Hugging Face's logo

hnv2520
/
excavator_Gemma3_12B_4bit_3e

Image-Text-to-Text
Transformers
Safetensors
English
gemma3
text-generation-inference
unsloth
conversational
4-bit precision
bitsandbytes
Model card Files Community
hnv2520 commited on
Commit
89b5571
·
verified ·
1 Parent(s): 40e2cd7

(Trained with Unsloth)

Browse files
Files changed (15) hide show
  1. .gitattributes +1 -0
  2. added_tokens.json +3 -0
  3. chat_template.json +3 -0
  4. config.json +308 -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 +29 -0
  11. processor_config.json +4 -0
  12. special_tokens_map.json +33 -0
  13. tokenizer.json +3 -0
  14. tokenizer.model +3 -0
  15. 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.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ {
2
+ "chat_template": "{{ bos_token }}\n{%- if messages[0]['role'] == 'system' -%}\n {%- if messages[0]['content'] is string -%}\n {%- set first_user_prefix = messages[0]['content'] + '\n\n' -%}\n {%- else -%}\n {%- set first_user_prefix = messages[0]['content'][0]['text'] + '\n\n' -%}\n {%- endif -%}\n {%- set loop_messages = messages[1:] -%}\n{%- else -%}\n {%- set first_user_prefix = \"\" -%}\n {%- set loop_messages = messages -%}\n{%- endif -%}\n{%- for message in loop_messages -%}\n {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}\n {{ raise_exception(\"Conversation roles must alternate user/assistant/user/assistant/...\") }}\n {%- endif -%}\n {%- if (message['role'] == 'assistant') -%}\n {%- set role = \"model\" -%}\n {%- else -%}\n {%- set role = message['role'] -%}\n {%- endif -%}\n {{ '<start_of_turn>' + role + '\n' + (first_user_prefix if loop.first else \"\") }}\n {%- if message['content'] is string -%}\n {{ message['content'] | trim }}\n {%- elif message['content'] is iterable -%}\n {%- for item in message['content'] -%}\n {%- if item['type'] == 'image' -%}\n {{ '<start_of_image>' }}\n {%- elif item['type'] == 'text' -%}\n {{ item['text'] | trim }}\n {%- endif -%}\n {%- endfor -%}\n {%- else -%}\n {{ raise_exception(\"Invalid content type\") }}\n {%- endif -%}\n {{ '<end_of_turn>\n' }}\n{%- endfor -%}\n{%- if add_generation_prompt -%}\n {{'<start_of_turn>model\n'}}\n{%- endif -%}\n"
3
+ }
config.json ADDED
@@ -0,0 +1,308 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Gemma3ForConditionalGeneration"
4
+ ],
5
+ "boi_token_index": 255999,
6
+ "bos_token_id": 2,
7
+ "eoi_token_index": 256000,
8
+ "eos_token_id": 106,
9
+ "image_token_index": 262144,
10
+ "initializer_range": 0.02,
11
+ "mm_tokens_per_image": 256,
12
+ "model_type": "gemma3",
13
+ "pad_token_id": 0,
14
+ "quantization_config": {
15
+ "bnb_4bit_compute_dtype": "bfloat16",
16
+ "bnb_4bit_quant_type": "nf4",
17
+ "bnb_4bit_use_double_quant": true,
18
+ "llm_int8_enable_fp32_cpu_offload": false,
19
+ "llm_int8_has_fp16_weight": false,
20
+ "llm_int8_skip_modules": [
21
+ "vision_tower.vision_model.encoder.layers.0.self_attn.k_proj",
22
+ "vision_tower.vision_model.encoder.layers.0.self_attn.v_proj",
23
+ "vision_tower.vision_model.encoder.layers.0.self_attn.q_proj",
24
+ "vision_tower.vision_model.encoder.layers.0.self_attn.out_proj",
25
+ "vision_tower.vision_model.encoder.layers.0.mlp.fc1",
26
+ "vision_tower.vision_model.encoder.layers.0.mlp.fc2",
27
+ "vision_tower.vision_model.encoder.layers.1.self_attn.k_proj",
