ "language_model.model.layers.3.mlp.gate_proj": {
785
+ "group_size": 64,
786
+ "bits": 8
787
+ },
788
+ "language_model.model.layers.3.mlp.down_proj": {
789
+ "group_size": 64,
790
+ "bits": 8
791
+ },
792
+ "language_model.model.layers.3.mlp.up_proj": {
793
+ "group_size": 64,
794
+ "bits": 8
795
+ },
796
+ "language_model.model.layers.4.mlp.gate_proj": {
797
+ "group_size": 64,
798
+ "bits": 8
799
+ },
800
+ "language_model.model.layers.4.mlp.down_proj": {
801
+ "group_size": 64,
802
+ "bits": 8
803
+ },
804
+ "language_model.model.layers.4.mlp.up_proj": {
805
+ "group_size": 64,
806
+ "bits": 8
807
+ },
808
+ "language_model.model.layers.5.mlp.gate_proj": {
809
+ "group_size": 64,
810
+ "bits": 8
811
+ },
812
+ "language_model.model.layers.5.mlp.down_proj": {
813
+ "group_size": 64,
814
+ "bits": 8
815
+ },
816
+ "language_model.model.layers.5.mlp.up_proj": {
817
+ "group_size": 64,
818
+ "bits": 8
819
+ },
820
+ "language_model.model.layers.6.mlp.gate_proj": {
821
+ "group_size": 64,
822
+ "bits": 8
823
+ },
824
+ "language_model.model.layers.6.mlp.down_proj": {
825
+ "group_size": 64,
826
+ "bits": 8
827
+ },
828
+ "language_model.model.layers.6.mlp.up_proj": {
829
+ "group_size": 64,
830
+ "bits": 8
831
+ },
832
+ "language_model.model.layers.7.mlp.gate_proj": {
833
+ "group_size": 64,
834
+ "bits": 8
835
+ },
836
+ "language_model.model.layers.7.mlp.down_proj": {
837
+ "group_size": 64,
838
+ "bits": 8
839
+ },
840
+ "language_model.model.layers.7.mlp.up_proj": {
841
+ "group_size": 64,
842
+ "bits": 8
843
+ },
844
+ "language_model.model.layers.8.mlp.gate_proj": {
845
+ "group_size": 64,
846
+ "bits": 8
847
+ },
848
+ "language_model.model.layers.8.mlp.down_proj": {
849
+ "group_size": 64,
850
+ "bits": 8
851
+ },
852
+ "language_model.model.layers.8.mlp.up_proj": {
853
+ "group_size": 64,
854
+ "bits": 8
855
+ },
856
+ "language_model.model.layers.9.mlp.gate_proj": {
857
+ "group_size": 64,
858
+ "bits": 8
859
+ },
860
+ "language_model.model.layers.9.mlp.down_proj": {
861
+ "group_size": 64,
862
+ "bits": 8
863
+ },
864
+ "language_model.model.layers.9.mlp.up_proj": {
865
+ "group_size": 64,
866
+ "bits": 8
867
+ },
868
+ "language_model.model.layers.10.mlp.gate_proj": {
869
+ "group_size": 64,
870
+ "bits": 8
871
+ },
872
+ "language_model.model.layers.10.mlp.down_proj": {
873
+ "group_size": 64,
874
+ "bits": 8
875
+ },
876
+ "language_model.model.layers.10.mlp.up_proj": {
877
+ "group_size": 64,
878
+ "bits": 8
879
+ },
880
+ "language_model.model.layers.11.mlp.gate_proj": {
881
+ "group_size": 64,
882
+ "bits": 8
883
+ },
884
+ "language_model.model.layers.11.mlp.down_proj": {
885
+ "group_size": 64,
886
+ "bits": 8
887
+ },
888
+ "language_model.model.layers.11.mlp.up_proj": {
889
+ "group_size": 64,
890
+ "bits": 8
891
+ },
892
+ "language_model.model.layers.12.mlp.gate_proj": {
893
+ "group_size": 64,
894
+ "bits": 8
895
+ },
896
+ "language_model.model.layers.12.mlp.down_proj": {
897
+ "group_size": 64,
898
+ "bits": 8
899
+ },
900
+ "language_model.model.layers.12.mlp.up_proj": {
901
+ "group_size": 64,
902
+ "bits": 8
903
+ },
904
+ "language_model.model.layers.13.mlp.gate_proj": {
905
+ "group_size": 64,
906
+ "bits": 8
907
+ },
908
+ "language_model.model.layers.13.mlp.down_proj": {
909
+ "group_size": 64,
910
+ "bits": 8
911
+ },
912
+ "language_model.model.layers.13.mlp.up_proj": {
913
+ "group_size": 64,
914
+ "bits": 8
915
+ },
916
+ "language_model.model.layers.14.mlp.gate_proj": {
917
+ "group_size": 64,
