Hugging Face's logo

ig1
/
medgemma-27b-it-FP8-Dynamic

Text Generation
Transformers
Safetensors
gemma3
any-to-any
medical
biomedical
multimodal
fp8
quantization
vllm
medgemma
conversational
text-generation-inference
compressed-tensors
Model card Files
xet
Community
ihebig1 commited on
Commit
09051b5
·
verified ·
1 Parent(s): ff950db

Upload folder using huggingface_hub

Browse files
Files changed (20) hide show
  1. .gitattributes +1 -0
  2. added_tokens.json +3 -0
  3. chat_template.jinja +47 -0
  4. config.json +334 -0
  5. generation_config.json +10 -0
  6. model-00001-of-00007.safetensors +3 -0
  7. model-00002-of-00007.safetensors +3 -0
  8. model-00003-of-00007.safetensors +3 -0
  9. model-00004-of-00007.safetensors +3 -0
  10. model-00005-of-00007.safetensors +3 -0
  11. model-00006-of-00007.safetensors +3 -0
  12. model-00007-of-00007.safetensors +3 -0
  13. model.safetensors.index.json +0 -0
  14. preprocessor_config.json +29 -0
  15. processor_config.json +4 -0
  16. recipe.yaml +7 -0
  17. special_tokens_map.json +33 -0
  18. tokenizer.json +3 -0
  19. tokenizer.model +3 -0
  20. tokenizer_config.json +0 -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
added_tokens.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ {
2
+ "<image_soft_token>": 262144
3
+ }
chat_template.jinja ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {{ bos_token }}
2
+ {%- if messages[0]['role'] == 'system' -%}
3
+ {%- if messages[0]['content'] is string -%}
4
+ {%- set first_user_prefix = messages[0]['content'] + '
5
+
6
+ ' -%}
7
+ {%- else -%}
8
+ {%- set first_user_prefix = messages[0]['content'][0]['text'] + '
9
+
10
+ ' -%}
11
+ {%- endif -%}
12
+ {%- set loop_messages = messages[1:] -%}
13
+ {%- else -%}
14
+ {%- set first_user_prefix = "" -%}
15
+ {%- set loop_messages = messages -%}
16
+ {%- endif -%}
17
+ {%- for message in loop_messages -%}
18
+ {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
19
+ {{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
20
+ {%- endif -%}
21
+ {%- if (message['role'] == 'assistant') -%}
22
+ {%- set role = "model" -%}
23
+ {%- else -%}
24
+ {%- set role = message['role'] -%}
25
+ {%- endif -%}
26
+ {{ '<start_of_turn>' + role + '
27
+ ' + (first_user_prefix if loop.first else "") }}
28
+ {%- if message['content'] is string -%}
29
+ {{ message['content'] | trim }}
30
+ {%- elif message['content'] is iterable -%}
31
+ {%- for item in message['content'] -%}
32
+ {%- if item['type'] == 'image' -%}
33
+ {{ '<start_of_image>' }}
34
+ {%- elif item['type'] == 'text' -%}
35
+ {{ item['text'] | trim }}
36
+ {%- endif -%}
37
+ {%- endfor -%}
38
+ {%- else -%}
39
+ {{ raise_exception("Invalid content type") }}
40
+ {%- endif -%}
41
+ {{ '<end_of_turn>
42
+ ' }}
43
+ {%- endfor -%}
44
+ {%- if add_generation_prompt -%}
45
+ {{'<start_of_turn>model
46
+ '}}
47
+ {%- endif -%}
config.json ADDED
@@ -0,0 +1,334 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Gemma3ForConditionalGeneration"
4
+ ],
5
+ "boi_token_index": 255999,
6
+ "dtype": "bfloat16",
7
+ "eoi_token_index": 256000,
8
+ "eos_token_id": [
9
+ 1,
10
+ 106
11
+ ],
12
+ "image_token_index": 262144,
13
+ "initializer_range": 0.02,
14
+ "mm_tokens_per_image": 256,
15
+ "model_type": "gemma3",
16