28
+ "vision_tower.vision_model.encoder.layers.1.self_attn.v_proj",
29
+ "vision_tower.vision_model.encoder.layers.1.self_attn.q_proj",
30
+ "vision_tower.vision_model.encoder.layers.1.self_attn.out_proj",
31
+ "vision_tower.vision_model.encoder.layers.1.mlp.fc1",
32
+ "vision_tower.vision_model.encoder.layers.1.mlp.fc2",
33
+ "vision_tower.vision_model.encoder.layers.2.self_attn.k_proj",
34
+ "vision_tower.vision_model.encoder.layers.2.self_attn.v_proj",
35
+ "vision_tower.vision_model.encoder.layers.2.self_attn.q_proj",
36
+ "vision_tower.vision_model.encoder.layers.2.self_attn.out_proj",
37
+ "vision_tower.vision_model.encoder.layers.2.mlp.fc1",
38
+ "vision_tower.vision_model.encoder.layers.2.mlp.fc2",
39
+ "vision_tower.vision_model.encoder.layers.3.self_attn.k_proj",
40
+ "vision_tower.vision_model.encoder.layers.3.self_attn.v_proj",
41
+ "vision_tower.vision_model.encoder.layers.3.self_attn.q_proj",
42
+ "vision_tower.vision_model.encoder.layers.3.self_attn.out_proj",
43
+ "vision_tower.vision_model.encoder.layers.3.mlp.fc1",
44
+ "vision_tower.vision_model.encoder.layers.3.mlp.fc2",
45
+ "vision_tower.vision_model.encoder.layers.4.self_attn.k_proj",
46
+ "vision_tower.vision_model.encoder.layers.4.self_attn.v_proj",
47
+ "vision_tower.vision_model.encoder.layers.4.self_attn.q_proj",
48
+ "vision_tower.vision_model.encoder.layers.4.self_attn.out_proj",
49
+ "vision_tower.vision_model.encoder.layers.4.mlp.fc1",
50
+ "vision_tower.vision_model.encoder.layers.4.mlp.fc2",
51
+ "vision_tower.vision_model.encoder.layers.5.self_attn.k_proj",
52
+ "vision_tower.vision_model.encoder.layers.5.self_attn.v_proj",
53
+ "vision_tower.vision_model.encoder.layers.5.self_attn.q_proj",
54
+ "vision_tower.vision_model.encoder.layers.5.self_attn.out_proj",
55
+ "vision_tower.vision_model.encoder.layers.5.mlp.fc1",
56
+ "vision_tower.vision_model.encoder.layers.5.mlp.fc2",
57
+ "vision_tower.vision_model.encoder.layers.6.self_attn.k_proj",
58
+ "vision_tower.vision_model.encoder.layers.6.self_attn.v_proj",
59
+ "vision_tower.vision_model.encoder.layers.6.self_attn.q_proj",
60
+ "vision_tower.vision_model.encoder.layers.6.self_attn.out_proj",
61
+ "vision_tower.vision_model.encoder.layers.6.mlp.fc1",
62
+ "vision_tower.vision_model.encoder.layers.6.mlp.fc2",
63
+ "vision_tower.vision_model.encoder.layers.7.self_attn.k_proj",
64
+ "vision_tower.vision_model.encoder.layers.7.self_attn.v_proj",
65
+ "vision_tower.vision_model.encoder.layers.7.self_attn.q_proj",
66
+ "vision_tower.vision_model.encoder.layers.7.self_attn.out_proj",
67
+ "vision_tower.vision_model.encoder.layers.7.mlp.fc1",
68
+ "vision_tower.vision_model.encoder.layers.7.mlp.fc2",
69
+ "vision_tower.vision_model.encoder.layers.8.self_attn.k_proj",
70
+ "vision_tower.vision_model.encoder.layers.8.self_attn.v_proj",
71
+ "vision_tower.vision_model.encoder.layers.8.self_attn.q_proj",
72
+ "vision_tower.vision_model.encoder.layers.8.self_attn.out_proj",
73
+ "vision_tower.vision_model.encoder.layers.8.mlp.fc1",
74
+ "vision_tower.vision_model.encoder.layers.8.mlp.fc2",
75
+ "vision_tower.vision_model.encoder.layers.9.self_attn.k_proj",
76
+ "vision_tower.vision_model.encoder.layers.9.self_attn.v_proj",
77
+ "vision_tower.vision_model.encoder.layers.9.self_attn.q_proj",
