Hugging Face's logo

Intel
/
gemma-4-26B-A4B-it-int4-mixed-AutoRound

Safetensors
gemma4
text-generation-inference
4-bit precision
auto-round
Model card Files
xet
 Community
1
wenhuach commited on
Commit
d2c00ca
·
verified ·
1 Parent(s): 028a867

Upload folder using huggingface_hub

Browse files
Files changed (19) hide show
  1. .gitattributes +1 -0
  2. chat_template.jinja +266 -0
  3. config.json +873 -0
  4. generation_config.json +14 -0
  5. model-00001-of-00009.safetensors +3 -0
  6. model-00002-of-00009.safetensors +3 -0
  7. model-00003-of-00009.safetensors +3 -0
  8. model-00004-of-00009.safetensors +3 -0
  9. model-00005-of-00009.safetensors +3 -0
  10. model-00006-of-00009.safetensors +3 -0
  11. model-00007-of-00009.safetensors +3 -0
  12. model-00008-of-00009.safetensors +3 -0
  13. model-00009-of-00009.safetensors +3 -0
  14. model.safetensors.index.json +0 -0
  15. preprocessor_config.json +23 -0
  16. processor_config.json +75 -0
  17. quantization_config.json +727 -0
  18. tokenizer.json +3 -0
  19. 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
chat_template.jinja ADDED
@@ -0,0 +1,266 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {%- macro format_parameters(properties, required) -%}
2
+ {%- set standard_keys = ['description', 'type', 'properties', 'required', 'nullable'] -%}
3
+ {%- set ns = namespace(found_first=false) -%}
4
+ {%- for key, value in properties | dictsort -%}
5
+ {%- set add_comma = false -%}
6
+ {%- if key not in standard_keys -%}
7
+ {%- if ns.found_first %},{% endif -%}
8
+ {%- set ns.found_first = true -%}
9
+ {{ key }}:{
10
+ {%- if value['description'] -%}
11
+ description:<|"|>{{ value['description'] }}<|"|>
12
+ {%- set add_comma = true -%}
13
+ {%- endif -%}
14
+ {%- if value['nullable'] %}
15
+ {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
16
+ nullable:true
17
+ {%- endif -%}
18
+ {%- if value['type'] | upper == 'STRING' -%}
19
+ {%- if value['enum'] -%}
20
+ {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
21
+ enum:{{ format_argument(value['enum']) }}
22
+ {%- endif -%}
23
+ {%- elif value['type'] | upper == 'OBJECT' -%}
24
+ ,properties:{
25
+ {%- if value['properties'] is defined and value['properties'] is mapping -%}
26
+ {{- format_parameters(value['properties'], value['required'] | default([])) -}}
27
+ {%- elif value is mapping -%}
28
+ {{- format_parameters(value, value['required'] | default([])) -}}
29
+ {%- endif -%}
30
+ }
31
+ {%- if value['required'] -%}
32
+ ,required:[
33
+ {%- for item in value['required'] | default([]) -%}
34
+ <|"|>{{- item -}}<|"|>
35
+ {%- if not loop.last %},{% endif -%}
36
+ {%- endfor -%}
37
+ ]
38
+ {%- endif -%}
39
+ {%- elif value['type'] | upper == 'ARRAY' -%}
40
+ {%- if value['items'] is mapping and value['items'] -%}
41
+ ,items:{
42
+ {%- set ns_items = namespace(found_first=false) -%}
43
+ {%- for item_key, item_value in value['items'] | dictsort -%}
44
+ {%- if item_value is not none -%}
45
+ {%- if ns_items.found_first %},{% endif -%}
46
+ {%- set ns_items.found_first = true -%}
47
+ {%- if item_key == 'properties' -%}
48
+ properties:{
49
+ {%- if item_value is mapping -%}
50
+ {{- format_parameters(item_value, value['items']['required'] | default([])) -}}
51
+ {%- endif -%}
52
+ }
53
+ {%- elif item_key == 'required' -%}
54
+ required:[
55
+ {%- for req_item in item_value -%}
56
+ <|"|>{{- req_item -}}<|"|>
57
+ {%- if not loop.last %},{% endif -%}
58
+ {%- endfor -%}
59
+ ]
60
+ {%- elif item_key == 'type' -%}
61
+ {%- if item_value is string -%}
62
+ type:{{ format_argument(item_value | upper) }}
63
+ {%- else -%}
64
+ type:{{ format_argument(item_value | map('upper') | list) }}
65
+ {%- endif -%}
66
+ {%- else -%}
67
+ {{ item_key }}:{{ format_argument(item_value) }}
68
+ {%- endif -%}
69
+ {%- endif -%}
70
+ {%- endfor -%}
71
+ }
72
+ {%- endif -%}
73
+ {%- endif -%}
74
+ {%- if add_comma %},{%- else -%} {%- set add_comma = true -%} {% endif -%}
75
+ type:<|"|>{{ value['type'] | upper }}<|"|>}
76
+ {%- endif -%}
77
+ {%- endfor -%}
78
+ {%- endmacro -%}
79
+ {%- macro format_function_declaration(tool_data) -%}
80
+ declaration:{{- tool_data['function']['name'] -}}{description:<|"|>{{- tool_data['function']['description'] -}}<|"|>
81
+ {%- set params = tool_data['function']['parameters'] -%}
82
+ {%- if params -%}
83
+ ,parameters:{
84
+ {%- if params['properties'] -%}
85
+ properties:{ {{- format_parameters(params['properties'], params['required']) -}} },
86
+ {%- endif -%}
87
+ {%- if params['required'] -%}
88
+ required:[
89
+ {%- for item in params['required'] -%}
90
+ <|"|>{{- item -}}<|"|>
91
+ {{- ',' if not loop.last -}}
92
+ {%- endfor -%}
93
+ ],
94
+ {%- endif -%}
95
+ {%- if params['type'] -%}
96
+ type:<|"|>{{- params['type'] | upper -}}<|"|>}
97
+ {%- endif -%}
98
+ {%- endif -%}
99
+ {%- if 'response' in tool_data['function'] -%}
100
+ {%- set response_declaration = tool_data['function']['response'] -%}
101
+ ,response:{
102
+ {%- if response_declaration['description'] -%}
103
+ description:<|"|>{{- response_declaration['description'] -}}<|"|>,
104
+ {%- endif -%}
105
+ {%- if response_declaration['type'] | upper == 'OBJECT' -%}
106
+ type:<|"|>{{- response_declaration['type'] | upper -}}<|"|>}
107
+ {%- endif -%}
108
+ {%- endif -%}
109
+ }
110
+ {%- endmacro -%}
111
+ {%- macro format_argument(argument, escape_keys=True) -%}
112
+ {%- if argument is string -%}
113
+ {{- '<|"|>' + argument + '<|"|>' -}}
114
+ {%- elif argument is boolean -%}
115
+ {{- 'true' if argument else 'false' -}}
116
+ {%- elif argument is mapping -%}
117
+ {{- '{' -}}
118
+ {%- set ns = namespace(found_first=false) -%}
119
+ {%- for key, value in argument | dictsort -%}
120
+ {%- if ns.found_first %},{% endif -%}
121
+ {%- set ns.found_first = true -%}
122
+ {%- if escape_keys -%}
123
+ {{- '<|"|>' + key + '<|"|>' -}}
124
+ {%- else -%}
125
+ {{- key -}}
126
+ {%- endif -%}
127
+ :{{- format_argument(value, escape_keys=escape_keys) -}}
128
+ {%- endfor -%}
129
+ {{- '}' -}}
130
+ {%- elif argument is sequence -%}
131
+ {{- '[' -}}
132
+ {%- for item in argument -%}
133
+ {{- format_argument(item, escape_keys=escape_keys) -}}
134
+ {%- if not loop.last %},{% endif -%}
135
+ {%- endfor -%}
136
+ {{- ']' -}}
137
+ {%- else -%}
138
+ {{- argument -}}
139
+ {%- endif -%}
140
+ {%- endmacro -%}
141
+ {%- macro strip_thinking(text) -%}
142
+ {%- set ns = namespace(result='') -%}
143
+ {%- for part in text.split('<channel|>') -%}
144
+ {%- if '<|channel>' in part -%}
145
+ {%- set ns.result = ns.result + part.split('<|channel>')[0] -%}
146
+ {%- else -%}
147
+ {%- set ns.result = ns.result + part -%}
148
+ {%- endif -%}
149
+ {%- endfor -%}
150
+ {{- ns.result | trim -}}
151
+ {%- endmacro -%}
152
+
153
+ {%- set ns = namespace(prev_message_type=None) -%}
154
+ {%- set loop_messages = messages -%}
155
+ {{ bos_token }}
156
+ {#- Handle System/Tool Definitions Block -#}
157
+ {%- if (enable_thinking is defined and enable_thinking) or tools or messages[0]['role'] in ['system', 'developer'] -%}
158
+ {{- '<|turn>system\n' -}}
159
+
160
+ {#- Inject Thinking token at the very top of the FIRST system turn -#}
161
+ {%- if enable_thinking is defined and enable_thinking -%}
162
+ {{- '<|think|>' -}}
163
+ {%- set ns.prev_message_type = 'think' -%}
164
+ {%- endif -%}
165
+
166
+ {%- if messages[0]['role'] in ['system', 'developer'] -%}
167
