Hugging Face's logo

inference-net
/
ClipTagger-12b

Image-Text-to-Text
Safetensors
English
gemma3
VLM
video-understanding
image-captioning
gemma
json-mode
structured-output
video-analysis
conversational
Eval Results
compressed-tensors
Model card Files
xet
Community
amarinference commited on
Commit
6733c82
·
verified ·
1 Parent(s): 4f5e1c2

Upload folder using huggingface_hub

Browse files
Files changed (16) hide show
  1. .gitattributes +1 -0
  2. added_tokens.json +3 -0
  3. chat_template.jinja +47 -0
  4. config.json +317 -0
  5. generation_config.json +13 -0
  6. model-00001-of-00003.safetensors +3 -0
  7. model-00002-of-00003.safetensors +3 -0
  8. model-00003-of-00003.safetensors +3 -0
  9. model.safetensors.index.json +0 -0
  10. preprocessor_config.json +37 -0
  11. processor_config.json +4 -0
  12. recipe.yaml +6 -0
  13. special_tokens_map.json +33 -0
  14. tokenizer.json +3 -0
  15. tokenizer.model +3 -0
  16. 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 +1,317 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Gemma3ForConditionalGeneration"
4
+ ],
5
+ "boi_token_index": 255999,
6
+ "eoi_token_index": 256000,
7
+ "eos_token_id": [
8
+ 1,
9
+ 106
10
+ ],
11
+ "image_token_index": 262144,
12
+ "initializer_range": 0.02,
13
+ "mm_tokens_per_image": 256,
14
+ "model_type": "gemma3",
15
+ "quantization_config": {
16
+ "config_groups": {
17
+ "group_0": {
18
+ "input_activations": {
19
+ "actorder": null,
20
+ "block_structure": null,
21
+ "dynamic": true,
22
+ "group_size": null,
23
+ "num_bits": 8,
24
+ "observer": null,
25
+ "observer_kwargs": {},
26
+ "strategy": "token",
27
+ "symmetric": true,
28
+ "type": "float"
29
+ },
30
+ "output_activations": null,
31
+ "targets": [
32
+ "Linear"
33
+ ],
34
+ "weights": {
35
+ "actorder": null,
36
+ "block_structure": null,
37
+ "dynamic": false,
38
+ "group_size": null,
39
+ "num_bits": 8,
40
+ "observer": "minmax",
41
+ "observer_kwargs": {},
42
+ "strategy": "channel",
43
+ "symmetric": true,
44
+ "type": "float"
45
+ }
46
+ }
47
+ },
48
+ "format": "float-quantized",
49
+ "global_compression_ratio": null,
50
+ "ignore": [
51
+ "model.vision_tower.vision_model.encoder.layers.0.self_attn.k_proj",
52
+ "model.vision_tower.vision_model.encoder.layers.0.self_attn.v_proj",
53
+ "model.vision_tower.vision_model.encoder.layers.0.self_attn.q_proj",
54
+ "model.vision_tower.vision_model.encoder.layers.0.self_attn.out_proj",
55
+ "model.vision_tower.vision_model.encoder.layers.0.mlp.fc1",
56
+ "model.vision_tower.vision_model.encoder.layers.0.mlp.fc2",
57
+ "model.vision_tower.vision_model.encoder.layers.1.self_attn.k_proj",
58
+ "model.vision_tower.vision_model.encoder.layers.1.self_attn.v_proj",
59
+ "model.vision_tower.vision_model.encoder.layers.1.self_attn.q_proj",
60
+ "model.vision_tower.vision_model.encoder.layers.1.self_attn.out_proj",
61
+ "model.vision_tower.vision_model.encoder.layers.1.mlp.fc1",
62
+ "model.vision_tower.vision_model.encoder.layers.1.mlp.fc2",
63
+ "model.vision_tower.vision_model.encoder.layers.2.self_attn.k_proj",
64
+ "model.vision_tower.vision_model.encoder.layers.2.self_attn.v_proj",
65
+ "model.vision_tower.vision_model.encoder.layers.2.self_attn.q_proj",