918
+ "bits": 8
919
+ },
920
+ "language_model.model.layers.14.mlp.down_proj": {
921
+ "group_size": 64,
922
+ "bits": 8
923
+ },
924
+ "language_model.model.layers.14.mlp.up_proj": {
925
+ "group_size": 64,
926
+ "bits": 8
927
+ },
928
+ "language_model.model.layers.15.mlp.gate_proj": {
929
+ "group_size": 64,
930
+ "bits": 8
931
+ },
932
+ "language_model.model.layers.15.mlp.down_proj": {
933
+ "group_size": 64,
934
+ "bits": 8
935
+ },
936
+ "language_model.model.layers.15.mlp.up_proj": {
937
+ "group_size": 64,
938
+ "bits": 8
939
+ },
940
+ "language_model.model.layers.16.mlp.gate_proj": {
941
+ "group_size": 64,
942
+ "bits": 8
943
+ },
944
+ "language_model.model.layers.16.mlp.down_proj": {
945
+ "group_size": 64,
946
+ "bits": 8
947
+ },
948
+ "language_model.model.layers.16.mlp.up_proj": {
949
+ "group_size": 64,
950
+ "bits": 8
951
+ },
952
+ "language_model.model.layers.17.mlp.gate_proj": {
953
+ "group_size": 64,
954
+ "bits": 8
955
+ },
956
+ "language_model.model.layers.17.mlp.down_proj": {
957
+ "group_size": 64,
958
+ "bits": 8
959
+ },
960
+ "language_model.model.layers.17.mlp.up_proj": {
961
+ "group_size": 64,
962
+ "bits": 8
963
+ },
964
+ "language_model.model.layers.18.mlp.gate_proj": {
965
+ "group_size": 64,
966
+ "bits": 8
967
+ },
968
+ "language_model.model.layers.18.mlp.down_proj": {
969
+ "group_size": 64,
970
+ "bits": 8
971
+ },
972
+ "language_model.model.layers.18.mlp.up_proj": {
973
+ "group_size": 64,
974
+ "bits": 8
975
+ },
976
+ "language_model.model.layers.19.mlp.gate_proj": {
977
+ "group_size": 64,
978
+ "bits": 8
979
+ },
980
+ "language_model.model.layers.19.mlp.down_proj": {
981
+ "group_size": 64,
982
+ "bits": 8
983
+ },
984
+ "language_model.model.layers.19.mlp.up_proj": {
985
+ "group_size": 64,
986
+ "bits": 8
987
+ },
988
+ "language_model.model.layers.20.mlp.gate_proj": {
989
+ "group_size": 64,
990
+ "bits": 8
991
+ },
992
+ "language_model.model.layers.20.mlp.down_proj": {
993
+ "group_size": 64,
994
+ "bits": 8
995
+ },
996
+ "language_model.model.layers.20.mlp.up_proj": {
997
+ "group_size": 64,
998
+ "bits": 8
999
+ },
1000
+ "language_model.model.layers.21.mlp.gate_proj": {
1001
+ "group_size": 64,
1002
+ "bits": 8
1003
+ },
1004
+ "language_model.model.layers.21.mlp.down_proj": {
1005
+ "group_size": 64,
1006
+ "bits": 8
1007
+ },
1008
+ "language_model.model.layers.21.mlp.up_proj": {
1009
+ "group_size": 64,
1010
+ "bits": 8
1011
+ },
1012
+ "language_model.model.layers.22.mlp.gate_proj": {
1013
+ "group_size": 64,
1014
+ "bits": 8
1015
+ },
1016
+ "language_model.model.layers.22.mlp.down_proj": {
1017
+ "group_size": 64,
1018
+ "bits": 8
1019
+ },
1020
+ "language_model.model.layers.22.mlp.up_proj": {
1021
+ "group_size": 64,
1022
+ "bits": 8
1023
+ },
1024
+ "language_model.model.layers.23.mlp.gate_proj": {
1025
+ "group_size": 64,
1026
+ "bits": 8
1027
+ },
1028
+ "language_model.model.layers.23.mlp.down_proj": {
1029
+ "group_size": 64,
1030
+ "bits": 8
1031
+ },
1032
+ "language_model.model.layers.23.mlp.up_proj": {
1033
+ "group_size": 64,
1034
+ "bits": 8
1035
+ },
1036
+ "language_model.model.layers.24.mlp.gate_proj": {
1037
+ "group_size": 64,
1038
+ "bits": 8
1039
+ },
1040
+ "language_model.model.layers.24.mlp.down_proj": {
1041
+ "group_size": 64,
1042
+ "bits": 8
1043
+ },
1044
+ "language_model.model.layers.24.mlp.up_proj": {
1045
+ "group_size": 64,
1046
+ "bits": 8
1047
+ },
1048
+ "language_model.model.layers.25.mlp.gate_proj": {
1049
+ "group_size": 64,