+ {{- messages[0]['content'] | trim -}}
168
+ {%- set loop_messages = messages[1:] -%}
169
+ {%- endif -%}
170
+
171
+ {%- if tools -%}
172
+ {%- for tool in tools %}
173
+ {{- '<|tool>' -}}
174
+ {{- format_function_declaration(tool) | trim -}}
175
+ {{- '<tool|>' -}}
176
+ {%- endfor %}
177
+ {%- set ns.prev_message_type = 'tool' -%}
178
+ {%- endif -%}
179
+
180
+ {{- '<turn|>\n' -}}
181
+ {%- endif %}
182
+
183
+ {#- Loop through messages -#}
184
+ {%- for message in loop_messages -%}
185
+ {%- set ns.prev_message_type = None -%}
186
+ {%- set role = 'model' if message['role'] == 'assistant' else message['role'] -%}
187
+ {{- '<|turn>' + role + '\n' }}
188
+
189
+ {%- if message['tool_calls'] -%}
190
+ {%- for tool_call in message['tool_calls'] -%}
191
+ {%- set function = tool_call['function'] -%}
192
+ {{- '<|tool_call>call:' + function['name'] + '{' -}}
193
+ {%- if function['arguments'] is mapping -%}
194
+ {%- set ns_args = namespace(found_first=false) -%}
195
+ {%- for key, value in function['arguments'] | dictsort -%}
196
+ {%- if ns_args.found_first %},{% endif -%}
197
+ {%- set ns_args.found_first = true -%}
198
+ {{- key -}}:{{- format_argument(value, escape_keys=False) -}}
199
+ {%- endfor -%}
200
+ {%- elif function['arguments'] is string -%}
201
+ {{- function['arguments'] -}}
202
+ {%- endif -%}
203
+ {{- '}<tool_call|>' -}}
204
+ {%- endfor -%}
205
+ {%- set ns.prev_message_type = 'tool_call' -%}
206
+ {%- endif -%}
207
+
208
+ {%- if message['tool_responses'] -%}
209
+ {#- Tool Response handling -#}
210
+ {%- for tool_response in message['tool_responses'] -%}
211
+ {{- '<|tool_response>' -}}
212
+ {%- if tool_response['response'] is mapping -%}
213
+ {{- 'response:' + tool_response['name'] | default('unknown') + '{' -}}
214
+ {%- for key, value in tool_response['response'] | dictsort -%}
215
+ {{- key -}}:{{- format_argument(value, escape_keys=False) -}}
216
+ {%- if not loop.last %},{% endif -%}
217
+ {%- endfor -%}
218
+ {{- '}' -}}
219
+ {%- else -%}
220
+ {{- 'response:' + tool_response['name'] | default('unknown') + '{value:' + format_argument(tool_response['response'], escape_keys=False) + '}' -}}
221
+ {%- endif -%}
222
+ {{- '<tool_response|>' -}}
223
+ {%- endfor -%}
224
+ {%- set ns.prev_message_type = 'tool_response' -%}
225
+ {%- endif -%}
226
+
227
+ {%- if message['content'] is string -%}
228
+ {%- if role == 'model' -%}
229
+ {{- strip_thinking(message['content']) -}}
230
+ {%- else -%}
231
+ {{- message['content'] | trim -}}
232
+ {%- endif -%}
233
+ {%- elif message['content'] is sequence -%}
234
+ {%- for item in message['content'] -%}
235
+ {%- if item['type'] == 'text' -%}
236
+ {%- if role == 'model' -%}
237
+ {{- strip_thinking(item['text']) -}}
238
+ {%- else -%}
239
+ {{- item['text'] | trim -}}
240
+ {%- endif -%}
241
+ {%- elif item['type'] == 'image' -%}
242
+ {{- '\n\n<|image|>\n\n' -}}
243
+ {%- set ns.prev_message_type = 'image' -%}
244
+ {%- elif item['type'] == 'audio' -%}
245
+ {{- '<|audio|>' -}}
246
+ {%- set ns.prev_message_type = 'audio' -%}
247
+ {%- elif item['type'] == 'video' -%}
248
+ {{- '\n\n<|video|>\n\n' -}}
249
+ {%- set ns.prev_message_type = 'video' -%}
250
+ {%- endif -%}
251
+ {%- endfor -%}
252
+ {%- endif -%}
253
+
254
+ {%- if not (message['tool_responses'] and not message['content']) -%}
255
+ {{- '<turn|>\n' -}}
256
+ {%- endif -%}
257
+ {%- endfor -%}
258
+
259
+ {%- if add_generation_prompt -%}
260
+ {%- if ns.prev_message_type != 'tool_response' -%}
261
+ {{- '<|turn>model\n' -}}
262
+ {%- endif -%}
263
+ {%- if not enable_thinking | default(false) -%}
264
+ {{- '<|channel>thought\n<channel|>' -}}
265
+ {%- endif -%}
266
+ {%- endif -%}
config.json ADDED
@@ -0,0 +1,873 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "dtype": "bfloat16",
10
+ "eoa_token_id": 258883,
11
+ "eoa_token_index": 258883,
12
+ "eoi_token_id": 258882,
13
+ "eos_token_id": [
14
+ 1,
15
+ 106
16
+ ],
17
+ "image_token_id": 258880,
18
+ "initializer_range": 0.02,
19
+ "model_type": "gemma4",
20
+ "quantization_config": {
21
+ "autoround_version": "0.13.0",
22
+ "bits": 4,
23
+ "block_name_to_quantize": "model.language_model.layers",
24
+ "data_type": "int",
25
+ "extra_config": {
26
+ ".*model\\.language_model\\.layers\\.\\d+\\.mlp\\..*": {
27
+ "bits": 8
28
+ },
29
+ ".*model\\.language_model\\.layers\\.\\d+\\.router\\..*": {
30
+ "bits": 8
31
+ },
32
+ ".*model\\.language_model\\.layers\\.\\d+\\.self_attn\\..*": {
33
+ "bits": 8
34
+ },
35
+ "model.language_model.layers.0.mlp.down_proj": {
36
+ "bits": 8
37
+ },
38
+ "model.language_model.layers.0.mlp.gate_proj": {
39
+ "bits": 8
40
+ },
41
+ "model.language_model.layers.0.mlp.up_proj": {
42
+ "bits": 8
43
+ },
44
+ "model.language_model.layers.0.router.proj": {
45
+ "bits": 8
46
+ },
47
+ "model.language_model.layers.0.self_attn.k_proj": {
48
+ "bits": 8
49
+ },
50
+ "model.language_model.layers.0.self_attn.o_proj": {
51
+ "bits": 8
52
+ },
53
+ "model.language_model.layers.0.self_attn.q_proj": {
54
+ "bits": 8
55
+ },
56
+ "model.language_model.layers.0.self_attn.v_proj": {
57
+ "bits": 8
58
+ },
59
+ "model.language_model.layers.1.mlp.down_proj": {
60
+ "bits": 8
61
+ },
62
+ "model.language_model.layers.1.mlp.gate_proj": {
63
+ "bits": 8
64
+ },
65
+ "model.language_model.layers.1.mlp.up_proj": {
66
+ "bits": 8
67
+ },
68
+ "model.language_model.layers.1.router.proj": {
69
+ "bits": 8
70
+ },
71
+ "model.language_model.layers.1.self_attn.k_proj": {
72
+ "bits": 8
73
+ },
74
+ "model.language_model.layers.1.self_attn.o_proj": {
75
+ "bits": 8
76
+ },
77
+ "model.language_model.layers.1.self_attn.q_proj": {
78
+ "bits": 8
79
+ },
80
+ "model.language_model.layers.1.self_attn.v_proj": {
81
+ "bits": 8
82
+ },
83
+ "model.language_model.layers.10.mlp.down_proj": {
84
+ "bits": 8
85
+ },
86
+ "model.language_model.layers.10.mlp.gate_proj": {
87
+ "bits": 8
88
+ },
89
+ "model.language_model.layers.10.mlp.up_proj": {
90
+ "bits": 8
91
+ },
92
+ "model.language_model.layers.10.router.proj": {
93
+ "bits": 8
94
+ },
95
+ "model.language_model.layers.10.self_attn.k_proj": {
96
+ "bits": 8
97
+ },
98
+ "model.language_model.layers.10.self_attn.o_proj": {
99
+ "bits": 8
100
+ },
101
+ "model.language_model.layers.10.self_attn.q_proj": {
102
+ "bits": 8
103
+ },
104
+ "model.language_model.layers.10.self_attn.v_proj": {
105
+ "bits": 8
106
+ },
107
+ "model.language_model.layers.11.mlp.down_proj": {
108
+ "bits": 8
109
+ },
110
+ "model.language_model.layers.11.mlp.gate_proj": {
111
+ "bits": 8
112
+ },
113
+ "model.language_model.layers.11.mlp.up_proj": {
114
+ "bits": 8
115
+ },
116
+ "model.language_model.layers.11.router.proj": {
117
+ "bits": 8
118
+ },
119
+ "model.language_model.layers.11.self_attn.k_proj": {
120
+ "bits": 8
121
+ },
122
+ "model.language_model.layers.11.self_attn.o_proj": {
123
+ "bits": 8
124
+ },
125
+ "model.language_model.layers.11.self_attn.q_proj": {
126
+ "bits": 8
127
+ },
128
+ "model.language_model.layers.12.mlp.down_proj": {
129
+ "bits": 8
130
+ },
131
+ "model.language_model.layers.12.mlp.gate_proj": {
132
+ "bits": 8
133
+ },
134
+ "model.language_model.layers.12.mlp.up_proj": {
135
+ "bits": 8
136
+ },
137
+ "model.language_model.layers.12.router.proj": {
138
+ "bits": 8
139
+ },
140
+ "model.language_model.layers.12.self_attn.k_proj": {
141
+ "bits": 8
142
+ },
143
+ "model.language_model.layers.12.self_attn.o_proj": {