78
+ "vision_tower.vision_model.encoder.layers.9.self_attn.out_proj",
79
+ "vision_tower.vision_model.encoder.layers.9.mlp.fc1",
80
+ "vision_tower.vision_model.encoder.layers.9.mlp.fc2",
81
+ "vision_tower.vision_model.encoder.layers.10.self_attn.k_proj",
82
+ "vision_tower.vision_model.encoder.layers.10.self_attn.v_proj",
83
+ "vision_tower.vision_model.encoder.layers.10.self_attn.q_proj",
84
+ "vision_tower.vision_model.encoder.layers.10.self_attn.out_proj",
85
+ "vision_tower.vision_model.encoder.layers.10.mlp.fc1",
86
+ "vision_tower.vision_model.encoder.layers.10.mlp.fc2",
87
+ "vision_tower.vision_model.encoder.layers.11.self_attn.k_proj",
88
+ "vision_tower.vision_model.encoder.layers.11.self_attn.v_proj",
89
+ "vision_tower.vision_model.encoder.layers.11.self_attn.q_proj",
90
+ "vision_tower.vision_model.encoder.layers.11.self_attn.out_proj",
91
+ "vision_tower.vision_model.encoder.layers.11.mlp.fc1",
92
+ "vision_tower.vision_model.encoder.layers.11.mlp.fc2",
93
+ "vision_tower.vision_model.encoder.layers.12.self_attn.k_proj",
94
+ "vision_tower.vision_model.encoder.layers.12.self_attn.v_proj",
95
+ "vision_tower.vision_model.encoder.layers.12.self_attn.q_proj",
96
+ "vision_tower.vision_model.encoder.layers.12.self_attn.out_proj",
97
+ "vision_tower.vision_model.encoder.layers.12.mlp.fc1",
98
+ "vision_tower.vision_model.encoder.layers.12.mlp.fc2",
99
+ "vision_tower.vision_model.encoder.layers.13.self_attn.k_proj",
100
+ "vision_tower.vision_model.encoder.layers.13.self_attn.v_proj",
101
+ "vision_tower.vision_model.encoder.layers.13.self_attn.q_proj",
102
+ "vision_tower.vision_model.encoder.layers.13.self_attn.out_proj",
103
+ "vision_tower.vision_model.encoder.layers.13.mlp.fc1",
104
+ "vision_tower.vision_model.encoder.layers.13.mlp.fc2",
105
+ "vision_tower.vision_model.encoder.layers.14.self_attn.k_proj",
106
+ "vision_tower.vision_model.encoder.layers.14.self_attn.v_proj",
107
+ "vision_tower.vision_model.encoder.layers.14.self_attn.q_proj",
108
+ "vision_tower.vision_model.encoder.layers.14.self_attn.out_proj",
109
+ "vision_tower.vision_model.encoder.layers.14.mlp.fc1",
110
+ "vision_tower.vision_model.encoder.layers.14.mlp.fc2",
111
+ "vision_tower.vision_model.encoder.layers.15.self_attn.k_proj",
112
+ "vision_tower.vision_model.encoder.layers.15.self_attn.v_proj",
113
+ "vision_tower.vision_model.encoder.layers.15.self_attn.q_proj",
114
+ "vision_tower.vision_model.encoder.layers.15.self_attn.out_proj",
115
+ "vision_tower.vision_model.encoder.layers.15.mlp.fc1",
116
+ "vision_tower.vision_model.encoder.layers.15.mlp.fc2",
117
+ "vision_tower.vision_model.encoder.layers.16.self_attn.k_proj",
118
+ "vision_tower.vision_model.encoder.layers.16.self_attn.v_proj",
119
+ "vision_tower.vision_model.encoder.layers.16.self_attn.q_proj",
120
+ "vision_tower.vision_model.encoder.layers.16.self_attn.out_proj",
121
+ "vision_tower.vision_model.encoder.layers.16.mlp.fc1",
122
+ "vision_tower.vision_model.encoder.layers.16.mlp.fc2",
123
+ "vision_tower.vision_model.encoder.layers.17.self_attn.k_proj",
124
+ "vision_tower.vision_model.encoder.layers.17.self_attn.v_proj",
125
+ "vision_tower.vision_model.encoder.layers.17.self_attn.q_proj",
126
+ "vision_tower.vision_model.encoder.layers.17.self_attn.out_proj",