1050
+ "bits": 8
1051
+ },
1052
+ "language_model.model.layers.25.mlp.down_proj": {
1053
+ "group_size": 64,
1054
+ "bits": 8
1055
+ },
1056
+ "language_model.model.layers.25.mlp.up_proj": {
1057
+ "group_size": 64,
1058
+ "bits": 8
1059
+ },
1060
+ "language_model.model.layers.26.mlp.gate_proj": {
1061
+ "group_size": 64,
1062
+ "bits": 8
1063
+ },
1064
+ "language_model.model.layers.26.mlp.down_proj": {
1065
+ "group_size": 64,
1066
+ "bits": 8
1067
+ },
1068
+ "language_model.model.layers.26.mlp.up_proj": {
1069
+ "group_size": 64,
1070
+ "bits": 8
1071
+ },
1072
+ "language_model.model.layers.27.mlp.gate_proj": {
1073
+ "group_size": 64,
1074
+ "bits": 8
1075
+ },
1076
+ "language_model.model.layers.27.mlp.down_proj": {
1077
+ "group_size": 64,
1078
+ "bits": 8
1079
+ },
1080
+ "language_model.model.layers.27.mlp.up_proj": {
1081
+ "group_size": 64,
1082
+ "bits": 8
1083
+ },
1084
+ "language_model.model.layers.28.mlp.gate_proj": {
1085
+ "group_size": 64,
1086
+ "bits": 8
1087
+ },
1088
+ "language_model.model.layers.28.mlp.down_proj": {
1089
+ "group_size": 64,
1090
+ "bits": 8
1091
+ },
1092
+ "language_model.model.layers.28.mlp.up_proj": {
1093
+ "group_size": 64,
1094
+ "bits": 8
1095
+ },
1096
+ "language_model.model.layers.29.mlp.gate_proj": {
1097
+ "group_size": 64,
1098
+ "bits": 8
1099
+ },
1100
+ "language_model.model.layers.29.mlp.down_proj": {
1101
+ "group_size": 64,
1102
+ "bits": 8
1103
+ },
1104
+ "language_model.model.layers.29.mlp.up_proj": {
1105
+ "group_size": 64,
1106
+ "bits": 8
1107
+ },
1108
+ "language_model.model.layers.30.mlp.gate_proj": {
1109
+ "group_size": 64,
1110
+ "bits": 8
1111
+ },
1112
+ "language_model.model.layers.30.mlp.down_proj": {
1113
+ "group_size": 64,
1114
+ "bits": 8
1115
+ },
1116
+ "language_model.model.layers.30.mlp.up_proj": {
1117
+ "group_size": 64,
1118
+ "bits": 8
1119
+ },
1120
+ "language_model.model.layers.31.mlp.gate_proj": {
1121
+ "group_size": 64,
1122
+ "bits": 8
1123
+ },
1124
+ "language_model.model.layers.31.mlp.down_proj": {
1125
+ "group_size": 64,
1126
+ "bits": 8
1127
+ },
1128
+ "language_model.model.layers.31.mlp.up_proj": {
1129
+ "group_size": 64,
1130
+ "bits": 8
1131
+ },
1132
+ "language_model.model.layers.32.mlp.gate_proj": {
1133
+ "group_size": 64,
1134
+ "bits": 8
1135
+ },
1136
+ "language_model.model.layers.32.mlp.down_proj": {
1137
+ "group_size": 64,
1138
+ "bits": 8
1139
+ },
1140
+ "language_model.model.layers.32.mlp.up_proj": {
1141
+ "group_size": 64,
1142
+ "bits": 8
1143
+ },
1144
+ "language_model.model.layers.33.mlp.gate_proj": {
1145
+ "group_size": 64,
1146
+ "bits": 8
1147
+ },
1148
+ "language_model.model.layers.33.mlp.down_proj": {
1149
+ "group_size": 64,
1150
+ "bits": 8
1151
+ },
1152
+ "language_model.model.layers.33.mlp.up_proj": {
1153
+ "group_size": 64,
1154
+ "bits": 8
1155
+ },
1156
+ "language_model.model.layers.34.mlp.gate_proj": {
1157
+ "group_size": 64,
1158
+ "bits": 8
1159
+ },
1160
+ "language_model.model.layers.34.mlp.down_proj": {
1161
+ "group_size": 64,
1162
+ "bits": 8
1163
+ },
1164
+ "language_model.model.layers.34.mlp.up_proj": {
1165
+ "group_size": 64,
1166
+ "bits": 8
1167
+ },
1168
+ "language_model.model.layers.35.mlp.gate_proj": {
1169
+ "group_size": 64,
1170
+ "bits": 8
1171
+ },
1172
+ "language_model.model.layers.35.mlp.down_proj": {
1173
+ "group_size": 64,
1174
+ "bits": 8
1175
+ },
1176
+ "language_model.model.layers.35.mlp.up_proj": {
1177
+ "group_size": 64,
1178
+ "bits": 8