144
+ "bits": 8
145
+ },
146
+ "model.language_model.layers.12.self_attn.q_proj": {
147
+ "bits": 8
148
+ },
149
+ "model.language_model.layers.12.self_attn.v_proj": {
150
+ "bits": 8
151
+ },
152
+ "model.language_model.layers.13.mlp.down_proj": {
153
+ "bits": 8
154
+ },
155
+ "model.language_model.layers.13.mlp.gate_proj": {
156
+ "bits": 8
157
+ },
158
+ "model.language_model.layers.13.mlp.up_proj": {
159
+ "bits": 8
160
+ },
161
+ "model.language_model.layers.13.router.proj": {
162
+ "bits": 8
163
+ },
164
+ "model.language_model.layers.13.self_attn.k_proj": {
165
+ "bits": 8
166
+ },
167
+ "model.language_model.layers.13.self_attn.o_proj": {
168
+ "bits": 8
169
+ },
170
+ "model.language_model.layers.13.self_attn.q_proj": {
171
+ "bits": 8
172
+ },
173
+ "model.language_model.layers.13.self_attn.v_proj": {
174
+ "bits": 8
175
+ },
176
+ "model.language_model.layers.14.mlp.down_proj": {
177
+ "bits": 8
178
+ },
179
+ "model.language_model.layers.14.mlp.gate_proj": {
180
+ "bits": 8
181
+ },
182
+ "model.language_model.layers.14.mlp.up_proj": {
183
+ "bits": 8
184
+ },
185
+ "model.language_model.layers.14.router.proj": {
186
+ "bits": 8
187
+ },
188
+ "model.language_model.layers.14.self_attn.k_proj": {
189
+ "bits": 8
190
+ },
191
+ "model.language_model.layers.14.self_attn.o_proj": {
192
+ "bits": 8
193
+ },
194
+ "model.language_model.layers.14.self_attn.q_proj": {
195
+ "bits": 8
196
+ },
197
+ "model.language_model.layers.14.self_attn.v_proj": {
198
+ "bits": 8
199
+ },
200
+ "model.language_model.layers.15.mlp.down_proj": {
201
+ "bits": 8
202
+ },
203
+ "model.language_model.layers.15.mlp.gate_proj": {
204
+ "bits": 8
205
+ },
206
+ "model.language_model.layers.15.mlp.up_proj": {
207
+ "bits": 8
208
+ },
209
+ "model.language_model.layers.15.router.proj": {
210
+ "bits": 8
211
+ },
212
+ "model.language_model.layers.15.self_attn.k_proj": {
213
+ "bits": 8
214
+ },
215
+ "model.language_model.layers.15.self_attn.o_proj": {
216
+ "bits": 8
217
+ },
218
+ "model.language_model.layers.15.self_attn.q_proj": {
219
+ "bits": 8
220
+ },
221
+ "model.language_model.layers.15.self_attn.v_proj": {
222
+ "bits": 8
223
+ },
224
+ "model.language_model.layers.16.mlp.down_proj": {
225
+ "bits": 8
226
+ },
227
+ "model.language_model.layers.16.mlp.gate_proj": {
228
+ "bits": 8
229
+ },
230
+ "model.language_model.layers.16.mlp.up_proj": {
231
+ "bits": 8
232
+ },
233
+ "model.language_model.layers.16.router.proj": {
234
+ "bits": 8
235
+ },
236
+ "model.language_model.layers.16.self_attn.k_proj": {
237
+ "bits": 8
238
+ },
239
+ "model.language_model.layers.16.self_attn.o_proj": {
240
+ "bits": 8
241
+ },
242
+ "model.language_model.layers.16.self_attn.q_proj": {
243
+ "bits": 8
244
+ },
245
+ "model.language_model.layers.16.self_attn.v_proj": {
246
+ "bits": 8
247
+ },
248
+ "model.language_model.layers.17.mlp.down_proj": {
249
+ "bits": 8
250
+ },
251
+ "model.language_model.layers.17.mlp.gate_proj": {
252
+ "bits": 8
253
+ },
254
+ "model.language_model.layers.17.mlp.up_proj": {
255
+ "bits": 8
256
+ },
257
+ "model.language_model.layers.17.router.proj": {
258
+ "bits": 8
259
+ },
260
+ "model.language_model.layers.17.self_attn.k_proj": {
261
+ "bits": 8
262
+ },
263
+ "model.language_model.layers.17.self_attn.o_proj": {
264
+ "bits": 8
265
+ },
266
+ "model.language_model.layers.17.self_attn.q_proj": {
267
+ "bits": 8
268
+ },
269
+ "model.language_model.layers.18.mlp.down_proj": {
270
+ "bits": 8
271
+ },
272
+ "model.language_model.layers.18.mlp.gate_proj": {
273
+ "bits": 8
274
+ },
275
+ "model.language_model.layers.18.mlp.up_proj": {
276
+ "bits": 8
277
+ },
278
+ "model.language_model.layers.18.router.proj": {
279
+ "bits": 8
280
+ },
281
+ "model.language_model.layers.18.self_attn.k_proj": {
282
+ "bits": 8
283
+ },
284
+ "model.language_model.layers.18.self_attn.o_proj": {
285
+ "bits": 8
286
+ },
287
+ "model.language_model.layers.18.self_attn.q_proj": {
288
+ "bits": 8
289
+ },
290
+ "model.language_model.layers.18.self_attn.v_proj": {
291
+ "bits": 8
292
+ },
293
+ "model.language_model.layers.19.mlp.down_proj": {
294
+ "bits": 8
295
+ },
296
+ "model.language_model.layers.19.mlp.gate_proj": {
297
+ "bits": 8
298
+ },
299
+ "model.language_model.layers.19.mlp.up_proj": {
300
+ "bits": 8
301
+ },
302
+ "model.language_model.layers.19.router.proj": {
303
+ "bits": 8
304
+ },
305
+ "model.language_model.layers.19.self_attn.k_proj": {
306
+ "bits": 8
307
+ },
308
+ "model.language_model.layers.19.self_attn.o_proj": {
309
+ "bits": 8
310
+ },
311
+ "model.language_model.layers.19.self_attn.q_proj": {
312
+ "bits": 8
313
+ },
314
+ "model.language_model.layers.19.self_attn.v_proj": {
315
+ "bits": 8
316
+ },
317
+ "model.language_model.layers.2.mlp.down_proj": {
318
+ "bits": 8
319
+ },
320
+ "model.language_model.layers.2.mlp.gate_proj": {
321
+ "bits": 8
322
+ },
323
+ "model.language_model.layers.2.mlp.up_proj": {
324
+ "bits": 8
325
+ },
326
+ "model.language_model.layers.2.router.proj": {
327
+ "bits": 8
328
+ },
329
+ "model.language_model.layers.2.self_attn.k_proj": {
330
+ "bits": 8
331
+ },
332
+ "model.language_model.layers.2.self_attn.o_proj": {
333
+ "bits": 8
334
+ },
335
+ "model.language_model.layers.2.self_attn.q_proj": {
336
+ "bits": 8
337
+ },
338
+ "model.language_model.layers.2.self_attn.v_proj": {
339
+ "bits": 8
340
+ },
341
+ "model.language_model.layers.20.mlp.down_proj": {
342
+ "bits": 8
343
+ },
344
+ "model.language_model.layers.20.mlp.gate_proj": {
345
+ "bits": 8
346
+ },
347
+ "model.language_model.layers.20.mlp.up_proj": {
348
+ "bits": 8
349
+ },
350
+ "model.language_model.layers.20.router.proj": {
351
+ "bits": 8
352
+ },
353
+ "model.language_model.layers.20.self_attn.k_proj": {
354
+ "bits": 8
355
+ },
356
+ "model.language_model.layers.20.self_attn.o_proj": {
357
+ "bits": 8
358
+ },
359
+ "model.language_model.layers.20.self_attn.q_proj": {
360
+ "bits": 8
361
+ },
362
+ "model.language_model.layers.20.self_attn.v_proj": {
363
+ "bits": 8
364
+ },
365
+ "model.language_model.layers.21.mlp.down_proj": {
366
+ "bits": 8
367
+ },
368
+ "model.language_model.layers.21.mlp.gate_proj": {
369
+ "bits": 8
370
+ },
371
+ "model.language_model.layers.21.mlp.up_proj": {
372
+ "bits": 8
373
+ },
374
+ "model.language_model.layers.21.router.proj": {
375
+ "bits": 8
376
+ },
377
+ "model.language_model.layers.21.self_attn.k_proj": {
378
+ "bits": 8
379
+ },
380
+ "model.language_model.layers.21.self_attn.o_proj": {
381
+ "bits": 8
382
+ },
383
+ "model.language_model.layers.21.self_attn.q_proj": {
384
+ "bits": 8
385
+ },
386
+ "model.language_model.layers.21.self_attn.v_proj": {
387
+ "bits": 8
388
+ },
389
+ "model.language_model.layers.22.mlp.down_proj": {
390
+ "bits": 8
391
+ },
392
+ "model.language_model.layers.22.mlp.gate_proj": {
393
+ "bits": 8
394
+ },
395
+ "model.language_model.layers.22.mlp.up_proj": {
396
+ "bits": 8
397
+ },
398
+ "model.language_model.layers.22.router.proj": {
399
+ "bits": 8
400
+ },
401
+ "model.language_model.layers.22.self_attn.k_proj": {
402
+ "bits": 8
403
+ },
404
+ "model.language_model.layers.22.self_attn.o_proj": {
405
+ "bits": 8
406
+ },
407
+ "model.language_model.layers.22.self_attn.q_proj": {
408
+ "bits": 8
409
+ },
410
+ "model.language_model.layers.22.self_attn.v_proj": {
411
+ "bits": 8
412
+ },
413
+ "model.language_model.layers.23.mlp.down_proj": {
414
+ "bits": 8
