Hugging Face's logo

Echo9Zulu
/
gemma-3-12b-it-int8_asym-ov

Image-Text-to-Text
OpenVINO
gemma3
OpenVINO
OpenArc
Optimum-Intel
conversational
Model card Files Community
1
Echo9Zulu commited on
Commit
ec4477d
·
verified ·
1 Parent(s): 869a351

Upload 21 files

Browse files
Files changed (22) hide show
  1. .gitattributes +1 -0
  2. added_tokens.json +3 -0
  3. chat_template.json +3 -0
  4. config.json +63 -0
  5. generation_config.json +13 -0
  6. openvino_config.json +27 -0
  7. openvino_detokenizer.bin +3 -0
  8. openvino_detokenizer.xml +364 -0
  9. openvino_language_model.bin +3 -0
  10. openvino_language_model.xml +0 -0
  11. openvino_text_embeddings_model.bin +3 -0
  12. openvino_text_embeddings_model.xml +207 -0
  13. openvino_tokenizer.bin +3 -0
  14. openvino_tokenizer.xml +837 -0
  15. openvino_vision_embeddings_model.bin +3 -0
  16. openvino_vision_embeddings_model.xml +0 -0
  17. preprocessor_config.json +29 -0
  18. processor_config.json +4 -0
  19. special_tokens_map.json +33 -0
  20. tokenizer.json +3 -0
  21. tokenizer.model +3 -0
  22. 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 +1,63 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_attn_implementation_autoset": true,
3
+ "architectures": [
4
+ "Gemma3ForConditionalGeneration"
5
+ ],
6
+ "boi_token_index": 255999,
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
+ "text_config": {
17
+ "attention_bias": false,
18
+ "attention_dropout": 0.0,
19
+ "attn_logit_softcapping": null,
20
+ "cache_implementation": "hybrid",
21
+ "final_logit_softcapping": null,
22
+ "head_dim": 256,
23
+ "hidden_activation": "gelu_pytorch_tanh",
24
+ "hidden_size": 3840,
25
+ "initializer_range": 0.02,
26
+ "intermediate_size": 15360,
27
+ "max_position_embeddings": 131072,
28
+ "model_type": "gemma3_text",
29
+ "num_attention_heads": 16,
30
+ "num_hidden_layers": 48,
31
+ "num_key_value_heads": 8,
32
+ "query_pre_attn_scalar": 256,
33
+ "rms_norm_eps": 1e-06,
34
+ "rope_local_base_freq": 10000.0,
35
+ "rope_scaling": {
36
+ "factor": 8.0,
37
+ "rope_type": "linear"
38
+ },
39
+ "rope_theta": 1000000.0,
40
+ "sliding_window": 1024,
41
+ "sliding_window_pattern": 6,
42
+ "torch_dtype": "bfloat16",
43
+ "use_cache": true,
44
+ "vocab_size": 262208
45
+ },
46
+ "torch_dtype": "bfloat16",
47
+ "transformers_version": "4.50.2",
48
+ "vision_config": {
49
+ "attention_dropout": 0.0,
50
+ "hidden_act": "gelu_pytorch_tanh",
51
+ "hidden_size": 1152,
52
+ "image_size": 896,
53
+ "intermediate_size": 4304,
54
+ "layer_norm_eps": 1e-06,
55
+ "model_type": "siglip_vision_model",
56
+ "num_attention_heads": 16,
57
+ "num_channels": 3,
58
+ "num_hidden_layers": 27,
59
+ "patch_size": 14,
60
+ "torch_dtype": "bfloat16",
61
+ "vision_use_head": false
62
+ }
63
+ }
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": "4.50.2"
13
+ }
openvino_config.json ADDED
@@ -0,0 +1,27 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dtype": "int8",
3
+ "input_info": null,
4
+ "optimum_version": "1.25.0.dev0",
5
+ "quantization_config": {
6
+ "all_layers": null,
7
+ "backup_precision": null,
8
+ "bits": 8,
9
+ "dataset": null,
10
+ "dtype": "int8",
11
+ "gptq": null,
12
+ "group_size": -1,
13