1179
+ },
1180
+ "language_model.model.layers.36.mlp.gate_proj": {
1181
+ "group_size": 64,
1182
+ "bits": 8
1183
+ },
1184
+ "language_model.model.layers.36.mlp.down_proj": {
1185
+ "group_size": 64,
1186
+ "bits": 8
1187
+ },
1188
+ "language_model.model.layers.36.mlp.up_proj": {
1189
+ "group_size": 64,
1190
+ "bits": 8
1191
+ },
1192
+ "language_model.model.layers.37.mlp.gate_proj": {
1193
+ "group_size": 64,
1194
+ "bits": 8
1195
+ },
1196
+ "language_model.model.layers.37.mlp.down_proj": {
1197
+ "group_size": 64,
1198
+ "bits": 8
1199
+ },
1200
+ "language_model.model.layers.37.mlp.up_proj": {
1201
+ "group_size": 64,
1202
+ "bits": 8
1203
+ },
1204
+ "language_model.model.layers.38.mlp.gate_proj": {
1205
+ "group_size": 64,
1206
+ "bits": 8
1207
+ },
1208
+ "language_model.model.layers.38.mlp.down_proj": {
1209
+ "group_size": 64,
1210
+ "bits": 8
1211
+ },
1212
+ "language_model.model.layers.38.mlp.up_proj": {
1213
+ "group_size": 64,
1214
+ "bits": 8
1215
+ },
1216
+ "language_model.model.layers.39.mlp.gate_proj": {
1217
+ "group_size": 64,
1218
+ "bits": 8
1219
+ },
1220
+ "language_model.model.layers.39.mlp.down_proj": {
1221
+ "group_size": 64,
1222
+ "bits": 8
1223
+ },
1224
+ "language_model.model.layers.39.mlp.up_proj": {
1225
+ "group_size": 64,
1226
+ "bits": 8
1227
+ },
1228
+ "language_model.model.layers.40.mlp.gate_proj": {
1229
+ "group_size": 64,
1230
+ "bits": 8
1231
+ },
1232
+ "language_model.model.layers.40.mlp.down_proj": {
1233
+ "group_size": 64,
1234
+ "bits": 8
1235
+ },
1236
+ "language_model.model.layers.40.mlp.up_proj": {
1237
+ "group_size": 64,
1238
+ "bits": 8
1239
+ },
1240
+ "language_model.model.layers.41.mlp.gate_proj": {
1241
+ "group_size": 64,
1242
+ "bits": 8
1243
+ },
1244
+ "language_model.model.layers.41.mlp.down_proj": {
1245
+ "group_size": 64,
1246
+ "bits": 8
1247
+ },
1248
+ "language_model.model.layers.41.mlp.up_proj": {
1249
+ "group_size": 64,
1250
+ "bits": 8
1251
+ },
1252
+ "language_model.model.layers.42.mlp.gate_proj": {
1253
+ "group_size": 64,
1254
+ "bits": 8
1255
+ },
1256
+ "language_model.model.layers.42.mlp.down_proj": {
1257
+ "group_size": 64,
1258
+ "bits": 8
1259
+ },
1260
+ "language_model.model.layers.42.mlp.up_proj": {
1261
+ "group_size": 64,
1262
+ "bits": 8
1263
+ },
1264
+ "language_model.model.layers.43.mlp.gate_proj": {
1265
+ "group_size": 64,
1266
+ "bits": 8
1267
+ },
1268
+ "language_model.model.layers.43.mlp.down_proj": {
1269
+ "group_size": 64,
1270
+ "bits": 8
1271
+ },
1272
+ "language_model.model.layers.43.mlp.up_proj": {
1273
+ "group_size": 64,
1274
+ "bits": 8
1275
+ },
1276
+ "language_model.model.layers.44.mlp.gate_proj": {
1277
+ "group_size": 64,
1278
+ "bits": 8
1279
+ },
1280
+ "language_model.model.layers.44.mlp.down_proj": {
1281
+ "group_size": 64,
1282
+ "bits": 8
1283
+ },
1284
+ "language_model.model.layers.44.mlp.up_proj": {
1285
+ "group_size": 64,
1286
+ "bits": 8
1287
+ },
1288
+ "language_model.model.layers.45.mlp.gate_proj": {
1289
+ "group_size": 64,
1290
+ "bits": 8
1291
+ },
1292
+ "language_model.model.layers.45.mlp.down_proj": {
1293
+ "group_size": 64,
1294
+ "bits": 8
1295
+ },
1296
+ "language_model.model.layers.45.mlp.up_proj": {
1297
+ "group_size": 64,
1298
+ "bits": 8
1299
+ },
1300
+ "language_model.model.layers.46.mlp.gate_proj": {
1301
+ "group_size": 64,
1302
+ "bits": 8
1303
+ },
1304
+ "language_model.model.layers.46.mlp.down_proj": {
1305
+ "group_size": 64,
1306
+ "bits": 8
1307
+ },
1308