415
+ },
416
+ "model.language_model.layers.23.mlp.gate_proj": {
417
+ "bits": 8
418
+ },
419
+ "model.language_model.layers.23.mlp.up_proj": {
420
+ "bits": 8
421
+ },
422
+ "model.language_model.layers.23.router.proj": {
423
+ "bits": 8
424
+ },
425
+ "model.language_model.layers.23.self_attn.k_proj": {
426
+ "bits": 8
427
+ },
428
+ "model.language_model.layers.23.self_attn.o_proj": {
429
+ "bits": 8
430
+ },
431
+ "model.language_model.layers.23.self_attn.q_proj": {
432
+ "bits": 8
433
+ },
434
+ "model.language_model.layers.24.mlp.down_proj": {
435
+ "bits": 8
436
+ },
437
+ "model.language_model.layers.24.mlp.gate_proj": {
438
+ "bits": 8
439
+ },
440
+ "model.language_model.layers.24.mlp.up_proj": {
441
+ "bits": 8
442
+ },
443
+ "model.language_model.layers.24.router.proj": {
444
+ "bits": 8
445
+ },
446
+ "model.language_model.layers.24.self_attn.k_proj": {
447
+ "bits": 8
448
+ },
449
+ "model.language_model.layers.24.self_attn.o_proj": {
450
+ "bits": 8
451
+ },
452
+ "model.language_model.layers.24.self_attn.q_proj": {
453
+ "bits": 8
454
+ },
455
+ "model.language_model.layers.24.self_attn.v_proj": {
456
+ "bits": 8
457
+ },
458
+ "model.language_model.layers.25.mlp.down_proj": {
459
+ "bits": 8
460
+ },
461
+ "model.language_model.layers.25.mlp.gate_proj": {
462
+ "bits": 8
463
+ },
464
+ "model.language_model.layers.25.mlp.up_proj": {
465
+ "bits": 8
466
+ },
467
+ "model.language_model.layers.25.router.proj": {
468
+ "bits": 8
469
+ },
470
+ "model.language_model.layers.25.self_attn.k_proj": {
471
+ "bits": 8
472
+ },
473
+ "model.language_model.layers.25.self_attn.o_proj": {
474
+ "bits": 8
475
+ },
476
+ "model.language_model.layers.25.self_attn.q_proj": {
477
+ "bits": 8
478
+ },
479
+ "model.language_model.layers.25.self_attn.v_proj": {
480
+ "bits": 8
481
+ },
482
+ "model.language_model.layers.26.mlp.down_proj": {
483
+ "bits": 8
484
+ },
485
+ "model.language_model.layers.26.mlp.gate_proj": {
486
+ "bits": 8
487
+ },
488
+ "model.language_model.layers.26.mlp.up_proj": {
489
+ "bits": 8
490
+ },
491
+ "model.language_model.layers.26.router.proj": {
492
+ "bits": 8
493
+ },
494
+ "model.language_model.layers.26.self_attn.k_proj": {
495
+ "bits": 8
496
+ },
497
+ "model.language_model.layers.26.self_attn.o_proj": {
498
+ "bits": 8
499
+ },
500
+ "model.language_model.layers.26.self_attn.q_proj": {
501
+ "bits": 8
502
+ },
503
+ "model.language_model.layers.26.self_attn.v_proj": {
504
+ "bits": 8
505
+ },
506
+ "model.language_model.layers.27.mlp.down_proj": {
507
+ "bits": 8
508
+ },
509
+ "model.language_model.layers.27.mlp.gate_proj": {
510
+ "bits": 8
511
+ },
512
+ "model.language_model.layers.27.mlp.up_proj": {
513
+ "bits": 8
514
+ },
515
+ "model.language_model.layers.27.router.proj": {
516
+ "bits": 8
517
+ },
518
+ "model.language_model.layers.27.self_attn.k_proj": {
519
+ "bits": 8
520
+ },
521
+ "model.language_model.layers.27.self_attn.o_proj": {
522
+ "bits": 8
523
+ },
524
+ "model.language_model.layers.27.self_attn.q_proj": {
525
+ "bits": 8
526
+ },
527
+ "model.language_model.layers.27.self_attn.v_proj": {
528
+ "bits": 8
529
+ },
530
+ "model.language_model.layers.28.mlp.down_proj": {
531
+ "bits": 8
532
+ },
533
+ "model.language_model.layers.28.mlp.gate_proj": {
534
+ "bits": 8
535
+ },
536
+ "model.language_model.layers.28.mlp.up_proj": {
537
+ "bits": 8
538
+ },
539
+ "model.language_model.layers.28.router.proj": {
540
+ "bits": 8
541
+ },
542
+ "model.language_model.layers.28.self_attn.k_proj": {
543
+ "bits": 8
544
+ },
545
+ "model.language_model.layers.28.self_attn.o_proj": {
546
+ "bits": 8
547
+ },
548
+ "model.language_model.layers.28.self_attn.q_proj": {
549
+ "bits": 8
550
+ },
551
+ "model.language_model.layers.28.self_attn.v_proj": {
552
+ "bits": 8
553
+ },
554
+ "model.language_model.layers.29.mlp.down_proj": {
555
+ "bits": 8
556
+ },
557
+ "model.language_model.layers.29.mlp.gate_proj": {
558
+ "bits": 8
559
+ },
560
+ "model.language_model.layers.29.mlp.up_proj": {
561
+ "bits": 8
562
+ },
563
+ "model.language_model.layers.29.router.proj": {
564
+ "bits": 8
565
+ },
566
+ "model.language_model.layers.29.self_attn.k_proj": {
567
+ "bits": 8
568
+ },
569
+ "model.language_model.layers.29.self_attn.o_proj": {
570
+ "bits": 8
571
+ },
572
+ "model.language_model.layers.29.self_attn.q_proj": {
573
+ "bits": 8
574
+ },
575
+ "model.language_model.layers.3.mlp.down_proj": {
576
+ "bits": 8
577
+ },
578
+ "model.language_model.layers.3.mlp.gate_proj": {
579
+ "bits": 8
580
+ },
581
+ "model.language_model.layers.3.mlp.up_proj": {
582
+ "bits": 8
583
+ },
584
+ "model.language_model.layers.3.router.proj": {
585
+ "bits": 8
586
+ },
587
+ "model.language_model.layers.3.self_attn.k_proj": {
588
+ "bits": 8
589
+ },
590
+ "model.language_model.layers.3.self_attn.o_proj": {
591
+ "bits": 8
592
+ },
593
+ "model.language_model.layers.3.self_attn.q_proj": {
594
+ "bits": 8
595
+ },
596
+ "model.language_model.layers.3.self_attn.v_proj": {
597
+ "bits": 8
598
+ },
599
+ "model.language_model.layers.4.mlp.down_proj": {
600
+ "bits": 8
601
+ },
602
+ "model.language_model.layers.4.mlp.gate_proj": {
603
+ "bits": 8
604
+ },
605
+ "model.language_model.layers.4.mlp.up_proj": {
606
+ "bits": 8
607
+ },
608
+ "model.language_model.layers.4.router.proj": {
609
+ "bits": 8
610
+ },
611
+ "model.language_model.layers.4.self_attn.k_proj": {
612
+ "bits": 8
613
+ },
614
+ "model.language_model.layers.4.self_attn.o_proj": {
615
+ "bits": 8
616
+ },
617
+ "model.language_model.layers.4.self_attn.q_proj": {
618
+ "bits": 8
619
+ },
620
+ "model.language_model.layers.4.self_attn.v_proj": {
621
+ "bits": 8
622
+ },
623
+ "model.language_model.layers.5.mlp.down_proj": {
624
+ "bits": 8
625
+ },
626
+ "model.language_model.layers.5.mlp.gate_proj": {
627
+ "bits": 8
628
+ },
629
+ "model.language_model.layers.5.mlp.up_proj": {
630
+ "bits": 8
631
+ },
632
+ "model.language_model.layers.5.router.proj": {
633
+ "bits": 8
634
+ },
635
+ "model.language_model.layers.5.self_attn.k_proj": {
636
+ "bits": 8
637
+ },
638
+ "model.language_model.layers.5.self_attn.o_proj": {
639
+ "bits": 8
640
+ },
641
+ "model.language_model.layers.5.self_attn.q_proj": {
642
+ "bits": 8
643
+ },
644
+ "model.language_model.layers.6.mlp.down_proj": {
645
+ "bits": 8
646
+ },
647
+ "model.language_model.layers.6.mlp.gate_proj": {
648
+ "bits": 8
649
+ },
650
+ "model.language_model.layers.6.mlp.up_proj": {
651
+ "bits": 8
652
+ },
653
+ "model.language_model.layers.6.router.proj": {
654
+ "bits": 8
655
+ },
656
+ "model.language_model.layers.6.self_attn.k_proj": {
657
+ "bits": 8
658
+ },
659
+ "model.language_model.layers.6.self_attn.o_proj": {
660
+ "bits": 8
661
+ },
662
+ "model.language_model.layers.6.self_attn.q_proj": {
663
+ "bits": 8
664
+ },
665
+ "model.language_model.layers.6.self_attn.v_proj": {
666
+ "bits": 8
667
+ },
668
+ "model.language_model.layers.7.mlp.down_proj": {
669
+ "bits": 8
670
+ },
671
+ "model.language_model.layers.7.mlp.gate_proj": {
672
+ "bits": 8
673
+ },
674
+ "model.language_model.layers.7.mlp.up_proj": {
675
+ "bits": 8
676
+ },
677
+ "model.language_model.layers.7.router.proj": {
678
+ "bits": 8
679
+ },
680
+ "model.language_model.layers.7.self_attn.k_proj": {
681
+ "bits": 8
682
+ },
683
+ "model.language_model.layers.7.self_attn.o_proj": {
684
+ "bits": 8
685
+ },
686
+ "model.language_model.layers.7.self_attn.q_proj": {