66
+ "model.vision_tower.vision_model.encoder.layers.2.self_attn.out_proj",
67
+ "model.vision_tower.vision_model.encoder.layers.2.mlp.fc1",
68
+ "model.vision_tower.vision_model.encoder.layers.2.mlp.fc2",
69
+ "model.vision_tower.vision_model.encoder.layers.3.self_attn.k_proj",
70
+ "model.vision_tower.vision_model.encoder.layers.3.self_attn.v_proj",
71
+ "model.vision_tower.vision_model.encoder.layers.3.self_attn.q_proj",
72
+ "model.vision_tower.vision_model.encoder.layers.3.self_attn.out_proj",
73
+ "model.vision_tower.vision_model.encoder.layers.3.mlp.fc1",
74
+ "model.vision_tower.vision_model.encoder.layers.3.mlp.fc2",
75
+ "model.vision_tower.vision_model.encoder.layers.4.self_attn.k_proj",
76
+ "model.vision_tower.vision_model.encoder.layers.4.self_attn.v_proj",
77
+ "model.vision_tower.vision_model.encoder.layers.4.self_attn.q_proj",
78
+ "model.vision_tower.vision_model.encoder.layers.4.self_attn.out_proj",
79
+ "model.vision_tower.vision_model.encoder.layers.4.mlp.fc1",
80
+ "model.vision_tower.vision_model.encoder.layers.4.mlp.fc2",
81
+ "model.vision_tower.vision_model.encoder.layers.5.self_attn.k_proj",
82
+ "model.vision_tower.vision_model.encoder.layers.5.self_attn.v_proj",
83
+ "model.vision_tower.vision_model.encoder.layers.5.self_attn.q_proj",
84
+ "model.vision_tower.vision_model.encoder.layers.5.self_attn.out_proj",
85
+ "model.vision_tower.vision_model.encoder.layers.5.mlp.fc1",
86
+ "model.vision_tower.vision_model.encoder.layers.5.mlp.fc2",
87
+ "model.vision_tower.vision_model.encoder.layers.6.self_attn.k_proj",
88
+ "model.vision_tower.vision_model.encoder.layers.6.self_attn.v_proj",
89
+ "model.vision_tower.vision_model.encoder.layers.6.self_attn.q_proj",
90
+ "model.vision_tower.vision_model.encoder.layers.6.self_attn.out_proj",
91
+ "model.vision_tower.vision_model.encoder.layers.6.mlp.fc1",
92
+ "model.vision_tower.vision_model.encoder.layers.6.mlp.fc2",
93
+ "model.vision_tower.vision_model.encoder.layers.7.self_attn.k_proj",
94
+ "model.vision_tower.vision_model.encoder.layers.7.self_attn.v_proj",
95
+ "model.vision_tower.vision_model.encoder.layers.7.self_attn.q_proj",
96
+ "model.vision_tower.vision_model.encoder.layers.7.self_attn.out_proj",
97
+ "model.vision_tower.vision_model.encoder.layers.7.mlp.fc1",
98
+ "model.vision_tower.vision_model.encoder.layers.7.mlp.fc2",
99
+ "model.vision_tower.vision_model.encoder.layers.8.self_attn.k_proj",
100
+ "model.vision_tower.vision_model.encoder.layers.8.self_attn.v_proj",
101
+ "model.vision_tower.vision_model.encoder.layers.8.self_attn.q_proj",
102
+ "model.vision_tower.vision_model.encoder.layers.8.self_attn.out_proj",
103
+ "model.vision_tower.vision_model.encoder.layers.8.mlp.fc1",
104
+ "model.vision_tower.vision_model.encoder.layers.8.mlp.fc2",
105
+ "model.vision_tower.vision_model.encoder.layers.9.self_attn.k_proj",
106
+ "model.vision_tower.vision_model.encoder.layers.9.self_attn.v_proj",
107
+ "model.vision_tower.vision_model.encoder.layers.9.self_attn.q_proj",
108
+ "model.vision_tower.vision_model.encoder.layers.9.self_attn.out_proj",
109
+ "model.vision_tower.vision_model.encoder.layers.9.mlp.fc1",
110