+ "language_model.model.layers.46.mlp.up_proj": {
1309
+ "group_size": 64,
1310
+ "bits": 8
1311
+ },
1312
+ "language_model.model.layers.47.mlp.gate_proj": {
1313
+ "group_size": 64,
1314
+ "bits": 8
1315
+ },
1316
+ "language_model.model.layers.47.mlp.down_proj": {
1317
+ "group_size": 64,
1318
+ "bits": 8
1319
+ },
1320
+ "language_model.model.layers.47.mlp.up_proj": {
1321
+ "group_size": 64,
1322
+ "bits": 8
1323
+ },
1324
+ "language_model.model.layers.48.mlp.gate_proj": {
1325
+ "group_size": 64,
1326
+ "bits": 8
1327
+ },
1328
+ "language_model.model.layers.48.mlp.down_proj": {
1329
+ "group_size": 64,
1330
+ "bits": 8
1331
+ },
1332
+ "language_model.model.layers.48.mlp.up_proj": {
1333
+ "group_size": 64,
1334
+ "bits": 8
1335
+ },
1336
+ "language_model.model.layers.49.mlp.gate_proj": {
1337
+ "group_size": 64,
1338
+ "bits": 8
1339
+ },
1340
+ "language_model.model.layers.49.mlp.down_proj": {
1341
+ "group_size": 64,
1342
+ "bits": 8
1343
+ },
1344
+ "language_model.model.layers.49.mlp.up_proj": {
1345
+ "group_size": 64,
1346
+ "bits": 8
1347
+ },
1348
+ "language_model.model.layers.50.mlp.gate_proj": {
1349
+ "group_size": 64,
1350
+ "bits": 8
1351
+ },
1352
+ "language_model.model.layers.50.mlp.down_proj": {
1353
+ "group_size": 64,
1354
+ "bits": 8
1355
+ },
1356
+ "language_model.model.layers.50.mlp.up_proj": {
1357
+ "group_size": 64,
1358
+ "bits": 8
1359
+ },
1360
+ "language_model.model.layers.51.mlp.gate_proj": {
1361
+ "group_size": 64,
1362
+ "bits": 8
1363
+ },
1364
+ "language_model.model.layers.51.mlp.down_proj": {
1365
+ "group_size": 64,
1366
+ "bits": 8
1367
+ },
1368
+ "language_model.model.layers.51.mlp.up_proj": {
1369
+ "group_size": 64,
1370
+ "bits": 8
1371
+ },
1372
+ "language_model.model.layers.52.mlp.gate_proj": {
1373
+ "group_size": 64,
1374
+ "bits": 8
1375
+ },
1376
+ "language_model.model.layers.52.mlp.down_proj": {
1377
+ "group_size": 64,
1378
+ "bits": 8
1379
+ },
1380
+ "language_model.model.layers.52.mlp.up_proj": {
1381
+ "group_size": 64,
1382
+ "bits": 8
1383
+ },
1384
+ "language_model.model.layers.53.mlp.gate_proj": {
1385
+ "group_size": 64,
1386
+ "bits": 8
1387
+ },
1388
+ "language_model.model.layers.53.mlp.down_proj": {
1389
+ "group_size": 64,
1390
+ "bits": 8
1391
+ },
1392
+ "language_model.model.layers.53.mlp.up_proj": {
1393
+ "group_size": 64,
1394
+ "bits": 8
1395
+ },
1396
+ "language_model.model.layers.54.mlp.gate_proj": {
1397
+ "group_size": 64,
1398
+ "bits": 8
1399
+ },
1400
+ "language_model.model.layers.54.mlp.down_proj": {
1401
+ "group_size": 64,
1402
+ "bits": 8
1403
+ },
1404
+ "language_model.model.layers.54.mlp.up_proj": {
1405
+ "group_size": 64,
1406
+ "bits": 8
1407
+ },
1408
+ "language_model.model.layers.55.mlp.gate_proj": {
1409
+ "group_size": 64,
1410
+ "bits": 8
1411
+ },
1412
+ "language_model.model.layers.55.mlp.down_proj": {
1413
+ "group_size": 64,
1414
+ "bits": 8
1415
+ },
1416
+ "language_model.model.layers.55.mlp.up_proj": {
1417
+ "group_size": 64,
1418
+ "bits": 8
1419
+ },
1420
+ "language_model.model.layers.56.mlp.gate_proj": {
1421
+ "group_size": 64,
1422
+ "bits": 8
1423
+ },
1424
+ "language_model.model.layers.56.mlp.down_proj": {
1425
+ "group_size": 64,
1426
+ "bits": 8
1427
+ },
1428
+ "language_model.model.layers.56.mlp.up_proj": {
1429
+ "group_size": 64,
1430
+ "bits": 8
1431
+ },
1432
+ "language_model.model.layers.57.mlp.gate_proj": {
1433
+ "group_size": 64,
1434
+ "bits": 8
1435
+ },
1436
+ "language_model.model.layers.57.mlp.down_proj": {