687
+ "bits": 8
688
+ },
689
+ "model.language_model.layers.7.self_attn.v_proj": {
690
+ "bits": 8
691
+ },
692
+ "model.language_model.layers.8.mlp.down_proj": {
693
+ "bits": 8
694
+ },
695
+ "model.language_model.layers.8.mlp.gate_proj": {
696
+ "bits": 8
697
+ },
698
+ "model.language_model.layers.8.mlp.up_proj": {
699
+ "bits": 8
700
+ },
701
+ "model.language_model.layers.8.router.proj": {
702
+ "bits": 8
703
+ },
704
+ "model.language_model.layers.8.self_attn.k_proj": {
705
+ "bits": 8
706
+ },
707
+ "model.language_model.layers.8.self_attn.o_proj": {
708
+ "bits": 8
709
+ },
710
+ "model.language_model.layers.8.self_attn.q_proj": {
711
+ "bits": 8
712
+ },
713
+ "model.language_model.layers.8.self_attn.v_proj": {
714
+ "bits": 8
715
+ },
716
+ "model.language_model.layers.9.mlp.down_proj": {
717
+ "bits": 8
718
+ },
719
+ "model.language_model.layers.9.mlp.gate_proj": {
720
+ "bits": 8
721
+ },
722
+ "model.language_model.layers.9.mlp.up_proj": {
723
+ "bits": 8
724
+ },
725
+ "model.language_model.layers.9.router.proj": {
726
+ "bits": 8
727
+ },
728
+ "model.language_model.layers.9.self_attn.k_proj": {
729
+ "bits": 8
730
+ },
731
+ "model.language_model.layers.9.self_attn.o_proj": {
732
+ "bits": 8
733
+ },
734
+ "model.language_model.layers.9.self_attn.q_proj": {
735
+ "bits": 8
736
+ },
737
+ "model.language_model.layers.9.self_attn.v_proj": {
738
+ "bits": 8
739
+ }
740
+ },
741
+ "group_size": 128,
742
+ "iters": 0,
743
+ "packing_format": "auto_round:auto_gptq",
744
+ "quant_method": "auto-round",
745
+ "sym": true
746
+ },
747
+ "text_config": {
748
+ "attention_bias": false,
749
+ "attention_dropout": 0.0,
750
+ "attention_k_eq_v": true,
751
+ "bos_token_id": 2,
752
+ "dtype": "bfloat16",
753
+ "enable_moe_block": true,
754
+ "eos_token_id": 1,
755
+ "final_logit_softcapping": 30.0,
756
+ "global_head_dim": 512,
757
+ "head_dim": 256,
758
+ "hidden_activation": "gelu_pytorch_tanh",
759
+ "hidden_size": 2816,
760
+ "hidden_size_per_layer_input": 0,
761
+ "initializer_range": 0.02,
762
+ "intermediate_size": 2112,
763
+ "layer_types": [
764
+ "sliding_attention",
765
+ "sliding_attention",
766
+ "sliding_attention",
767
+ "sliding_attention",
768
+ "sliding_attention",
769
+ "full_attention",
770
+ "sliding_attention",
771
+ "sliding_attention",
772
+ "sliding_attention",
773
+ "sliding_attention",
774
+ "sliding_attention",
775
+ "full_attention",
776
+ "sliding_attention",
777
+ "sliding_attention",
778
+ "sliding_attention",
779
+ "sliding_attention",
780
+ "sliding_attention",
781
+ "full_attention",
782
+ "sliding_attention",
783
+ "sliding_attention",
784
+ "sliding_attention",
785
+ "sliding_attention",
786
+ "sliding_attention",
787
+ "full_attention",
788
+ "sliding_attention",
789
+ "sliding_attention",
790
+ "sliding_attention",
791
+ "sliding_attention",
792
+ "sliding_attention",
793
+ "full_attention"
794
+ ],
795
+ "max_position_embeddings": 262144,
796
+ "model_type": "gemma4_text",
797
+ "moe_intermediate_size": 704,
798
+ "num_attention_heads": 16,
799
+ "num_experts": 128,
800
+ "num_global_key_value_heads": 2,
801
+ "num_hidden_layers": 30,
802
+ "num_key_value_heads": 8,
803
+ "num_kv_shared_layers": 0,
804
+ "pad_token_id": 0,
805
+ "rms_norm_eps": 1e-06,
806
+ "rope_parameters": {
807
+ "full_attention": {
808
+ "partial_rotary_factor": 0.25,
809
+ "rope_theta": 1000000.0,
810
+ "rope_type": "proportional"
811
+ },
812
+ "sliding_attention": {
813
+ "rope_theta": 10000.0,
814
+ "rope_type": "default"
815
+ }
816
+ },
817
+ "sliding_window": 1024,
818
+ "tie_word_embeddings": true,
819
+ "top_k_experts": 8,
820
+ "use_bidirectional_attention": "vision",
821
+ "use_cache": true,
822
+ "use_double_wide_mlp": false,
823
+ "vocab_size": 262144,
824
+ "vocab_size_per_layer_input": 262144
825
+ },
826
+ "tie_word_embeddings": true,
827
+ "transformers_version": "5.5.0",
828
+ "video_token_id": 258884,
829
+ "vision_config": {
830
+ "_name_or_path": "",
831
+ "architectures": null,
832
+ "attention_bias": false,
833
+ "attention_dropout": 0.0,
834
+ "chunk_size_feed_forward": 0,
835
+ "default_output_length": 280,
836
+ "dtype": "bfloat16",
837
+ "global_head_dim": 72,
838
+ "head_dim": 72,
839
+ "hidden_activation": "gelu_pytorch_tanh",
840
+ "hidden_size": 1152,
841
+ "id2label": {
842
+ "0": "LABEL_0",
843
+ "1": "LABEL_1"
844
+ },
845
+ "initializer_range": 0.02,
846
+ "intermediate_size": 4304,
847
+ "is_encoder_decoder": false,
848
+ "label2id": {
849
+ "LABEL_0": 0,
850
+ "LABEL_1": 1
851
+ },
852
+ "max_position_embeddings": 131072,
853
+ "model_type": "gemma4_vision",
854
+ "num_attention_heads": 16,
855
+ "num_hidden_layers": 27,
856
+ "num_key_value_heads": 16,
857
+ "output_attentions": false,
858
+ "output_hidden_states": false,
859
+ "patch_size": 16,
860
+ "pooling_kernel_size": 3,
861
+ "position_embedding_size": 10240,
862
+ "problem_type": null,
863
+ "return_dict": true,
864
+ "rms_norm_eps": 1e-06,
865
+ "rope_parameters": {
866
+ "rope_theta": 100.0,
867
+ "rope_type": "default"
868
+ },
869
+ "standardize": true,
870
+ "use_clipped_linears": false
871
+ },
872
+ "vision_soft_tokens_per_image": 280
873
+ }
generation_config.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 2,
3
+ "do_sample": true,
4
+ "eos_token_id": [
5
+ 1,
6
+ 106,
7
+ 50
8
+ ],
9
+ "pad_token_id": 0,
10
+ "temperature": 1.0,
11
+ "top_k": 64,
12
+ "top_p": 0.95,
13
+ "transformers_version": "5.5.0"
14
+ }
model-00001-of-00009.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:159f88bd73722eab8ba986ea66b8898ec7263b37009440b9b494c5527fb09596
3
+ size 2147833296
model-00002-of-00009.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fc3f583445cc56a1a0f81970e8c4e1e711e146290dad323f3c21079667f11d8e
3
+ size 2148171178
model-00003-of-00009.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8b6d37d1170ff13c79ea9e80d831428daa23a72235f4696ee3337f57c3aa7d09
3
+ size 2148176130
model-00004-of-00009.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:50e6fec593eccdfd77039f5c54add563cbd3444e5e5ffa61890f5d802e29861e
3
+ size 2147716168
model-00005-of-00009.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d64a2e4206a1764afdac8ab6acde6692c344f848c4c9b1dd2c2f9762b391176c
3
+ size 2148176482
model-00006-of-00009.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cb25e670e9a18df80a7ea8924dabc002bb33f149b7f57d5ef95783f3e91b2b88
3
+ size 2147862746
model-00007-of-00009.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:22e34d6141d700cf9fbeace304fd464bfdbb747f103bd05f4ac438ef06d3ab8e
3
+ size 691913700
model-00008-of-00009.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f25ffc75beeb6f0db75d6d671030598f318a03a87794119d81dfd28382d5e9ce
3
+ size 2141640496
model-00009-of-00009.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f2333fbf6106311ab5e6e3b7b09046f9792f999e15de75d752ca588e3663ca09
3
+ size 480398920
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
preprocessor_config.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_convert_rgb": true,
3
+ "do_normalize": false,
4
+ "do_rescale": true,
5
+ "do_resize": true,
6
+ "image_mean": [
7
+ 0.0,
8
+ 0.0,
9
+ 0.0
10
+ ],
11
+ "image_processor_type": "Gemma4ImageProcessor",
12
+ "image_seq_length": 280,
13
+ "image_std": [
14
+ 1.0,
15
+ 1.0,
16
+ 1.0
17
+ ],
18
+ "max_soft_tokens": 280,
19
+ "patch_size": 16,
20
+ "pooling_kernel_size": 3,
21