+ "quantization_config": {
17
+ "config_groups": {
18
+ "group_0": {
19
+ "format": "float-quantized",
20
+ "input_activations": {
21
+ "actorder": null,
22
+ "block_structure": null,
23
+ "dynamic": true,
24
+ "group_size": null,
25
+ "num_bits": 8,
26
+ "observer": null,
27
+ "observer_kwargs": {},
28
+ "strategy": "token",
29
+ "symmetric": true,
30
+ "type": "float"
31
+ },
32
+ "output_activations": null,
33
+ "targets": [
34
+ "Linear"
35
+ ],
36
+ "weights": {
37
+ "actorder": null,
38
+ "block_structure": null,
39
+ "dynamic": false,
40
+ "group_size": null,
41
+ "num_bits": 8,
42
+ "observer": "minmax",
43
+ "observer_kwargs": {},
44
+ "strategy": "channel",
45
+ "symmetric": true,
46
+ "type": "float"
47
+ }
48
+ }
49
+ },
50
+ "format": "float-quantized",
51
+ "global_compression_ratio": null,
52
+ "ignore": [
53
+ "model.vision_tower.vision_model.encoder.layers.0.self_attn.k_proj",
54
+ "model.vision_tower.vision_model.encoder.layers.0.self_attn.v_proj",
55
+ "model.vision_tower.vision_model.encoder.layers.0.self_attn.q_proj",
56
+ "model.vision_tower.vision_model.encoder.layers.0.self_attn.out_proj",
57
+ "model.vision_tower.vision_model.encoder.layers.0.mlp.fc1",
58
+ "model.vision_tower.vision_model.encoder.layers.0.mlp.fc2",
59
+ "model.vision_tower.vision_model.encoder.layers.1.self_attn.k_proj",
60
+ "model.vision_tower.vision_model.encoder.layers.1.self_attn.v_proj",
61
+ "model.vision_tower.vision_model.encoder.layers.1.self_attn.q_proj",
62
+ "model.vision_tower.vision_model.encoder.layers.1.self_attn.out_proj",
63
+ "model.vision_tower.vision_model.encoder.layers.1.mlp.fc1",
64
+ "model.vision_tower.vision_model.encoder.layers.1.mlp.fc2",
65
+ "model.vision_tower.vision_model.encoder.layers.2.self_attn.k_proj",
66
+ "model.vision_tower.vision_model.encoder.layers.2.self_attn.v_proj",
67
+ "model.vision_tower.vision_model.encoder.layers.2.self_attn.q_proj",
68
+ "model.vision_tower.vision_model.encoder.layers.2.self_attn.out_proj",
69
+ "model.vision_tower.vision_model.encoder.layers.2.mlp.fc1",
70
+ "model.vision_tower.vision_model.encoder.layers.2.mlp.fc2",
71
+ "model.vision_tower.vision_model.encoder.layers.3.self_attn.k_proj",
72
+ "model.vision_tower.vision_model.encoder.layers.3.self_attn.v_proj",
73
+ "model.vision_tower.vision_model.encoder.layers.3.self_attn.q_proj",
74
+ "model.vision_tower.vision_model.encoder.layers.3.self_attn.out_proj",
75
+ "model.vision_tower.vision_model.encoder.layers.3.mlp.fc1",
76
+ "model.vision_tower.vision_model.encoder.layers.3.mlp.fc2",
77
+ "model.vision_tower.vision_model.encoder.layers.4.self_attn.k_proj",
78
+ "model.vision_tower.vision_model.encoder.layers.4.self_attn.v_proj",
79
+ "model.vision_tower.vision_model.encoder.layers.4.self_attn.q_proj",
80
+ "model.vision_tower.vision_model.encoder.layers.4.self_attn.out_proj",
81
+ "model.vision_tower.vision_model.encoder.layers.4.mlp.fc1",
82
+ "model.vision_tower.vision_model.encoder.layers.4.mlp.fc2",
83
+ "model.vision_tower.vision_model.encoder.layers.5.self_attn.k_proj",
84
+ "model.vision_tower.vision_model.encoder.layers.5.self_attn.v_proj",
85
+ "model.vision_tower.vision_model.encoder.layers.5.self_attn.q_proj",
86
+ "model.vision_tower.vision_model.encoder.layers.5.self_attn.out_proj",
87