+ "model.vision_tower.vision_model.encoder.layers.9.mlp.fc2",
111
+ "model.vision_tower.vision_model.encoder.layers.10.self_attn.k_proj",
112
+ "model.vision_tower.vision_model.encoder.layers.10.self_attn.v_proj",
113
+ "model.vision_tower.vision_model.encoder.layers.10.self_attn.q_proj",
114
+ "model.vision_tower.vision_model.encoder.layers.10.self_attn.out_proj",
115
+ "model.vision_tower.vision_model.encoder.layers.10.mlp.fc1",
116
+ "model.vision_tower.vision_model.encoder.layers.10.mlp.fc2",
117
+ "model.vision_tower.vision_model.encoder.layers.11.self_attn.k_proj",
118
+ "model.vision_tower.vision_model.encoder.layers.11.self_attn.v_proj",
119
+ "model.vision_tower.vision_model.encoder.layers.11.self_attn.q_proj",
120
+ "model.vision_tower.vision_model.encoder.layers.11.self_attn.out_proj",
121
+ "model.vision_tower.vision_model.encoder.layers.11.mlp.fc1",
122
+ "model.vision_tower.vision_model.encoder.layers.11.mlp.fc2",
123
+ "model.vision_tower.vision_model.encoder.layers.12.self_attn.k_proj",
124
+ "model.vision_tower.vision_model.encoder.layers.12.self_attn.v_proj",
125
+ "model.vision_tower.vision_model.encoder.layers.12.self_attn.q_proj",
126
+ "model.vision_tower.vision_model.encoder.layers.12.self_attn.out_proj",
127
+ "model.vision_tower.vision_model.encoder.layers.12.mlp.fc1",
128
+ "model.vision_tower.vision_model.encoder.layers.12.mlp.fc2",
129
+ "model.vision_tower.vision_model.encoder.layers.13.self_attn.k_proj",
130
+ "model.vision_tower.vision_model.encoder.layers.13.self_attn.v_proj",
131
+ "model.vision_tower.vision_model.encoder.layers.13.self_attn.q_proj",
132
+ "model.vision_tower.vision_model.encoder.layers.13.self_attn.out_proj",
133
+ "model.vision_tower.vision_model.encoder.layers.13.mlp.fc1",
134
+ "model.vision_tower.vision_model.encoder.layers.13.mlp.fc2",
135
+ "model.vision_tower.vision_model.encoder.layers.14.self_attn.k_proj",
136
+ "model.vision_tower.vision_model.encoder.layers.14.self_attn.v_proj",
137
+ "model.vision_tower.vision_model.encoder.layers.14.self_attn.q_proj",
138
+ "model.vision_tower.vision_model.encoder.layers.14.self_attn.out_proj",
139
+ "model.vision_tower.vision_model.encoder.layers.14.mlp.fc1",
140
+ "model.vision_tower.vision_model.encoder.layers.14.mlp.fc2",
141
+ "model.vision_tower.vision_model.encoder.layers.15.self_attn.k_proj",
142
+ "model.vision_tower.vision_model.encoder.layers.15.self_attn.v_proj",
143
+ "model.vision_tower.vision_model.encoder.layers.15.self_attn.q_proj",
144
+ "model.vision_tower.vision_model.encoder.layers.15.self_attn.out_proj",
145
+ "model.vision_tower.vision_model.encoder.layers.15.mlp.fc1",
146
+ "model.vision_tower.vision_model.encoder.layers.15.mlp.fc2",
147
+ "model.vision_tower.vision_model.encoder.layers.16.self_attn.k_proj",
148
+ "model.vision_tower.vision_model.encoder.layers.16.self_attn.v_proj",
149
+ "model.vision_tower.vision_model.encoder.layers.16.self_attn.q_proj",
150
+ "model.vision_tower.vision_model.encoder.layers.16.self_attn.out_proj",
151
+ "model.vision_tower.vision_model.encoder.layers.16.mlp.fc1",
152
+ "model.vision_tower.vision_model.encoder.layers.16.mlp.fc2",
153
+ "model.vision_tower.vision_model.encoder.layers.17.self_attn.k_proj",