+ "resample": 3,
22
+ "rescale_factor": 0.00392156862745098
23
+ }
processor_config.json ADDED
@@ -0,0 +1,75 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "audio_ms_per_token": 40,
3
+ "audio_seq_length": 750,
4
+ "feature_extractor": {
5
+ "dither": 0.0,
6
+ "feature_extractor_type": "Gemma4AudioFeatureExtractor",
7
+ "feature_size": 128,
8
+ "fft_length": 512,
9
+ "fft_overdrive": false,
10
+ "frame_length": 320,
11
+ "hop_length": 160,
12
+ "input_scale_factor": 1.0,
13
+ "max_frequency": 8000.0,
14
+ "mel_floor": 0.001,
15
+ "min_frequency": 0.0,
16
+ "padding_side": "right",
17
+ "padding_value": 0.0,
18
+ "per_bin_mean": null,
19
+ "per_bin_stddev": null,
20
+ "preemphasis": 0.0,
21
+ "preemphasis_htk_flavor": true,
22
+ "return_attention_mask": true,
23
+ "sampling_rate": 16000
24
+ },
25
+ "image_processor": {
26
+ "do_convert_rgb": true,
27
+ "do_normalize": false,
28
+ "do_rescale": true,
29
+ "do_resize": true,
30
+ "image_mean": [
31
+ 0.0,
32
+ 0.0,
33
+ 0.0
34
+ ],
35
+ "image_processor_type": "Gemma4ImageProcessor",
36
+ "image_seq_length": 280,
37
+ "image_std": [
38
+ 1.0,
39
+ 1.0,
40
+ 1.0
41
+ ],
42
+ "max_soft_tokens": 280,
43
+ "patch_size": 16,
44
+ "pooling_kernel_size": 3,
45
+ "resample": 3,
46
+ "rescale_factor": 0.00392156862745098
47
+ },
48
+ "image_seq_length": 280,
49
+ "processor_class": "Gemma4Processor",
50
+ "video_processor": {
51
+ "do_convert_rgb": true,
52
+ "do_normalize": true,
53
+ "do_rescale": true,
54
+ "do_resize": true,
55
+ "do_sample_frames": true,
56
+ "image_mean": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.0
60
+ ],
61
+ "image_std": [
62
+ 1.0,
63
+ 1.0,
64
+ 1.0
65
+ ],
66
+ "max_soft_tokens": 70,
67
+ "num_frames": 32,
68
+ "patch_size": 16,
69
+ "pooling_kernel_size": 3,
70
+ "resample": 3,
71
+ "rescale_factor": 0.00392156862745098,
72
+ "return_metadata": false,
73
+ "video_processor_type": "Gemma4VideoProcessor"
74
+ }
75
+ }
quantization_config.json ADDED
@@ -0,0 +1,727 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bits": 4,
3
+ "data_type": "int",
4
+ "group_size": 128,
5
+ "sym": true,
6
+ "iters": 0,
7
+ "autoround_version": "0.13.0",
8
+ "block_name_to_quantize": "model.language_model.layers",
9
+ "quant_method": "auto-round",
10
+ "packing_format": "auto_round:auto_gptq",
11
+ "extra_config": {
12
+ "model.language_model.layers.0.self_attn.k_proj": {
13
+ "bits": 8
14
+ },
15
+ "model.language_model.layers.0.self_attn.q_proj": {
16
+ "bits": 8
17
+ },
18
+ "model.language_model.layers.0.self_attn.v_proj": {
19
+ "bits": 8
20
+ },
21
+ "model.language_model.layers.0.self_attn.o_proj": {
22
+ "bits": 8
23
+ },
24
+ "model.language_model.layers.1.self_attn.k_proj": {
25
+ "bits": 8
26
+ },
27
+ "model.language_model.layers.1.self_attn.q_proj": {
28
+ "bits": 8
29
+ },
30
+ "model.language_model.layers.1.self_attn.v_proj": {
31
+ "bits": 8
32
+ },
33
+ "model.language_model.layers.1.self_attn.o_proj": {
34
+ "bits": 8
35
+ },
36
+ "model.language_model.layers.2.self_attn.k_proj": {
37
+ "bits": 8
38
+ },
39
+ "model.language_model.layers.2.self_attn.q_proj": {
40
+ "bits": 8
41
+ },
42
+ "model.language_model.layers.2.self_attn.v_proj": {
43
+ "bits": 8
44
+ },
45
+ "model.language_model.layers.2.self_attn.o_proj": {
46
+ "bits": 8
47
+ },
48
+ "model.language_model.layers.3.self_attn.k_proj": {
49
+ "bits": 8
50
+ },
51
+ "model.language_model.layers.3.self_attn.q_proj": {
52
+ "bits": 8
53
+ },
54
+ "model.language_model.layers.3.self_attn.v_proj": {
55
+ "bits": 8
56
+ },
57
+ "model.language_model.layers.3.self_attn.o_proj": {
58
+ "bits": 8
59
+ },
60
+ "model.language_model.layers.4.self_attn.k_proj": {
61
+ "bits": 8
62
+ },
63
+ "model.language_model.layers.4.self_attn.q_proj": {
64
+ "bits": 8
65
+ },
66
+ "model.language_model.layers.4.self_attn.v_proj": {
67
+ "bits": 8
68
+ },
69
+ "model.language_model.layers.4.self_attn.o_proj": {
70
+ "bits": 8
71
+ },
72
+ "model.language_model.layers.5.self_attn.k_proj": {
73
+ "bits": 8
74
+ },
75
+ "model.language_model.layers.5.self_attn.q_proj": {
76
+ "bits": 8
77
+ },
78
+ "model.language_model.layers.5.self_attn.o_proj": {
79
+ "bits": 8
80
+ },
81
+ "model.language_model.layers.6.self_attn.k_proj": {
82
+ "bits": 8
83
+ },
84
+ "model.language_model.layers.6.self_attn.q_proj": {
85
+ "bits": 8
86
+ },
87
+ "model.language_model.layers.6.self_attn.v_proj": {
88
+ "bits": 8
89
+ },
90
+ "model.language_model.layers.6.self_attn.o_proj": {
91
+ "bits": 8
92
+ },
93
+ "model.language_model.layers.7.self_attn.k_proj": {
94
+ "bits": 8
95
+ },
96
+ "model.language_model.layers.7.self_attn.q_proj": {
97
+ "bits": 8
98
+ },
99
+ "model.language_model.layers.7.self_attn.v_proj": {
100
+ "bits": 8
101
+ },
102
+ "model.language_model.layers.7.self_attn.o_proj": {
103
+ "bits": 8
104
+ },
105
+ "model.language_model.layers.8.self_attn.k_proj": {
106
+ "bits": 8
107
+ },
108
+ "model.language_model.layers.8.self_attn.q_proj": {
109
+ "bits": 8
110
+ },
111
+ "model.language_model.layers.8.self_attn.v_proj": {
112
+ "bits": 8
113
+ },
114
+ "model.language_model.layers.8.self_attn.o_proj": {
115
+ "bits": 8
116
+ },
117
+ "model.language_model.layers.9.self_attn.k_proj": {
118
+ "bits": 8
119
+ },
120
+ "model.language_model.layers.9.self_attn.q_proj": {
121
+ "bits": 8
122
+ },
123
+ "model.language_model.layers.9.self_attn.v_proj": {
124
+ "bits": 8
125
+ },
126
+ "model.language_model.layers.9.self_attn.o_proj": {
127
+ "bits": 8
128
+ },
129
+ "model.language_model.layers.10.self_attn.k_proj": {
130
+ "bits": 8
131
+ },
132
+ "model.language_model.layers.10.self_attn.q_proj": {
133
+ "bits": 8
134
+ },
135
+ "model.language_model.layers.10.self_attn.v_proj": {
136
+ "bits": 8
137
+ },
138
+ "model.language_model.layers.10.self_attn.o_proj": {
139
+ "bits": 8
140
+ },
141
+ "model.language_model.layers.11.self_attn.k_proj": {
142
+ "bits": 8
143
+ },
144
+ "model.language_model.layers.11.self_attn.q_proj": {
145
+ "bits": 8
146
+ },
147
+ "model.language_model.layers.11.self_attn.o_proj": {
148
+ "bits": 8
149
+ },
150
+ "model.language_model.layers.12.self_attn.k_proj": {
151
+ "bits": 8
152
+ },
153
+ "model.language_model.layers.12.self_attn.q_proj": {
154
+ "bits": 8
155
+ },
156
+ "model.language_model.layers.12.self_attn.v_proj": {
157
+ "bits": 8
158
+ },
159
+ "model.language_model.layers.12.self_attn.o_proj": {
160
+ "bits": 8
161
+ },
162
+ "model.language_model.layers.13.self_attn.k_proj": {
163
+ "bits": 8
164
+ },
165
+ "model.language_model.layers.13.self_attn.q_proj": {
166
+ "bits": 8
167
+ },
168
+ "model.language_model.layers.13.self_attn.v_proj": {
169
+ "bits": 8
170
+ },
171
+ "model.language_model.layers.13.self_attn.o_proj": {
172
+ "bits": 8
173
+ },
174
+ "model.language_model.layers.14.self_attn.k_proj": {
175
+ "bits": 8
176
+ },
177
+ "model.language_model.layers.14.self_attn.q_proj": {
178
+ "bits": 8
179
+ },
180
+ "model.language_model.layers.14.self_attn.v_proj": {
181
+ "bits": 8
182
+ },
183
+ "model.language_model.layers.14.self_attn.o_proj": {
184
+ "bits": 8
185
+ },
186
+ "model.language_model.layers.15.self_attn.k_proj": {
187
+ "bits": 8
188
+ },
189
+ "model.language_model.layers.15.self_attn.q_proj": {
190
+ "bits": 8
191
+ },
192
+ "model.language_model.layers.15.self_attn.v_proj": {
193
+ "bits": 8
194
+ },
195
+ "model.language_model.layers.15.self_attn.o_proj": {
196