+ "model.vision_tower.vision_model.encoder.layers.5.mlp.fc1",
88
+ "model.vision_tower.vision_model.encoder.layers.5.mlp.fc2",
89
+ "model.vision_tower.vision_model.encoder.layers.6.self_attn.k_proj",
90
+ "model.vision_tower.vision_model.encoder.layers.6.self_attn.v_proj",
91
+ "model.vision_tower.vision_model.encoder.layers.6.self_attn.q_proj",
92
+ "model.vision_tower.vision_model.encoder.layers.6.self_attn.out_proj",
93
+ "model.vision_tower.vision_model.encoder.layers.6.mlp.fc1",
94
+ "model.vision_tower.vision_model.encoder.layers.6.mlp.fc2",
95
+ "model.vision_tower.vision_model.encoder.layers.7.self_attn.k_proj",
96
+ "model.vision_tower.vision_model.encoder.layers.7.self_attn.v_proj",
97
+ "model.vision_tower.vision_model.encoder.layers.7.self_attn.q_proj",
98
+ "model.vision_tower.vision_model.encoder.layers.7.self_attn.out_proj",
99
+ "model.vision_tower.vision_model.encoder.layers.7.mlp.fc1",
100
+ "model.vision_tower.vision_model.encoder.layers.7.mlp.fc2",
101
+ "model.vision_tower.vision_model.encoder.layers.8.self_attn.k_proj",
102
+ "model.vision_tower.vision_model.encoder.layers.8.self_attn.v_proj",
103
+ "model.vision_tower.vision_model.encoder.layers.8.self_attn.q_proj",
104
+ "model.vision_tower.vision_model.encoder.layers.8.self_attn.out_proj",
105
+ "model.vision_tower.vision_model.encoder.layers.8.mlp.fc1",
106
+ "model.vision_tower.vision_model.encoder.layers.8.mlp.fc2",
107
+ "model.vision_tower.vision_model.encoder.layers.9.self_attn.k_proj",
108
+ "model.vision_tower.vision_model.encoder.layers.9.self_attn.v_proj",
109
+ "model.vision_tower.vision_model.encoder.layers.9.self_attn.q_proj",
110
+ "model.vision_tower.vision_model.encoder.layers.9.self_attn.out_proj",
111
+ "model.vision_tower.vision_model.encoder.layers.9.mlp.fc1",
112
+ "model.vision_tower.vision_model.encoder.layers.9.mlp.fc2",
113
+ "model.vision_tower.vision_model.encoder.layers.10.self_attn.k_proj",
114
+ "model.vision_tower.vision_model.encoder.layers.10.self_attn.v_proj",
115
+ "model.vision_tower.vision_model.encoder.layers.10.self_attn.q_proj",
116
+ "model.vision_tower.vision_model.encoder.layers.10.self_attn.out_proj",
117
+ "model.vision_tower.vision_model.encoder.layers.10.mlp.fc1",
118
+ "model.vision_tower.vision_model.encoder.layers.10.mlp.fc2",
119
+ "model.vision_tower.vision_model.encoder.layers.11.self_attn.k_proj",
120
+ "model.vision_tower.vision_model.encoder.layers.11.self_attn.v_proj",
121
+ "model.vision_tower.vision_model.encoder.layers.11.self_attn.q_proj",
122
+ "model.vision_tower.vision_model.encoder.layers.11.self_attn.out_proj",
123
+ "model.vision_tower.vision_model.encoder.layers.11.mlp.fc1",
124
+ "model.vision_tower.vision_model.encoder.layers.11.mlp.fc2",
125
+ "model.vision_tower.vision_model.encoder.layers.12.self_attn.k_proj",
126
+ "model.vision_tower.vision_model.encoder.layers.12.self_attn.v_proj",
127
+ "model.vision_tower.vision_model.encoder.layers.12.self_attn.q_proj",
128
+ "model.vision_tower.vision_model.encoder.layers.12.self_attn.out_proj",
129
+ "model.vision_tower.vision_model.encoder.layers.12.mlp.fc1",
130
+ "model.vision_tower.vision_model.encoder.layers.12.mlp.fc2",
131
+ "model.vision_tower.vision_model.encoder.layers.13.self_attn.k_proj",
132