127
+ "vision_tower.vision_model.encoder.layers.17.mlp.fc1",
128
+ "vision_tower.vision_model.encoder.layers.17.mlp.fc2",
129
+ "vision_tower.vision_model.encoder.layers.18.self_attn.k_proj",
130
+ "vision_tower.vision_model.encoder.layers.18.self_attn.v_proj",
131
+ "vision_tower.vision_model.encoder.layers.18.self_attn.q_proj",
132
+ "vision_tower.vision_model.encoder.layers.18.self_attn.out_proj",
133
+ "vision_tower.vision_model.encoder.layers.18.mlp.fc1",
134
+ "vision_tower.vision_model.encoder.layers.18.mlp.fc2",
135
+ "vision_tower.vision_model.encoder.layers.19.self_attn.k_proj",
136
+ "vision_tower.vision_model.encoder.layers.19.self_attn.v_proj",
137
+ "vision_tower.vision_model.encoder.layers.19.self_attn.q_proj",
138
+ "vision_tower.vision_model.encoder.layers.19.self_attn.out_proj",
139
+ "vision_tower.vision_model.encoder.layers.19.mlp.fc1",
140
+ "vision_tower.vision_model.encoder.layers.19.mlp.fc2",
141
+ "vision_tower.vision_model.encoder.layers.20.self_attn.k_proj",
142
+ "vision_tower.vision_model.encoder.layers.20.self_attn.v_proj",
143
+ "vision_tower.vision_model.encoder.layers.20.self_attn.q_proj",
144
+ "vision_tower.vision_model.encoder.layers.20.self_attn.out_proj",
145
+ "vision_tower.vision_model.encoder.layers.20.mlp.fc1",
146
+ "vision_tower.vision_model.encoder.layers.20.mlp.fc2",
147
+ "vision_tower.vision_model.encoder.layers.21.self_attn.k_proj",
148
+ "vision_tower.vision_model.encoder.layers.21.self_attn.v_proj",
149
+ "vision_tower.vision_model.encoder.layers.21.self_attn.q_proj",
150
+ "vision_tower.vision_model.encoder.layers.21.self_attn.out_proj",
151
+ "vision_tower.vision_model.encoder.layers.21.mlp.fc1",
152
+ "vision_tower.vision_model.encoder.layers.21.mlp.fc2",
153
+ "vision_tower.vision_model.encoder.layers.22.self_attn.k_proj",
154
+ "vision_tower.vision_model.encoder.layers.22.self_attn.v_proj",
155
+ "vision_tower.vision_model.encoder.layers.22.self_attn.q_proj",
156
+ "vision_tower.vision_model.encoder.layers.22.self_attn.out_proj",
157
+ "vision_tower.vision_model.encoder.layers.22.mlp.fc1",
158
+ "vision_tower.vision_model.encoder.layers.22.mlp.fc2",
159
+ "vision_tower.vision_model.encoder.layers.23.self_attn.k_proj",
160
+ "vision_tower.vision_model.encoder.layers.23.self_attn.v_proj",
161
+ "vision_tower.vision_model.encoder.layers.23.self_attn.q_proj",
162
+ "vision_tower.vision_model.encoder.layers.23.self_attn.out_proj",
163
+ "vision_tower.vision_model.encoder.layers.23.mlp.fc1",
164
+ "vision_tower.vision_model.encoder.layers.23.mlp.fc2",
165
+ "vision_tower.vision_model.encoder.layers.24.self_attn.k_proj",
166
+ "vision_tower.vision_model.encoder.layers.24.self_attn.v_proj",
167
+ "vision_tower.vision_model.encoder.layers.24.self_attn.q_proj",
168
+ "vision_tower.vision_model.encoder.layers.24.self_attn.out_proj",
169
+ "vision_tower.vision_model.encoder.layers.24.mlp.fc1",
170
+ "vision_tower.vision_model.encoder.layers.24.mlp.fc2",
171
+ "vision_tower.vision_model.encoder.layers.25.self_attn.k_proj",
172
+ "vision_tower.vision_model.encoder.layers.25.self_attn.v_proj",
173
+ "vision_tower.vision_model.encoder.layers.25.self_attn.q_proj",
174
+ "vision_tower.vision_model.encoder.layers.25.self_attn.out_proj",
175
+ "vision_tower.vision_model.encoder.layers.25.mlp.fc1",