+ "ignored_scope": null,
14
+ "lora_correction": null,
15
+ "num_samples": null,
16
+ "processor": null,
17
+ "quant_method": "default",
18
+ "ratio": 1.0,
19
+ "scale_estimation": null,
20
+ "sensitivity_metric": null,
21
+ "sym": false,
22
+ "tokenizer": null,
23
+ "trust_remote_code": false
24
+ },
25
+ "save_onnx_model": false,
26
+ "transformers_version": "4.50.2"
27
+ }
openvino_detokenizer.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:665e8659ed3667b8a167ec9abf6e8bf413d149838db3633061a9604c7d2bf86d
3
+ size 3365022
openvino_detokenizer.xml ADDED
@@ -0,0 +1,364 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <?xml version="1.0"?>
2
+ <net name="detokenizer" version="11">
3
+ <layers>
4
+ <layer id="0" name="Parameter_2718652" type="Parameter" version="opset1">
5
+ <data shape="?,?" element_type="i64" />
6
+ <output>
7
+ <port id="0" precision="I64" names="Parameter_2718652">
8
+ <dim>-1</dim>
9
+ <dim>-1</dim>
10
+ </port>
11
+ </output>
12
+ </layer>
13
+ <layer id="1" name="Convert_2718674" type="Convert" version="opset1">
14
+ <data destination_type="i32" />
15
+ <input>
16
+ <port id="0" precision="I64">
17
+ <dim>-1</dim>
18
+ <dim>-1</dim>
19
+ </port>
20
+ </input>
21
+ <output>
22
+ <port id="1" precision="I32">
23
+ <dim>-1</dim>
24
+ <dim>-1</dim>
25
+ </port>
26
+ </output>
27
+ </layer>
28
+ <layer id="2" name="Constant_2718619" type="Const" version="opset1">
29
+ <data element_type="u8" shape="3364962" offset="0" size="3364962" />
30
+ <output>
31
+ <port id="0" precision="U8">
32
+ <dim>3364962</dim>
33
+ </port>
34
+ </output>
35
+ </layer>
36
+ <layer id="3" name="StringTensorUnpack_2718620" type="StringTensorUnpack" version="extension">
37
+ <data mode="begins_ends" />
38
+ <input>
39
+ <port id="0" precision="U8">
40
+ <dim>3364962</dim>
41
+ </port>
42
+ </input>
43
+ <output>
44
+ <port id="1" precision="I32">
45
+ <dim>-1</dim>
46
+ </port>
47
+ <port id="2" precision="I32">
48
+ <dim>-1</dim>
49
+ </port>
50
+ <port id="3" precision="U8">
51
+ <dim>-1</dim>
52
+ </port>
53
+ </output>
54
+ </layer>
55
+ <layer id="4" name="Constant_2718656" type="Const" version="opset1">
56
+ <data element_type="i32" shape="9" offset="3364962" size="36" />
57
+ <output>
58
+ <port id="0" precision="I32">
59
+ <dim>9</dim>
60
+ </port>
61
+ </output>
62
+ </layer>
63
+ <layer id="5" name="Constant_2718654" type="Const" version="opset1">
64
+ <data element_type="i32" shape="1" offset="3364998" size="4" />
65
+ <output>
66
+ <port id="0" precision="I32">
67
+ <dim>1</dim>
68
+ </port>
69
+ </output>
70
+ </layer>
71
+ <layer id="6" name="Constant_2718653" type="Const" version="opset1">
72
+ <data element_type="i32" shape="1" offset="3365002" size="4" />
73
+ <output>
74
+ <port id="0" precision="I32">
75
+ <dim>1</dim>
76
+ </port>
77
+ </output>
78
+ </layer>
79
+ <layer id="7" name="Constant_2718655" type="Const" version="opset1">
80
+ <data element_type="i32" shape="1" offset="3365006" size="4" />
81
+ <output>
82
+ <port id="0" precision="I32">
83
+ <dim>1</dim>
84
+ </port>
85
+ </output>
86
+ </layer>
87
+ <layer id="8" name="Constant_2718658" type="Const" version="opset1">
88
+ <data element_type="i64" shape="1" offset="3365010" size="8" />