+ "model.vision_tower.vision_model.encoder.layers.13.self_attn.v_proj",
133
+ "model.vision_tower.vision_model.encoder.layers.13.self_attn.q_proj",
134
+ "model.vision_tower.vision_model.encoder.layers.13.self_attn.out_proj",
135
+ "model.vision_tower.vision_model.encoder.layers.13.mlp.fc1",
136
+ "model.vision_tower.vision_model.encoder.layers.13.mlp.fc2",
137
+ "model.vision_tower.vision_model.encoder.layers.14.self_attn.k_proj",
138
+ "model.vision_tower.vision_model.encoder.layers.14.self_attn.v_proj",
139
+ "model.vision_tower.vision_model.encoder.layers.14.self_attn.q_proj",
140
+ "model.vision_tower.vision_model.encoder.layers.14.self_attn.out_proj",
141
+ "model.vision_tower.vision_model.encoder.layers.14.mlp.fc1",
142
+ "model.vision_tower.vision_model.encoder.layers.14.mlp.fc2",
143
+ "model.vision_tower.vision_model.encoder.layers.15.self_attn.k_proj",
144
+ "model.vision_tower.vision_model.encoder.layers.15.self_attn.v_proj",
145
+ "model.vision_tower.vision_model.encoder.layers.15.self_attn.q_proj",
146
+ "model.vision_tower.vision_model.encoder.layers.15.self_attn.out_proj",
147
+ "model.vision_tower.vision_model.encoder.layers.15.mlp.fc1",
148
+ "model.vision_tower.vision_model.encoder.layers.15.mlp.fc2",
149
+ "model.vision_tower.vision_model.encoder.layers.16.self_attn.k_proj",
150
+ "model.vision_tower.vision_model.encoder.layers.16.self_attn.v_proj",
151
+ "model.vision_tower.vision_model.encoder.layers.16.self_attn.q_proj",
152
+ "model.vision_tower.vision_model.encoder.layers.16.self_attn.out_proj",
153
+ "model.vision_tower.vision_model.encoder.layers.16.mlp.fc1",
154
+ "model.vision_tower.vision_model.encoder.layers.16.mlp.fc2",
155
+ "model.vision_tower.vision_model.encoder.layers.17.self_attn.k_proj",
156
+ "model.vision_tower.vision_model.encoder.layers.17.self_attn.v_proj",
157
+ "model.vision_tower.vision_model.encoder.layers.17.self_attn.q_proj",
158
+ "model.vision_tower.vision_model.encoder.layers.17.self_attn.out_proj",
159
+ "model.vision_tower.vision_model.encoder.layers.17.mlp.fc1",
160
+ "model.vision_tower.vision_model.encoder.layers.17.mlp.fc2",
161
+ "model.vision_tower.vision_model.encoder.layers.18.self_attn.k_proj",
162
+ "model.vision_tower.vision_model.encoder.layers.18.self_attn.v_proj",
163
+ "model.vision_tower.vision_model.encoder.layers.18.self_attn.q_proj",
164
+ "model.vision_tower.vision_model.encoder.layers.18.self_attn.out_proj",
165
+ "model.vision_tower.vision_model.encoder.layers.18.mlp.fc1",
166
+ "model.vision_tower.vision_model.encoder.layers.18.mlp.fc2",
167
+ "model.vision_tower.vision_model.encoder.layers.19.self_attn.k_proj",
168
+ "model.vision_tower.vision_model.encoder.layers.19.self_attn.v_proj",
169
+ "model.vision_tower.vision_model.encoder.layers.19.self_attn.q_proj",
170
+ "model.vision_tower.vision_model.encoder.layers.19.self_attn.out_proj",
171
+ "model.vision_tower.vision_model.encoder.layers.19.mlp.fc1",
172
+ "model.vision_tower.vision_model.encoder.layers.19.mlp.fc2",
173
+ "model.vision_tower.vision_model.encoder.layers.20.self_attn.k_proj",
174
+ "model.vision_tower.vision_model.encoder.layers.20.self_attn.v_proj",
175
+ "model.vision_tower.vision_model.encoder.layers.20.self_attn.q_proj",
176
+ "model.vision_tower.vision_model.encoder.layers.20.self_attn.out_proj",