1437
+ "group_size": 64,
1438
+ "bits": 8
1439
+ },
1440
+ "language_model.model.layers.57.mlp.up_proj": {
1441
+ "group_size": 64,
1442
+ "bits": 8
1443
+ },
1444
+ "language_model.model.layers.58.mlp.gate_proj": {
1445
+ "group_size": 64,
1446
+ "bits": 8
1447
+ },
1448
+ "language_model.model.layers.58.mlp.down_proj": {
1449
+ "group_size": 64,
1450
+ "bits": 8
1451
+ },
1452
+ "language_model.model.layers.58.mlp.up_proj": {
1453
+ "group_size": 64,
1454
+ "bits": 8
1455
+ },
1456
+ "language_model.model.layers.59.mlp.gate_proj": {
1457
+ "group_size": 64,
1458
+ "bits": 8
1459
+ },
1460
+ "language_model.model.layers.59.mlp.down_proj": {
1461
+ "group_size": 64,
1462
+ "bits": 8
1463
+ },
1464
+ "language_model.model.layers.59.mlp.up_proj": {
1465
+ "group_size": 64,
1466
+ "bits": 8
1467
+ }
1468
+ },
1469
+ "text_config": {
1470
+ "attention_bias": false,
1471
+ "attention_dropout": 0.0,
1472
+ "attention_k_eq_v": true,
1473
+ "bos_token_id": 2,
1474
+ "torch_dtype": "bfloat16",
1475
+ "enable_moe_block": false,
1476
+ "eos_token_id": 1,
1477
+ "expert_intermediate_size": null,
1478
+ "final_logit_softcapping": 30.0,
1479
+ "global_head_dim": 512,
1480
+ "head_dim": 256,
1481
+ "hidden_activation": "gelu_pytorch_tanh",
1482
+ "hidden_size": 5376,
1483
+ "hidden_size_per_layer_input": 0,
1484
+ "initializer_range": 0.02,
1485
+ "intermediate_size": 21504,
1486
+ "layer_types": [
1487
+ "sliding_attention",
1488
+ "sliding_attention",
1489
+ "sliding_attention",
1490
+ "sliding_attention",
1491
+ "sliding_attention",
1492
+ "full_attention",
1493
+ "sliding_attention",
1494
+ "sliding_attention",
1495
+ "sliding_attention",
1496
+ "sliding_attention",
1497
+ "sliding_attention",
1498
+ "full_attention",
1499
+ "sliding_attention",
1500
+ "sliding_attention",
1501
+ "sliding_attention",
1502
+ "sliding_attention",
1503
+ "sliding_attention",
1504
+ "full_attention",
1505
+ "sliding_attention",
1506
+ "sliding_attention",
1507
+ "sliding_attention",
1508
+ "sliding_attention",
1509
+ "sliding_attention",
1510
+ "full_attention",
1511
+ "sliding_attention",
1512
+ "sliding_attention",
1513
+ "sliding_attention",
1514
+ "sliding_attention",
1515
+ "sliding_attention",
1516
+ "full_attention",
1517
+ "sliding_attention",
1518
+ "sliding_attention",
1519
+ "sliding_attention",
1520
+ "sliding_attention",
1521
+ "sliding_attention",
1522
+ "full_attention",
1523
+ "sliding_attention",
1524
+ "sliding_attention",
1525
+ "sliding_attention",
1526
+ "sliding_attention",
1527
+ "sliding_attention",
1528
+ "full_attention",
1529
+ "sliding_attention",
1530
+ "sliding_attention",
1531
+ "sliding_attention",
1532
+ "sliding_attention",
1533
+ "sliding_attention",
1534
+ "full_attention",
1535
+ "sliding_attention",
1536
+ "sliding_attention",
1537
+ "sliding_attention",
1538
+ "sliding_attention",
1539
+ "sliding_attention",
1540
+ "full_attention",
1541
+ "sliding_attention",
1542
+ "sliding_attention",
1543
+ "sliding_attention",
1544
+ "sliding_attention",
1545
+ "sliding_attention",
1546
+ "full_attention"
1547
+ ],
1548
+ "max_position_embeddings": 262144,
1549
+ "model_type": "gemma4_text",
1550
+ "moe_intermediate_size": null,
1551
+ "num_attention_heads": 32,
1552
+ "num_experts": null,
1553
+ "num_global_key_value_heads": 4,
1554
+ "num_hidden_layers": 60,
1555
+ "num_key_value_heads": 16,
1556
+ "num_kv_shared_layers": 0,
1557
+ "pad_token_id": 0,
1558
+ "rms_norm_eps": 1e-06,
1559
+ "rope_parameters": {