154
+ "model.vision_tower.vision_model.encoder.layers.17.self_attn.v_proj",
155
+ "model.vision_tower.vision_model.encoder.layers.17.self_attn.q_proj",
156
+ "model.vision_tower.vision_model.encoder.layers.17.self_attn.out_proj",
157
+ "model.vision_tower.vision_model.encoder.layers.17.mlp.fc1",
158
+ "model.vision_tower.vision_model.encoder.layers.17.mlp.fc2",
159
+ "model.vision_tower.vision_model.encoder.layers.18.self_attn.k_proj",
160
+ "model.vision_tower.vision_model.encoder.layers.18.self_attn.v_proj",
161
+ "model.vision_tower.vision_model.encoder.layers.18.self_attn.q_proj",
162
+ "model.vision_tower.vision_model.encoder.layers.18.self_attn.out_proj",
163
+ "model.vision_tower.vision_model.encoder.layers.18.mlp.fc1",
164
+ "model.vision_tower.vision_model.encoder.layers.18.mlp.fc2",
165
+ "model.vision_tower.vision_model.encoder.layers.19.self_attn.k_proj",
166
+ "model.vision_tower.vision_model.encoder.layers.19.self_attn.v_proj",
167
+ "model.vision_tower.vision_model.encoder.layers.19.self_attn.q_proj",
168
+ "model.vision_tower.vision_model.encoder.layers.19.self_attn.out_proj",
169
+ "model.vision_tower.vision_model.encoder.layers.19.mlp.fc1",
170
+ "model.vision_tower.vision_model.encoder.layers.19.mlp.fc2",
171
+ "model.vision_tower.vision_model.encoder.layers.20.self_attn.k_proj",
172
+ "model.vision_tower.vision_model.encoder.layers.20.self_attn.v_proj",
173
+ "model.vision_tower.vision_model.encoder.layers.20.self_attn.q_proj",
174
+ "model.vision_tower.vision_model.encoder.layers.20.self_attn.out_proj",
175
+ "model.vision_tower.vision_model.encoder.layers.20.mlp.fc1",
176
+ "model.vision_tower.vision_model.encoder.layers.20.mlp.fc2",
177
+ "model.vision_tower.vision_model.encoder.layers.21.self_attn.k_proj",
178
+ "model.vision_tower.vision_model.encoder.layers.21.self_attn.v_proj",
179
+ "model.vision_tower.vision_model.encoder.layers.21.self_attn.q_proj",
180
+ "model.vision_tower.vision_model.encoder.layers.21.self_attn.out_proj",
181
+ "model.vision_tower.vision_model.encoder.layers.21.mlp.fc1",
182
+ "model.vision_tower.vision_model.encoder.layers.21.mlp.fc2",
183
+ "model.vision_tower.vision_model.encoder.layers.22.self_attn.k_proj",
184
+ "model.vision_tower.vision_model.encoder.layers.22.self_attn.v_proj",
185
+ "model.vision_tower.vision_model.encoder.layers.22.self_attn.q_proj",
186
+ "model.vision_tower.vision_model.encoder.layers.22.self_attn.out_proj",
187
+ "model.vision_tower.vision_model.encoder.layers.22.mlp.fc1",
188
+ "model.vision_tower.vision_model.encoder.layers.22.mlp.fc2",
189
+ "model.vision_tower.vision_model.encoder.layers.23.self_attn.k_proj",
190
+ "model.vision_tower.vision_model.encoder.layers.23.self_attn.v_proj",
191
+ "model.vision_tower.vision_model.encoder.layers.23.self_attn.q_proj",
192
+ "model.vision_tower.vision_model.encoder.layers.23.self_attn.out_proj",
193
+ "model.vision_tower.vision_model.encoder.layers.23.mlp.fc1",
194
+ "model.vision_tower.vision_model.encoder.layers.23.mlp.fc2",
195
+ "model.vision_tower.vision_model.encoder.layers.24.self_attn.k_proj",
196
+ "model.vision_tower.vision_model.encoder.layers.24.self_attn.v_proj",
197