177
+ "model.vision_tower.vision_model.encoder.layers.20.mlp.fc1",
178
+ "model.vision_tower.vision_model.encoder.layers.20.mlp.fc2",
179
+ "model.vision_tower.vision_model.encoder.layers.21.self_attn.k_proj",
180
+ "model.vision_tower.vision_model.encoder.layers.21.self_attn.v_proj",
181
+ "model.vision_tower.vision_model.encoder.layers.21.self_attn.q_proj",
182
+ "model.vision_tower.vision_model.encoder.layers.21.self_attn.out_proj",
183
+ "model.vision_tower.vision_model.encoder.layers.21.mlp.fc1",
184
+ "model.vision_tower.vision_model.encoder.layers.21.mlp.fc2",
185
+ "model.vision_tower.vision_model.encoder.layers.22.self_attn.k_proj",
186
+ "model.vision_tower.vision_model.encoder.layers.22.self_attn.v_proj",
187
+ "model.vision_tower.vision_model.encoder.layers.22.self_attn.q_proj",
188
+ "model.vision_tower.vision_model.encoder.layers.22.self_attn.out_proj",
189
+ "model.vision_tower.vision_model.encoder.layers.22.mlp.fc1",
190
+ "model.vision_tower.vision_model.encoder.layers.22.mlp.fc2",
191
+ "model.vision_tower.vision_model.encoder.layers.23.self_attn.k_proj",
192
+ "model.vision_tower.vision_model.encoder.layers.23.self_attn.v_proj",
193
+ "model.vision_tower.vision_model.encoder.layers.23.self_attn.q_proj",
194
+ "model.vision_tower.vision_model.encoder.layers.23.self_attn.out_proj",
195
+ "model.vision_tower.vision_model.encoder.layers.23.mlp.fc1",
196
+ "model.vision_tower.vision_model.encoder.layers.23.mlp.fc2",
197
+ "model.vision_tower.vision_model.encoder.layers.24.self_attn.k_proj",
198
+ "model.vision_tower.vision_model.encoder.layers.24.self_attn.v_proj",
199
+ "model.vision_tower.vision_model.encoder.layers.24.self_attn.q_proj",
200
+ "model.vision_tower.vision_model.encoder.layers.24.self_attn.out_proj",
201
+ "model.vision_tower.vision_model.encoder.layers.24.mlp.fc1",
202
+ "model.vision_tower.vision_model.encoder.layers.24.mlp.fc2",
203
+ "model.vision_tower.vision_model.encoder.layers.25.self_attn.k_proj",
204
+ "model.vision_tower.vision_model.encoder.layers.25.self_attn.v_proj",
205
+ "model.vision_tower.vision_model.encoder.layers.25.self_attn.q_proj",
206
+ "model.vision_tower.vision_model.encoder.layers.25.self_attn.out_proj",
207
+ "model.vision_tower.vision_model.encoder.layers.25.mlp.fc1",
208
+ "model.vision_tower.vision_model.encoder.layers.25.mlp.fc2",
209
+ "model.vision_tower.vision_model.encoder.layers.26.self_attn.k_proj",
210
+ "model.vision_tower.vision_model.encoder.layers.26.self_attn.v_proj",
211
+ "model.vision_tower.vision_model.encoder.layers.26.self_attn.q_proj",
212
+ "model.vision_tower.vision_model.encoder.layers.26.self_attn.out_proj",
213
+ "model.vision_tower.vision_model.encoder.layers.26.mlp.fc1",
214
+ "model.vision_tower.vision_model.encoder.layers.26.mlp.fc2",
215
+ "lm_head"
216
+ ],
217
+ "kv_cache_scheme": null,
218
+ "quant_method": "compressed-tensors",
219
+ "quantization_status": "compressed",
220
+ "sparsity_config": {},
221
+ "transform_config": {},
222
+ "version": "0.12.2"
223
+ },
224
+ "text_config": {
225
+ "_sliding_window_pattern": 6,
226
+ "attention_bias": false,
227
+ "attention_dropout": 0.0,
228
+ "attn_logit_softcapping": null,
229
+ "dtype": "bfloat16",
230
+ "final_logit_softcapping": null,