+ "bits": 8
197
+ },
198
+ "model.language_model.layers.16.self_attn.k_proj": {
199
+ "bits": 8
200
+ },
201
+ "model.language_model.layers.16.self_attn.q_proj": {
202
+ "bits": 8
203
+ },
204
+ "model.language_model.layers.16.self_attn.v_proj": {
205
+ "bits": 8
206
+ },
207
+ "model.language_model.layers.16.self_attn.o_proj": {
208
+ "bits": 8
209
+ },
210
+ "model.language_model.layers.17.self_attn.k_proj": {
211
+ "bits": 8
212
+ },
213
+ "model.language_model.layers.17.self_attn.q_proj": {
214
+ "bits": 8
215
+ },
216
+ "model.language_model.layers.17.self_attn.o_proj": {
217
+ "bits": 8
218
+ },
219
+ "model.language_model.layers.18.self_attn.k_proj": {
220
+ "bits": 8
221
+ },
222
+ "model.language_model.layers.18.self_attn.q_proj": {
223
+ "bits": 8
224
+ },
225
+ "model.language_model.layers.18.self_attn.v_proj": {
226
+ "bits": 8
227
+ },
228
+ "model.language_model.layers.18.self_attn.o_proj": {
229
+ "bits": 8
230
+ },
231
+ "model.language_model.layers.19.self_attn.k_proj": {
232
+ "bits": 8
233
+ },
234
+ "model.language_model.layers.19.self_attn.q_proj": {
235
+ "bits": 8
236
+ },
237
+ "model.language_model.layers.19.self_attn.v_proj": {
238
+ "bits": 8
239
+ },
240
+ "model.language_model.layers.19.self_attn.o_proj": {
241
+ "bits": 8
242
+ },
243
+ "model.language_model.layers.20.self_attn.k_proj": {
244
+ "bits": 8
245
+ },
246
+ "model.language_model.layers.20.self_attn.q_proj": {
247
+ "bits": 8
248
+ },
249
+ "model.language_model.layers.20.self_attn.v_proj": {
250
+ "bits": 8
251
+ },
252
+ "model.language_model.layers.20.self_attn.o_proj": {
253
+ "bits": 8
254
+ },
255
+ "model.language_model.layers.21.self_attn.k_proj": {
256
+ "bits": 8
257
+ },
258
+ "model.language_model.layers.21.self_attn.q_proj": {
259
+ "bits": 8
260
+ },
261
+ "model.language_model.layers.21.self_attn.v_proj": {
262
+ "bits": 8
263
+ },
264
+ "model.language_model.layers.21.self_attn.o_proj": {
265
+ "bits": 8
266
+ },
267
+ "model.language_model.layers.22.self_attn.k_proj": {
268
+ "bits": 8
269
+ },
270
+ "model.language_model.layers.22.self_attn.q_proj": {
271
+ "bits": 8
272
+ },
273
+ "model.language_model.layers.22.self_attn.v_proj": {
274
+ "bits": 8
275
+ },
276
+ "model.language_model.layers.22.self_attn.o_proj": {
277
+ "bits": 8
278
+ },
279
+ "model.language_model.layers.23.self_attn.k_proj": {
280
+ "bits": 8
281
+ },
282
+ "model.language_model.layers.23.self_attn.q_proj": {
283
+ "bits": 8
284
+ },
285
+ "model.language_model.layers.23.self_attn.o_proj": {
286
+ "bits": 8
287
+ },
288
+ "model.language_model.layers.24.self_attn.k_proj": {
289
+ "bits": 8
290
+ },
291
+ "model.language_model.layers.24.self_attn.q_proj": {
292
+ "bits": 8
293
+ },
294
+ "model.language_model.layers.24.self_attn.v_proj": {
295
+ "bits": 8
296
+ },
297
+ "model.language_model.layers.24.self_attn.o_proj": {
298
+ "bits": 8
299
+ },
300
+ "model.language_model.layers.25.self_attn.k_proj": {
301
+ "bits": 8
302
+ },
303
+ "model.language_model.layers.25.self_attn.q_proj": {
304
+ "bits": 8
305
+ },
306
+ "model.language_model.layers.25.self_attn.v_proj": {
307
+ "bits": 8
308
+ },
309
+ "model.language_model.layers.25.self_attn.o_proj": {
310
+ "bits": 8
311
+ },
312
+ "model.language_model.layers.26.self_attn.k_proj": {
313
+ "bits": 8
314
+ },
315
+ "model.language_model.layers.26.self_attn.q_proj": {
316
+ "bits": 8
317
+ },
318
+ "model.language_model.layers.26.self_attn.v_proj": {
319
+ "bits": 8
320
+ },
321
+ "model.language_model.layers.26.self_attn.o_proj": {
322
+ "bits": 8
323
+ },
324
+ "model.language_model.layers.27.self_attn.k_proj": {
325
+ "bits": 8
326
+ },
327
+ "model.language_model.layers.27.self_attn.q_proj": {
328
+ "bits": 8
329
+ },
330
+ "model.language_model.layers.27.self_attn.v_proj": {
331
+ "bits": 8
332
+ },
333
+ "model.language_model.layers.27.self_attn.o_proj": {
334
+ "bits": 8
335
+ },
336
+ "model.language_model.layers.28.self_attn.k_proj": {
337
+ "bits": 8
338
+ },
339
+ "model.language_model.layers.28.self_attn.q_proj": {
340
+ "bits": 8
341
+ },
342
+ "model.language_model.layers.28.self_attn.v_proj": {
343
+ "bits": 8
344
+ },
345
+ "model.language_model.layers.28.self_attn.o_proj": {
346
+ "bits": 8
347
+ },
348
+ "model.language_model.layers.29.self_attn.k_proj": {
349
+ "bits": 8
350
+ },
351
+ "model.language_model.layers.29.self_attn.q_proj": {
352
+ "bits": 8
353
+ },
354
+ "model.language_model.layers.29.self_attn.o_proj": {
355
+ "bits": 8
356
+ },
357
+ "model.language_model.layers.0.mlp.gate_proj": {
358
+ "bits": 8
359
+ },
360
+ "model.language_model.layers.0.mlp.up_proj": {
361
+ "bits": 8
362
+ },
363
+ "model.language_model.layers.0.mlp.down_proj": {
364
+ "bits": 8
365
+ },
366
+ "model.language_model.layers.1.mlp.gate_proj": {
367
+ "bits": 8
368
+ },
369
+ "model.language_model.layers.1.mlp.up_proj": {
370
+ "bits": 8
371
+ },
372
+ "model.language_model.layers.1.mlp.down_proj": {
373
+ "bits": 8
374
+ },
375
+ "model.language_model.layers.2.mlp.gate_proj": {
376
+ "bits": 8
377
+ },
378
+ "model.language_model.layers.2.mlp.up_proj": {
379
+ "bits": 8
380
+ },
381
+ "model.language_model.layers.2.mlp.down_proj": {
382
+ "bits": 8
383
+ },
384
+ "model.language_model.layers.3.mlp.gate_proj": {
385
+ "bits": 8
386
+ },
387
+ "model.language_model.layers.3.mlp.up_proj": {
388
+ "bits": 8
389
+ },
390
+ "model.language_model.layers.3.mlp.down_proj": {
391
+ "bits": 8
392
+ },
393
+ "model.language_model.layers.4.mlp.gate_proj": {
394
+ "bits": 8
395
+ },
396
+ "model.language_model.layers.4.mlp.up_proj": {
397
+ "bits": 8
398
+ },
399
+ "model.language_model.layers.4.mlp.down_proj": {
400
+ "bits": 8
401
+ },
402
+ "model.language_model.layers.5.mlp.gate_proj": {
403
+ "bits": 8
404
+ },
405
+ "model.language_model.layers.5.mlp.up_proj": {
406
+ "bits": 8
407
+ },
408
+ "model.language_model.layers.5.mlp.down_proj": {
409
+ "bits": 8
410
+ },
411
+ "model.language_model.layers.6.mlp.gate_proj": {
412
+ "bits": 8
413
+ },
414
+ "model.language_model.layers.6.mlp.up_proj": {
415
+ "bits": 8
416
+ },
417
+ "model.language_model.layers.6.mlp.down_proj": {
418
+ "bits": 8
419
+ },
420
+ "model.language_model.layers.7.mlp.gate_proj": {
421
+ "bits": 8
422
+ },
423
+ "model.language_model.layers.7.mlp.up_proj": {
424
+ "bits": 8
425
+ },
426
+ "model.language_model.layers.7.mlp.down_proj": {
427
+ "bits": 8
428
+ },
429
+ "model.language_model.layers.8.mlp.gate_proj": {
430
+ "bits": 8
431
+ },
432
+ "model.language_model.layers.8.mlp.up_proj": {
433
+ "bits": 8
434
+ },
435
+ "model.language_model.layers.8.mlp.down_proj": {
436
+ "bits": 8
437
+ },
438
+ "model.language_model.layers.9.mlp.gate_proj": {
439
+ "bits": 8
440
+ },
441
+ "model.language_model.layers.9.mlp.up_proj": {
442
+ "bits": 8
443
+ },
444
+ "model.language_model.layers.9.mlp.down_proj": {
445
+ "bits": 8
446
+ },
447
+ "model.language_model.layers.10.mlp.gate_proj": {
448
+ "bits": 8
449
+ },
450
+ "model.language_model.layers.10.mlp.up_proj": {
451
+ "bits": 8
452
+ },
453
+ "model.language_model.layers.10.mlp.down_proj": {
454
+ "bits": 8
455
+ },
456
+ "model.language_model.layers.11.mlp.gate_proj": {
457
+ "bits": 8
458
+ },
459
+ "model.language_model.layers.11.mlp.up_proj": {
460
+ "bits": 8
461
+ },
462
+ "model.language_model.layers.11.mlp.down_proj": {
463
+ "bits": 8
464
+ },
465
+ "model.language_model.layers.12.mlp.gate_proj": {