1560
+ "full_attention": {
1561
+ "partial_rotary_factor": 0.25,
1562
+ "rope_theta": 1000000.0,
1563
+ "rope_type": "proportional"
1564
+ },
1565
+ "sliding_attention": {
1566
+ "rope_theta": 10000.0,
1567
+ "rope_type": "default"
1568
+ }
1569
+ },
1570
+ "sliding_window": 1024,
1571
+ "tie_word_embeddings": true,
1572
+ "top_k_experts": null,
1573
+ "use_bidirectional_attention": "vision",
1574
+ "use_cache": true,
1575
+ "use_double_wide_mlp": false,
1576
+ "vocab_size": 262144,
1577
+ "vocab_size_per_layer_input": 262144
1578
+ },
1579
+ "tie_word_embeddings": true,
1580
+ "unsloth_fixed": true,
1581
+ "unsloth_version": "2026.4.4",
1582
+ "use_cache": false,
1583
+ "video_token_id": 258884,
1584
+ "vision_config": {
1585
+ "_name_or_path": "",
1586
+ "architectures": null,
1587
+ "attention_bias": false,
1588
+ "attention_dropout": 0.0,
1589
+ "chunk_size_feed_forward": 0,
1590
+ "default_output_length": 280,
1591
+ "torch_dtype": "bfloat16",
1592
+ "global_head_dim": 72,
1593
+ "head_dim": 72,
1594
+ "hidden_activation": "gelu_pytorch_tanh",
1595
+ "hidden_size": 1152,
1596
+ "id2label": {
1597
+ "0": "LABEL_0",
1598
+ "1": "LABEL_1"
1599
+ },
1600
+ "initializer_range": 0.02,
1601
+ "intermediate_size": 4304,
1602
+ "is_encoder_decoder": false,
1603
+ "label2id": {
1604
+ "LABEL_0": 0,
1605
+ "LABEL_1": 1
1606
+ },
1607
+ "max_position_embeddings": 131072,
1608
+ "model_type": "gemma4_vision",
1609
+ "num_attention_heads": 16,
1610
+ "num_hidden_layers": 27,
1611
+ "num_key_value_heads": 16,
1612
+ "output_attentions": false,
1613
+ "output_hidden_states": false,
1614
+ "patch_size": 16,
1615
+ "pooling_kernel_size": 3,
1616
+ "position_embedding_size": 10240,
1617
+ "problem_type": null,
1618
+ "return_dict": true,
1619
+ "rms_norm_eps": 1e-06,
1620
+ "rope_parameters": {
1621
+ "rope_theta": 100.0,
1622
+ "rope_type": "default"
1623
+ },
1624
+ "standardize": true,
1625
+ "use_clipped_linears": false
1626
+ },
1627
+ "vision_soft_tokens_per_image": 280
1628
+ }
model-00001-of-00006.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f95474740be5f15a60ffe878d1bb7da2fac1c249151371561d2ec6763e22c265
3
- size 1772077056
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7f80b26dc64e89529931ced458ed73c3fee307c0159940e7abbfdb51cb0fc3a1
3
+ size 5326585088
model-00002-of-00006.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4616e1c7c831f8989a3eaf8a263b944e69c1d7b6facd3dd70030de16a72ba0e1
3
+ size 5351448920
model-00003-of-00006.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c3de20ea2cca81d89a3aea1dbe171562f3360e250880cde693a5c0fdfa749627
3
+ size 5363836511
model-00004-of-00006.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:85274d4d47ea68fda24d5927244f7ac43ae54225ea12756c4a4bbd75b682030f
3
+ size 5363836509
model-00005-of-00006.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a4dde640b866e6da71e419d97bb0ba99a2b5fdd06a9791521a2a2b8e339b9406
3
+ size 5351450167
model-00006-of-00006.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d0e11f1a62dac394ed33970bb8a3753eff497a6d60c4631c3c815ee86c57cfa
3
+ size 2059324259
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
processor_config.json ADDED
@@ -0,0 +1,42 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "audio_seq_length": 750,
3
+ "image_processor": {
4
+ "do_convert_rgb": true,
5
+ "do_normalize": false,
6
+ "do_rescale": true,
7
+ "do_resize": true,
8
+ "image_mean": [
9
+ 0.0,
10
+ 0.0,
11
+ 0.0
12
+ ],
13
+ "image_processor_type": "Gemma4ImageProcessor",