231
+ "head_dim": 128,
232
+ "hidden_activation": "gelu_pytorch_tanh",
233
+ "hidden_size": 5376,
234
+ "initializer_range": 0.02,
235
+ "intermediate_size": 21504,
236
+ "layer_types": [
237
+ "sliding_attention",
238
+ "sliding_attention",
239
+ "sliding_attention",
240
+ "sliding_attention",
241
+ "sliding_attention",
242
+ "full_attention",
243
+ "sliding_attention",
244
+ "sliding_attention",
245
+ "sliding_attention",
246
+ "sliding_attention",
247
+ "sliding_attention",
248
+ "full_attention",
249
+ "sliding_attention",
250
+ "sliding_attention",
251
+ "sliding_attention",
252
+ "sliding_attention",
253
+ "sliding_attention",
254
+ "full_attention",
255
+ "sliding_attention",
256
+ "sliding_attention",
257
+ "sliding_attention",
258
+ "sliding_attention",
259
+ "sliding_attention",
260
+ "full_attention",
261
+ "sliding_attention",
262
+ "sliding_attention",
263
+ "sliding_attention",
264
+ "sliding_attention",
265
+ "sliding_attention",
266
+ "full_attention",
267
+ "sliding_attention",
268
+ "sliding_attention",
269
+ "sliding_attention",
270
+ "sliding_attention",
271
+ "sliding_attention",
272
+ "full_attention",
273
+ "sliding_attention",
274
+ "sliding_attention",
275
+ "sliding_attention",
276
+ "sliding_attention",
277
+ "sliding_attention",
278
+ "full_attention",
279
+ "sliding_attention",
280
+ "sliding_attention",
281
+ "sliding_attention",
282
+ "sliding_attention",
283
+ "sliding_attention",
284
+ "full_attention",
285
+ "sliding_attention",
286
+ "sliding_attention",
287
+ "sliding_attention",
288
+ "sliding_attention",
289
+ "sliding_attention",
290
+ "full_attention",
291
+ "sliding_attention",
292
+ "sliding_attention",
293
+ "sliding_attention",
294
+ "sliding_attention",
295
+ "sliding_attention",
296
+ "full_attention",
297
+ "sliding_attention",
298
+ "sliding_attention"
299
+ ],
300
+ "max_position_embeddings": 131072,
301
+ "model_type": "gemma3_text",
302
+ "num_attention_heads": 32,
303
+ "num_hidden_layers": 62,
304
+ "num_key_value_heads": 16,
305
+ "query_pre_attn_scalar": 168,
306
+ "rms_norm_eps": 1e-06,
307
+ "rope_local_base_freq": 10000,
308
+ "rope_scaling": {
309
+ "factor": 8.0,
310
+ "rope_type": "linear"
311
+ },
312
+ "rope_theta": 1000000,
313
+ "sliding_window": 1024,
314
+ "use_cache": true,
315
+ "vocab_size": 262208
316
+ },
317
+ "tie_word_embeddings": false,
318
+ "transformers_version": "4.56.2",
319
+ "vision_config": {
320
+ "attention_dropout": 0.0,
321
+ "dtype": "bfloat16",
322
+ "hidden_act": "gelu_pytorch_tanh",
323
+ "hidden_size": 1152,
324
+ "image_size": 896,
325
+ "intermediate_size": 4304,
326
+ "layer_norm_eps": 1e-06,
327
+ "model_type": "siglip_vision_model",
328
+ "num_attention_heads": 16,
329
+ "num_channels": 3,
330
+ "num_hidden_layers": 27,
331
+ "patch_size": 14,
332
+ "vision_use_head": false
333
+ }
334
+ }
generation_config.json ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 2,
4
+ "eos_token_id": [
5
+ 1,
6
+ 106
7
+ ],
8
+ "pad_token_id": 0,
9
+ "transformers_version": "4.56.2"
10
+ }
model-00001-of-00007.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:92a360f6a192bb43f68ac7a3193dc8b480097e1fa851e617a07ee449b55617cb
3
+ size 4970671784