+ "model.vision_tower.vision_model.encoder.layers.24.self_attn.q_proj",
198
+ "model.vision_tower.vision_model.encoder.layers.24.self_attn.out_proj",
199
+ "model.vision_tower.vision_model.encoder.layers.24.mlp.fc1",
200
+ "model.vision_tower.vision_model.encoder.layers.24.mlp.fc2",
201
+ "model.vision_tower.vision_model.encoder.layers.25.self_attn.k_proj",
202
+ "model.vision_tower.vision_model.encoder.layers.25.self_attn.v_proj",
203
+ "model.vision_tower.vision_model.encoder.layers.25.self_attn.q_proj",
204
+ "model.vision_tower.vision_model.encoder.layers.25.self_attn.out_proj",
205
+ "model.vision_tower.vision_model.encoder.layers.25.mlp.fc1",
206
+ "model.vision_tower.vision_model.encoder.layers.25.mlp.fc2",
207
+ "model.vision_tower.vision_model.encoder.layers.26.self_attn.k_proj",
208
+ "model.vision_tower.vision_model.encoder.layers.26.self_attn.v_proj",
209
+ "model.vision_tower.vision_model.encoder.layers.26.self_attn.q_proj",
210
+ "model.vision_tower.vision_model.encoder.layers.26.self_attn.out_proj",
211
+ "model.vision_tower.vision_model.encoder.layers.26.mlp.fc1",
212
+ "model.vision_tower.vision_model.encoder.layers.26.mlp.fc2",
213
+ "lm_head"
214
+ ],
215
+ "kv_cache_scheme": null,
216
+ "quant_method": "compressed-tensors",
217
+ "quantization_status": "compressed"
218
+ },
219
+ "text_config": {
220
+ "_sliding_window_pattern": 6,
221
+ "attention_bias": false,
222
+ "attention_dropout": 0.0,
223
+ "attn_logit_softcapping": null,
224
+ "cache_implementation": "hybrid",
225
+ "final_logit_softcapping": null,
226
+ "head_dim": 256,
227
+ "hidden_activation": "gelu_pytorch_tanh",
228
+ "hidden_size": 3840,
229
+ "initializer_range": 0.02,
230
+ "intermediate_size": 15360,
231
+ "layer_types": [
232
+ "sliding_attention",
233
+ "sliding_attention",
234
+ "sliding_attention",
235
+ "sliding_attention",
236
+ "sliding_attention",
237
+ "full_attention",
238
+ "sliding_attention",
239
+ "sliding_attention",
240
+ "sliding_attention",
241
+ "sliding_attention",
242
+ "sliding_attention",
243
+ "full_attention",
244
+ "sliding_attention",
245
+ "sliding_attention",
246
+ "sliding_attention",
247
+ "sliding_attention",
248
+ "sliding_attention",
249
+ "full_attention",
250
+ "sliding_attention",
251
+ "sliding_attention",
252
+ "sliding_attention",
253
+ "sliding_attention",
254
+ "sliding_attention",
255
+ "full_attention",
256
+ "sliding_attention",
257
+ "sliding_attention",
258
+ "sliding_attention",
259
+ "sliding_attention",
260
+ "sliding_attention",
261
+ "full_attention",
262
+ "sliding_attention",
263
+ "sliding_attention",
264
+ "sliding_attention",
265
+ "sliding_attention",
266
+ "sliding_attention",
267
+ "full_attention",
268
+ "sliding_attention",
269
+ "sliding_attention",
270
+ "sliding_attention",
271
+ "sliding_attention",
272
+ "sliding_attention",
273
+ "full_attention",
274
+ "sliding_attention",
275
+ "sliding_attention",
276
+ "sliding_attention",
277
+ "sliding_attention",
278
+ "sliding_attention",
279
+ "full_attention"
280
+ ],
281
+ "max_position_embeddings": 131072,
282
+ "model_type": "gemma3_text",
283
+ "num_attention_heads": 16,
284
+ "num_hidden_layers": 48,
285
+ "num_key_value_heads": 8,
286