176
+ "vision_tower.vision_model.encoder.layers.25.mlp.fc2",
177
+ "vision_tower.vision_model.encoder.layers.26.self_attn.k_proj",
178
+ "vision_tower.vision_model.encoder.layers.26.self_attn.v_proj",
179
+ "vision_tower.vision_model.encoder.layers.26.self_attn.q_proj",
180
+ "vision_tower.vision_model.encoder.layers.26.self_attn.out_proj",
181
+ "vision_tower.vision_model.encoder.layers.26.mlp.fc1",
182
+ "vision_tower.vision_model.encoder.layers.26.mlp.fc2",
183
+ "language_model.model.layers.1.mlp.gate_proj",
184
+ "language_model.model.layers.1.mlp.up_proj",
185
+ "language_model.model.layers.1.mlp.down_proj",
186
+ "language_model.model.layers.2.self_attn.q_proj",
187
+ "language_model.model.layers.2.self_attn.k_proj",
188
+ "language_model.model.layers.2.self_attn.v_proj",
189
+ "language_model.model.layers.2.self_attn.o_proj",
190
+ "language_model.model.layers.2.mlp.gate_proj",
191
+ "language_model.model.layers.2.mlp.up_proj",
192
+ "language_model.model.layers.2.mlp.down_proj",
193
+ "language_model.model.layers.3.self_attn.q_proj",
194
+ "language_model.model.layers.3.self_attn.k_proj",
195
+ "language_model.model.layers.3.self_attn.v_proj",
196
+ "language_model.model.layers.3.self_attn.o_proj",
197
+ "language_model.model.layers.3.mlp.gate_proj",
198
+ "language_model.model.layers.3.mlp.up_proj",
199
+ "language_model.model.layers.3.mlp.down_proj",
200
+ "language_model.model.layers.4.mlp.gate_proj",
201
+ "language_model.model.layers.4.mlp.up_proj",
202
+ "language_model.model.layers.4.mlp.down_proj",
203
+ "language_model.model.layers.5.self_attn.q_proj",
204
+ "language_model.model.layers.5.self_attn.k_proj",
205
+ "language_model.model.layers.5.self_attn.v_proj",
206
+ "language_model.model.layers.5.self_attn.o_proj",
207
+ "language_model.model.layers.5.mlp.gate_proj",
208
+ "language_model.model.layers.5.mlp.up_proj",
209
+ "language_model.model.layers.5.mlp.down_proj",
210
+ "language_model.model.layers.6.self_attn.q_proj",
211
+ "language_model.model.layers.6.self_attn.k_proj",
212
+ "language_model.model.layers.6.self_attn.v_proj",
213
+ "language_model.model.layers.6.self_attn.o_proj",
214
+ "language_model.model.layers.6.mlp.gate_proj",
215
+ "language_model.model.layers.6.mlp.up_proj",
216
+ "language_model.model.layers.6.mlp.down_proj",
217
+ "language_model.model.layers.7.self_attn.q_proj",
218
+ "language_model.model.layers.7.self_attn.k_proj",
219
+ "language_model.model.layers.7.self_attn.v_proj",
220
+ "language_model.model.layers.7.self_attn.o_proj",
221
+ "language_model.model.layers.7.mlp.gate_proj",
222
+ "language_model.model.layers.7.mlp.up_proj",
223
+ "language_model.model.layers.7.mlp.down_proj",
224
+ "language_model.model.layers.8.self_attn.q_proj",
225
+ "language_model.model.layers.8.self_attn.k_proj",
226
+ "language_model.model.layers.8.self_attn.v_proj",
227
+ "language_model.model.layers.8.self_attn.o_proj",
228
+ "language_model.model.layers.8.mlp.gate_proj",
229
+ "language_model.model.layers.8.mlp.up_proj",
230
+ "language_model.model.layers.8.mlp.down_proj",
231
+ "language_model.model.layers.9.mlp.gate_proj",
232
+ "language_model.model.layers.9.mlp.up_proj",
233
+ "language_model.model.layers.9.mlp.down_proj",
234
+ "language_model.model.layers.10.mlp.gate_proj",
235