89
+ <output>
90
+ <port id="0" precision="I64">
91
+ <dim>1</dim>
92
+ </port>
93
+ </output>
94
+ </layer>
95
+ <layer id="9" name="Slice_2718657" type="Slice" version="opset8">
96
+ <input>
97
+ <port id="0" precision="I32">
98
+ <dim>9</dim>
99
+ </port>
100
+ <port id="1" precision="I32">
101
+ <dim>1</dim>
102
+ </port>
103
+ <port id="2" precision="I32">
104
+ <dim>1</dim>
105
+ </port>
106
+ <port id="3" precision="I32">
107
+ <dim>1</dim>
108
+ </port>
109
+ <port id="4" precision="I64">
110
+ <dim>1</dim>
111
+ </port>
112
+ </input>
113
+ <output>
114
+ <port id="5" precision="I32">
115
+ <dim>9</dim>
116
+ </port>
117
+ </output>
118
+ </layer>
119
+ <layer id="10" name="VocabDecoder_2718659" type="VocabDecoder" version="extension">
120
+ <data skip_tokens="" />
121
+ <input>
122
+ <port id="0" precision="I32">
123
+ <dim>-1</dim>
124
+ <dim>-1</dim>
125
+ </port>
126
+ <port id="1" precision="I32">
127
+ <dim>-1</dim>
128
+ </port>
129
+ <port id="2" precision="I32">
130
+ <dim>-1</dim>
131
+ </port>
132
+ <port id="3" precision="U8">
133
+ <dim>-1</dim>
134
+ </port>
135
+ <port id="4" precision="I32">
136
+ <dim>9</dim>
137
+ </port>
138
+ </input>
139
+ <output>
140
+ <port id="5" precision="I32">
141
+ <dim>-1</dim>
142
+ </port>
143
+ <port id="6" precision="I32">
144
+ <dim>-1</dim>
145
+ </port>
146
+ <port id="7" precision="I32">
147
+ <dim>-1</dim>
148
+ </port>
149
+ <port id="8" precision="I32">
150
+ <dim>-1</dim>
151
+ </port>
152
+ <port id="9" precision="U8">
153
+ <dim>-1</dim>
154
+ </port>
155
+ </output>
156
+ </layer>
157
+ <layer id="11" name="Constant_2718661" type="Const" version="opset1">
158
+ <data element_type="u8" shape="3" offset="3365018" size="3" />
159
+ <output>
160
+ <port id="0" precision="U8">
161
+ <dim>3</dim>
162
+ </port>
163
+ </output>
164
+ </layer>
165
+ <layer id="12" name="Constant_2718663" type="Const" version="opset1">
166
+ <data element_type="u8" shape="1" offset="3365021" size="1" />
167
+ <output>
168
+ <port id="0" precision="U8">
169
+ <dim>1</dim>
170
+ </port>
171
+ </output>
172
+ </layer>
173
+ <layer id="13" name="RegexNormalization_2718664" type="RegexNormalization" version="extension">
174
+ <data global_replace="true" />
175
+ <input>
176
+ <port id="0" precision="I32">
177
+ <dim>-1</dim>
178
+ </port>
179
+ <port id="1" precision="I32">
180
+ <dim>-1</dim>
181
+ </port>
182
+ <port id="2" precision="U8">
183
+ <dim>-1</dim>
184
+ </port>
185
+ <port id="3" precision="U8">
186
+ <dim>3</dim>
187
+ </port>
188
+ <port id="4" precision="U8">
189
+ <dim>1</dim>
190
+ </port>
191
+ </input>
192
+ <output>
193
+ <port id="5" precision="I32">
194
+ <dim>-1</dim>
195
+ </port>
196
+ <port id="6" precision="I32">
197
+ <dim>-1</dim>
198
+ </port>
199
+ <port id="7" precision="U8">
200
+ <dim>-1</dim>
201
+ </port>
202
+ </output>
203
+ </layer>
204
+ <layer id="14" name="ByteFallback_2718665" type="ByteFallback" version="extension">
205
+ <input>
206
+ <port id="0" precision="I32">
207
+ <dim>-1</dim>
208
+ </port>
209
+ <port id="1" precision="I32">
210
+ <dim>-1</dim>
211
+ </port>
212
+ <port id="2" precision="U8">
213
+ <dim>-1</dim>
214
+ </port>
215