14
+ "image_seq_length": 280,
15
+ "image_std": [
16
+ 1.0,
17
+ 1.0,
18
+ 1.0
19
+ ],
20
+ "max_soft_tokens": 280,
21
+ "patch_size": 16,
22
+ "pooling_kernel_size": 3,
23
+ "resample": 3,
24
+ "rescale_factor": 0.00392156862745098,
25
+ "size": {
26
+ "height": 224,
27
+ "width": 224
28
+ }
29
+ },
30
+ "image_seq_length": 280,
31
+ "processor_class": "Gemma4Processor",
32
+ "feature_extractor": {
33
+ "feature_extractor_type": "Gemma4AudioFeatureExtractor",
34
+ "sampling_rate": 16000,
35
+ "num_mel_filters": 128,
36
+ "fft_length": 512,
37
+ "hop_length": 160,
38
+ "chunk_duration": 8.0,
39
+ "overlap_duration": 1.0
40
+ },
41
+ "audio_ms_per_token": 40
42
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc8d3a0ce36466ccc1278bf987df5f71db1719b9ca6b4118264f45cb627bfe0f
3
+ size 32169626
tokenizer_config.json ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "audio_token": "<|audio|>",
3
+ "backend": "tokenizers",
4
+ "boa_token": "<|audio>",
5
+ "boi_token": "<|image>",
6
+ "bos_token": "<bos>",
7
+ "eoa_token": "<audio|>",
8
+ "eoc_token": "<channel|>",
9
+ "eoi_token": "<image|>",
10
+ "eos_token": "<turn|>",
11
+ "eot_token": "<turn|>",
12
+ "escape_token": "<|\"|>",
13
+ "etc_token": "<tool_call|>",
14
+ "etd_token": "<tool|>",
15
+ "etr_token": "<tool_response|>",
16
+ "extra_special_tokens": [
17
+ "<|video|>"
18
+ ],
19
+ "image_token": "<|image|>",
20
+ "is_local": true,
21
+ "mask_token": "<mask>",
22
+ "model_max_length": 262144,
23
+ "model_specific_special_tokens": {
24
+ "audio_token": "<|audio|>",
25
+ "boa_token": "<|audio>",
26
+ "boi_token": "<|image>",
27
+ "eoa_token": "<audio|>",
28
+ "eoc_token": "<channel|>",
29
+ "eoi_token": "<image|>",
30
+ "eot_token": "<turn|>",
31
+ "escape_token": "<|\"|>",
32
+ "etc_token": "<tool_call|>",
33
+ "etd_token": "<tool|>",
34
+ "etr_token": "<tool_response|>",
35
+ "image_token": "<|image|>",
36
+ "soc_token": "<|channel>",
37
+ "sot_token": "<|turn>",
38
+ "stc_token": "<|tool_call>",
39
+ "std_token": "<|tool>",
40
+ "str_token": "<|tool_response>",
41
+ "think_token": "<|think|>"
42
+ },
43
+ "pad_token": "<pad>",
44
+ "padding_side": "right",
45
+ "processor_class": "Gemma4Processor",
46
+ "response_schema": {
47
+ "properties": {
48
+ "content": {
49
+ "type": "string"
50
+ },
51
+ "role": {
52
+ "const": "assistant"
53
+ },
54
+ "thinking": {
55
+ "type": "string"
56
+ },
57
+ "tool_calls": {
58
+ "items": {
59
+ "properties": {
60
+ "function": {
61
+ "properties": {
62
+ "arguments": {
63
+ "additionalProperties": {},
64
+ "type": "object",
65
+ "x-parser": "gemma4-tool-call"
66
+ },
67
+ "name": {
68
+ "type": "string"
69
+ }
70
+ },
71
+ "type": "object",
72
+ "x-regex": "call\\:(?P<name>\\w+)(?P<arguments>\\{.*\\})"
73
+ },
74
+ "type": {
75
+ "const": "function"
76
+ }
77
+ },
78
+ "type": "object"
79
+ },
80
+ "type": "array",
81
+ "x-regex-iterator": "<\\|tool_call>(.*?)<tool_call\\|>"
82
+ }
83
+ },
84
+ "type": "object",
85
+ "x-regex": "(\\<\\|channel\\>thought\\n(?P<thinking>.*?)\\<channel\\|\\>)?(?P<tool_calls>\\<\\|tool_call\\>.*\\<tool_call\\|\\>)?(?P<content>(?:(?!\\<turn\\|\\>)(?!\\<\\|tool_response\\>).)+)?(?:\\<turn\\|\\>|\\<\\|tool_response\\>)?"
86
+ },
87
+ "soc_token": "<|channel>",
88
+ "sot_token": "<|turn>",
89
+ "stc_token": "<|tool_call>",
90
+ "std_token": "<|tool>",
91
+ "str_token": "<|tool_response>",
92
+ "think_token": "<|think|>",
93
+ "tokenizer_class": "GemmaTokenizer",
94
+ "unk_token": "<unk>"
95
+ }