+ "language_model.model.layers.10.mlp.up_proj",
236
+ "language_model.model.layers.10.mlp.down_proj",
237
+ "language_model.model.layers.11.mlp.gate_proj",
238
+ "language_model.model.layers.11.mlp.up_proj",
239
+ "language_model.model.layers.11.mlp.down_proj",
240
+ "language_model.model.layers.12.mlp.gate_proj",
241
+ "language_model.model.layers.12.mlp.up_proj",
242
+ "language_model.model.layers.12.mlp.down_proj",
243
+ "language_model.model.layers.13.mlp.gate_proj",
244
+ "language_model.model.layers.13.mlp.up_proj",
245
+ "language_model.model.layers.13.mlp.down_proj",
246
+ "language_model.model.layers.14.mlp.gate_proj",
247
+ "language_model.model.layers.14.mlp.up_proj",
248
+ "language_model.model.layers.14.mlp.down_proj",
249
+ "language_model.model.layers.16.mlp.gate_proj",
250
+ "language_model.model.layers.16.mlp.up_proj",
251
+ "language_model.model.layers.16.mlp.down_proj",
252
+ "language_model.lm_head"
253
+ ],
254
+ "llm_int8_threshold": 6.0,
255
+ "load_in_4bit": true,
256
+ "load_in_8bit": false,
257
+ "quant_method": "bitsandbytes"
258
+ },
259
+ "text_config": {
260
+ "attention_bias": false,
261
+ "attention_dropout": 0.0,
262
+ "attn_logit_softcapping": null,
263
+ "cache_implementation": "hybrid",
264
+ "final_logit_softcapping": null,
265
+ "head_dim": 256,
266
+ "hidden_activation": "gelu_pytorch_tanh",
267
+ "hidden_size": 3840,
268
+ "initializer_range": 0.02,
269
+ "intermediate_size": 15360,
270
+ "max_position_embeddings": 131072,
271
+ "model_type": "gemma3_text",
272
+ "num_attention_heads": 16,
273
+ "num_hidden_layers": 48,
274
+ "num_key_value_heads": 8,
275
+ "query_pre_attn_scalar": 256,
276
+ "rms_norm_eps": 1e-06,
277
+ "rope_local_base_freq": 10000.0,
278
+ "rope_scaling": {
279
+ "factor": 8.0,
280
+ "rope_type": "linear"
281
+ },
282
+ "rope_theta": 1000000.0,
283
+ "sliding_window": 1024,
284
+ "sliding_window_pattern": 6,
285
+ "torch_dtype": "bfloat16",
286
+ "use_cache": true,
287
+ "vocab_size": 262208
288
+ },
289
+ "torch_dtype": "bfloat16",
290
+ "transformers_version": "4.51.3",
291
+ "unsloth_fixed": true,
292
+ "unsloth_version": "2025.5.3",
293
+ "vision_config": {
294
+ "attention_dropout": 0.0,
295
+ "hidden_act": "gelu_pytorch_tanh",
296
+ "hidden_size": 1152,
297
+ "image_size": 896,
298
+ "intermediate_size": 4304,
299
+ "layer_norm_eps": 1e-06,
300
+ "model_type": "siglip_vision_model",
301
+ "num_attention_heads": 16,
302
+ "num_channels": 3,
303
+ "num_hidden_layers": 27,
304
+ "patch_size": 14,
305
+ "torch_dtype": "bfloat16",
306
+ "vision_use_head": false
307
+ }
308
+ }
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": "4.51.3"
13
+ }
model-00001-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:859aa63b3c402773d20578308beea1d1911d08f280f45191241e592618312178
3
+ size 4955527747
model-00002-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3189a601ad04a0b4f08961d4beb4f9db26a558dfc7e9574bba8b110878e01fcb
3
+ size 4975309842
model-00003-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:67b44a4529632c877e60e8c96b46f3bf955f1cf5a5484bd56a902e44a900d3b3
3
+ size 2836985258
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
+ }
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": "<end_of_turn>",
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