+ "query_pre_attn_scalar": 256,
287
+ "rms_norm_eps": 1e-06,
288
+ "rope_local_base_freq": 10000.0,
289
+ "rope_scaling": {
290
+ "factor": 8.0,
291
+ "rope_type": "linear"
292
+ },
293
+ "rope_theta": 1000000.0,
294
+ "sliding_window": 1024,
295
+ "torch_dtype": "bfloat16",
296
+ "use_cache": true,
297
+ "vocab_size": 262208
298
+ },
299
+ "torch_dtype": "bfloat16",
300
+ "transformers_version": "4.53.2",
301
+ "use_cache": false,
302
+ "vision_config": {
303
+ "attention_dropout": 0.0,
304
+ "hidden_act": "gelu_pytorch_tanh",
305
+ "hidden_size": 1152,
306
+ "image_size": 896,
307
+ "intermediate_size": 4304,
308
+ "layer_norm_eps": 1e-06,
309
+ "model_type": "siglip_vision_model",
310
+ "num_attention_heads": 16,
311
+ "num_channels": 3,
312
+ "num_hidden_layers": 27,
313
+ "patch_size": 14,
314
+ "torch_dtype": "bfloat16",
315
+ "vision_use_head": false
316
+ }
317
+ }
generation_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 2,
3
+ "cache_implementation": "hybrid",
4
+ "do_sample": true,
5
+ "eos_token_id": [
6
+ 1,
7
+ 106
8
+ ],
9
+ "pad_token_id": 0,
10
+ "top_k": 64,
11
+ "top_p": 0.95,
12
+ "transformers_version": "4.53.2"
13
+ }
model-00001-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5077835068293aa879153c607c672bae3a1fc68e86761c402fa046c69302a809
3
+ size 4980972040
model-00002-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f630add84b579638cc8d96914acc75562c407a41ac7a272ba78157013f06486
3
+ size 4992747400
model-00003-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:24e4a5de1e4ebb73a4d965a81f4f19bb21212307d6dafdc403f5222fb0572092
3
+ size 3647199496
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
preprocessor_config.json ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "crop_size": null,
3
+ "data_format": "channels_first",
4
+ "default_to_square": true,
5
+ "device": null,
6
+ "disable_grouping": null,
7
+ "do_center_crop": null,
8
+ "do_convert_rgb": null,
9
+ "do_normalize": true,
10
+ "do_pan_and_scan": null,
11
+ "do_rescale": true,
12
+ "do_resize": true,
13
+ "image_mean": [
14
+ 0.5,
15
+ 0.5,
16
+ 0.5
17
+ ],
18
+ "image_processor_type": "Gemma3ImageProcessorFast",
19
+ "image_seq_length": 256,
20
+ "image_std": [
21
+ 0.5,
22
+ 0.5,
23
+ 0.5
24
+ ],
25
+ "input_data_format": null,
26
+ "pan_and_scan_max_num_crops": null,
27
+ "pan_and_scan_min_crop_size": null,
28
+ "pan_and_scan_min_ratio_to_activate": null,
29
+ "processor_class": "Gemma3Processor",
30
+ "resample": 2,
31
+ "rescale_factor": 0.00392156862745098,
32
+ "return_tensors": null,
33
+ "size": {
34
+ "height": 896,
35
+ "width": 896
36
+ }
37
+ }
processor_config.json ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+ {
2
+ "image_seq_length": 256,
3
+ "processor_class": "Gemma3Processor"
4
+ }
recipe.yaml ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ default_stage:
2
+ default_modifiers:
3
+ QuantizationModifier:
4
+ targets: [Linear]
5
+ ignore: [lm_head, 're:model\.vision_tower.*', 're:model\.multi_modal_projector.*']
6
+ 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:d786405177734910d7a3db625c2826640964a0b4e5cdbbd70620ae3313a01bef
3
+ size 33384722
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