model-00002-of-00007.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:181c49d6d33660daa0915270f18fd0f95720046816e31a729a8f0c5cb57e630b
3
+ size 4956561616
model-00003-of-00007.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f9ccc2e0670e6adf652c4eb6176835bab3043b002814867c57a97a6b70424811
3
+ size 4956561760
model-00004-of-00007.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6ed7960b50129caeb7e10d45953f02f0b63a8a5743a9b01db480c45edd0a0cdb
3
+ size 4956561760
model-00005-of-00007.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e93cea95903f907ec6298a751cd254076f78cd865b2e22f3324737100e429e9e
3
+ size 4956561760
model-00006-of-00007.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c7bf41322a921cd219621aa1b44d75cec8b3d2f9d4974e3207d2edc37d79ec2f
3
+ size 4477436544
model-00007-of-00007.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:05312f74cf750bd4b32af28802624fef104582e65e350ea502a9a29f23c5dc4a
3
+ size 2819260560
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
preprocessor_config.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_convert_rgb": null,
3
+ "do_normalize": true,
4
+ "do_pan_and_scan": null,
5
+ "do_rescale": true,
6
+ "do_resize": true,
7
+ "image_mean": [
8
+ 0.5,
9
+ 0.5,
10
+ 0.5
11
+ ],
12
+ "image_processor_type": "Gemma3ImageProcessor",
13
+ "image_seq_length": 256,
14
+ "image_std": [
15
+ 0.5,
16
+ 0.5,
17
+ 0.5
18
+ ],
19
+ "pan_and_scan_max_num_crops": null,
20
+ "pan_and_scan_min_crop_size": null,
21
+ "pan_and_scan_min_ratio_to_activate": null,
22
+ "processor_class": "Gemma3Processor",
23
+ "resample": 2,
24
+ "rescale_factor": 0.00392156862745098,
25
+ "size": {
26
+ "height": 896,
27
+ "width": 896
28
+ }
29
+ }
processor_config.json ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {
2
+ "image_seq_length": 256,
3
+ "processor_class": "Gemma3Processor"
4
+ }
recipe.yaml ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ default_stage:
2
+ default_modifiers:
3
+ QuantizationModifier:
4
+ targets: [Linear]
5
+ ignore: [lm_head, 're:visual.*', 're:model\.visual.*', 're:vision.*', 're:.*vision_tower.*',
6
+ 're:.*multi_modal_projector.*', 're:.*vision_model.*']
7
+ scheme: FP8_DYNAMIC
special_tokens_map.json ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "boi_token": "<start_of_image>",
3
+ "bos_token": {
4
+ "content": "<bos>",
5
+ "lstrip": false,
6
+ "normalized": false,
7
+ "rstrip": false,
8
+ "single_word": false
9
+ },
10
+ "eoi_token": "<end_of_image>",
11
+ "eos_token": {
12
+ "content": "<eos>",
13
+ "lstrip": false,
14
+ "normalized": false,
15
+ "rstrip": false,
16
+ "single_word": false
17
+ },
18
+ "image_token": "<image_soft_token>",
19
+ "pad_token": {
20
+ "content": "<pad>",
21
+ "lstrip": false,
22
+ "normalized": false,
23
+ "rstrip": false,
24
+ "single_word": false
25
+ },
26
+ "unk_token": {
27
+ "content": "<unk>",
28
+ "lstrip": false,
29
+ "normalized": false,
30
+ "rstrip": false,
31
+ "single_word": false
32
+ }
33
+ }
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4667f2089529e8e7657cfb6d1c19910ae71ff5f28aa7ab2ff2763330affad795
3
+ size 33384568
tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1299c11d7cf632ef3b4e11937501358ada021bbdf7c47638d13c0ee982f2e79c
3
+ size 4689074
tokenizer_config.json ADDED
The diff for this file is too large to render. See raw diff