466
+ "bits": 8
467
+ },
468
+ "model.language_model.layers.12.mlp.up_proj": {
469
+ "bits": 8
470
+ },
471
+ "model.language_model.layers.12.mlp.down_proj": {
472
+ "bits": 8
473
+ },
474
+ "model.language_model.layers.13.mlp.gate_proj": {
475
+ "bits": 8
476
+ },
477
+ "model.language_model.layers.13.mlp.up_proj": {
478
+ "bits": 8
479
+ },
480
+ "model.language_model.layers.13.mlp.down_proj": {
481
+ "bits": 8
482
+ },
483
+ "model.language_model.layers.14.mlp.gate_proj": {
484
+ "bits": 8
485
+ },
486
+ "model.language_model.layers.14.mlp.up_proj": {
487
+ "bits": 8
488
+ },
489
+ "model.language_model.layers.14.mlp.down_proj": {
490
+ "bits": 8
491
+ },
492
+ "model.language_model.layers.15.mlp.gate_proj": {
493
+ "bits": 8
494
+ },
495
+ "model.language_model.layers.15.mlp.up_proj": {
496
+ "bits": 8
497
+ },
498
+ "model.language_model.layers.15.mlp.down_proj": {
499
+ "bits": 8
500
+ },
501
+ "model.language_model.layers.16.mlp.gate_proj": {
502
+ "bits": 8
503
+ },
504
+ "model.language_model.layers.16.mlp.up_proj": {
505
+ "bits": 8
506
+ },
507
+ "model.language_model.layers.16.mlp.down_proj": {
508
+ "bits": 8
509
+ },
510
+ "model.language_model.layers.17.mlp.gate_proj": {
511
+ "bits": 8
512
+ },
513
+ "model.language_model.layers.17.mlp.up_proj": {
514
+ "bits": 8
515
+ },
516
+ "model.language_model.layers.17.mlp.down_proj": {
517
+ "bits": 8
518
+ },
519
+ "model.language_model.layers.18.mlp.gate_proj": {
520
+ "bits": 8
521
+ },
522
+ "model.language_model.layers.18.mlp.up_proj": {
523
+ "bits": 8
524
+ },
525
+ "model.language_model.layers.18.mlp.down_proj": {
526
+ "bits": 8
527
+ },
528
+ "model.language_model.layers.19.mlp.gate_proj": {
529
+ "bits": 8
530
+ },
531
+ "model.language_model.layers.19.mlp.up_proj": {
532
+ "bits": 8
533
+ },
534
+ "model.language_model.layers.19.mlp.down_proj": {
535
+ "bits": 8
536
+ },
537
+ "model.language_model.layers.20.mlp.gate_proj": {
538
+ "bits": 8
539
+ },
540
+ "model.language_model.layers.20.mlp.up_proj": {
541
+ "bits": 8
542
+ },
543
+ "model.language_model.layers.20.mlp.down_proj": {
544
+ "bits": 8
545
+ },
546
+ "model.language_model.layers.21.mlp.gate_proj": {
547
+ "bits": 8
548
+ },
549
+ "model.language_model.layers.21.mlp.up_proj": {
550
+ "bits": 8
551
+ },
552
+ "model.language_model.layers.21.mlp.down_proj": {
553
+ "bits": 8
554
+ },
555
+ "model.language_model.layers.22.mlp.gate_proj": {
556
+ "bits": 8
557
+ },
558
+ "model.language_model.layers.22.mlp.up_proj": {
559
+ "bits": 8
560
+ },
561
+ "model.language_model.layers.22.mlp.down_proj": {
562
+ "bits": 8
563
+ },
564
+ "model.language_model.layers.23.mlp.gate_proj": {
565
+ "bits": 8
566
+ },
567
+ "model.language_model.layers.23.mlp.up_proj": {
568
+ "bits": 8
569
+ },
570
+ "model.language_model.layers.23.mlp.down_proj": {
571
+ "bits": 8
572
+ },
573
+ "model.language_model.layers.24.mlp.gate_proj": {
574
+ "bits": 8
575
+ },
576
+ "model.language_model.layers.24.mlp.up_proj": {
577
+ "bits": 8
578
+ },
579
+ "model.language_model.layers.24.mlp.down_proj": {
580
+ "bits": 8
581
+ },
582
+ "model.language_model.layers.25.mlp.gate_proj": {
583
+ "bits": 8
584
+ },
585
+ "model.language_model.layers.25.mlp.up_proj": {
586
+ "bits": 8
587
+ },
588
+ "model.language_model.layers.25.mlp.down_proj": {
589
+ "bits": 8
590
+ },
591
+ "model.language_model.layers.26.mlp.gate_proj": {
592
+ "bits": 8
593
+ },
594
+ "model.language_model.layers.26.mlp.up_proj": {
595
+ "bits": 8
596
+ },
597
+ "model.language_model.layers.26.mlp.down_proj": {
598
+ "bits": 8
599
+ },
600
+ "model.language_model.layers.27.mlp.gate_proj": {
601
+ "bits": 8
602
+ },
603
+ "model.language_model.layers.27.mlp.up_proj": {
604
+ "bits": 8
605
+ },
606
+ "model.language_model.layers.27.mlp.down_proj": {
607
+ "bits": 8
608
+ },
609
+ "model.language_model.layers.28.mlp.gate_proj": {
610
+ "bits": 8
611
+ },
612
+ "model.language_model.layers.28.mlp.up_proj": {
613
+ "bits": 8
614
+ },
615
+ "model.language_model.layers.28.mlp.down_proj": {
616
+ "bits": 8
617
+ },
618
+ "model.language_model.layers.29.mlp.gate_proj": {
619
+ "bits": 8
620
+ },
621
+ "model.language_model.layers.29.mlp.up_proj": {
622
+ "bits": 8
623
+ },
624
+ "model.language_model.layers.29.mlp.down_proj": {
625
+ "bits": 8
626
+ },
627
+ "model.language_model.layers.0.router.proj": {
628
+ "bits": 8
629
+ },
630
+ "model.language_model.layers.1.router.proj": {
631
+ "bits": 8
632
+ },
633
+ "model.language_model.layers.2.router.proj": {
634
+ "bits": 8
635
+ },
636
+ "model.language_model.layers.3.router.proj": {
637
+ "bits": 8
638
+ },
639
+ "model.language_model.layers.4.router.proj": {
640
+ "bits": 8
641
+ },
642
+ "model.language_model.layers.5.router.proj": {
643
+ "bits": 8
644
+ },
645
+ "model.language_model.layers.6.router.proj": {
646
+ "bits": 8
647
+ },
648
+ "model.language_model.layers.7.router.proj": {
649
+ "bits": 8
650
+ },
651
+ "model.language_model.layers.8.router.proj": {
652
+ "bits": 8
653
+ },
654
+ "model.language_model.layers.9.router.proj": {
655
+ "bits": 8
656
+ },
657
+ "model.language_model.layers.10.router.proj": {
658
+ "bits": 8
659
+ },
660
+ "model.language_model.layers.11.router.proj": {
661
+ "bits": 8
662
+ },
663
+ "model.language_model.layers.12.router.proj": {
664
+ "bits": 8
665
+ },
666
+ "model.language_model.layers.13.router.proj": {
667
+ "bits": 8
668
+ },
669
+ "model.language_model.layers.14.router.proj": {
670
+ "bits": 8
671
+ },
672
+ "model.language_model.layers.15.router.proj": {
673
+ "bits": 8
674
+ },
675
+ "model.language_model.layers.16.router.proj": {
676
+ "bits": 8
677
+ },
678
+ "model.language_model.layers.17.router.proj": {
679
+ "bits": 8
680
+ },
681
+ "model.language_model.layers.18.router.proj": {
682
+ "bits": 8
683
+ },
684
+ "model.language_model.layers.19.router.proj": {
685
+ "bits": 8
686
+ },
687
+ "model.language_model.layers.20.router.proj": {
688
+ "bits": 8
689
+ },
690
+ "model.language_model.layers.21.router.proj": {
691
+ "bits": 8
692
+ },
693
+ "model.language_model.layers.22.router.proj": {
694
+ "bits": 8
695
+ },
696
+ "model.language_model.layers.23.router.proj": {
697
+ "bits": 8
698
+ },
699
+ "model.language_model.layers.24.router.proj": {
700
+ "bits": 8
701
+ },
702
+ "model.language_model.layers.25.router.proj": {
703
+ "bits": 8
704
+ },
705
+ "model.language_model.layers.26.router.proj": {
706
+ "bits": 8
707
+ },
708
+ "model.language_model.layers.27.router.proj": {
709
+ "bits": 8
710
+ },
711
+ "model.language_model.layers.28.router.proj": {
712
+ "bits": 8
713
+ },
714
+ "model.language_model.layers.29.router.proj": {
715
+ "bits": 8
716
+ },
717
+ ".*model\\.language_model\\.layers\\.\\d+\\.self_attn\\..*": {
718
+ "bits": 8
719
+ },
720
+ ".*model\\.language_model\\.layers\\.\\d+\\.mlp\\..*": {
721
+ "bits": 8
722
+ },
723
+ ".*model\\.language_model\\.layers\\.\\d+\\.router\\..*": {
724
+ "bits": 8
725
+ }
726
+ }
727
+ }
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": "<eos>",
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": 1000000000000000019884624838656,
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": "left",
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<content>(?:(?!\\<\\|tool_call\\>)(?!\\<turn\\|\\>).)+)?(?P<tool_calls>\\<\\|tool_call\\>.*\\<tool_call\\|\\>)?(?:\\<turn\\|\\>)?"
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
+ }