+ </input>
216
+ <output>
217
+ <port id="3" precision="I32">
218
+ <dim>-1</dim>
219
+ </port>
220
+ <port id="4" precision="I32">
221
+ <dim>-1</dim>
222
+ </port>
223
+ <port id="5" precision="U8">
224
+ <dim>-1</dim>
225
+ </port>
226
+ </output>
227
+ </layer>
228
+ <layer id="15" name="FuzeRagged_2718666" type="FuzeRagged" version="extension">
229
+ <input>
230
+ <port id="0" precision="I32">
231
+ <dim>-1</dim>
232
+ </port>
233
+ <port id="1" precision="I32">
234
+ <dim>-1</dim>
235
+ </port>
236
+ <port id="2" precision="I32">
237
+ <dim>-1</dim>
238
+ </port>
239
+ <port id="3" precision="I32">
240
+ <dim>-1</dim>
241
+ </port>
242
+ </input>
243
+ <output>
244
+ <port id="4" precision="I32">
245
+ <dim>-1</dim>
246
+ </port>
247
+ <port id="5" precision="I32">
248
+ <dim>-1</dim>
249
+ </port>
250
+ </output>
251
+ </layer>
252
+ <layer id="16" name="UTF8Validate_2718667" type="UTF8Validate" version="extension">
253
+ <data replace_mode="true" />
254
+ <input>
255
+ <port id="0" precision="I32">
256
+ <dim>-1</dim>
257
+ </port>
258
+ <port id="1" precision="I32">
259
+ <dim>-1</dim>
260
+ </port>
261
+ <port id="2" precision="U8">
262
+ <dim>-1</dim>
263
+ </port>
264
+ </input>
265
+ <output>
266
+ <port id="3" precision="I32">
267
+ <dim>-1</dim>
268
+ </port>
269
+ <port id="4" precision="I32">
270
+ <dim>-1</dim>
271
+ </port>
272
+ <port id="5" precision="U8">
273
+ <dim>-1</dim>
274
+ </port>
275
+ </output>
276
+ </layer>
277
+ <layer id="17" name="StringTensorPack_2718668" type="StringTensorPack" version="extension">
278
+ <data mode="begins_ends" />
279
+ <input>
280
+ <port id="0" precision="I32">
281
+ <dim>-1</dim>
282
+ </port>
283
+ <port id="1" precision="I32">
284
+ <dim>-1</dim>
285
+ </port>
286
+ <port id="2" precision="U8">
287
+ <dim>-1</dim>
288
+ </port>
289
+ </input>
290
+ <output>
291
+ <port id="3" precision="STRING" names="string_output">
292
+ <dim>-1</dim>
293
+ </port>
294
+ </output>
295
+ </layer>
296
+ <layer id="18" name="Result_2718669" type="Result" version="opset1">
297
+ <input>
298
+ <port id="0" precision="STRING">
299
+ <dim>-1</dim>
300
+ </port>
301
+ </input>
302
+ </layer>
303
+ </layers>
304
+ <edges>
305
+ <edge from-layer="0" from-port="0" to-layer="1" to-port="0" />
306
+ <edge from-layer="1" from-port="1" to-layer="10" to-port="0" />
307
+ <edge from-layer="2" from-port="0" to-layer="3" to-port="0" />
308
+ <edge from-layer="3" from-port="2" to-layer="10" to-port="2" />
309
+ <edge from-layer="3" from-port="3" to-layer="10" to-port="3" />
310
+ <edge from-layer="3" from-port="1" to-layer="10" to-port="1" />
311
+ <edge from-layer="4" from-port="0" to-layer="9" to-port="0" />
312
+ <edge from-layer="5" from-port="0" to-layer="9" to-port="1" />
313
+ <edge from-layer="6" from-port="0" to-layer="9" to-port="2" />
314
+ <edge from-layer="7" from-port="0" to-layer="9" to-port="3" />
315
+ <edge from-layer="8" from-port="0" to-layer="9" to-port="4" />
316
+ <edge from-layer="9" from-port="5" to-layer="10" to-port="4" />
317
+ <edge from-layer="10" from-port="7" to-layer="13" to-port="0" />
318
+ <edge from-layer="10" from-port="8" to-layer="13" to-port="1" />
319
+ <edge from-layer="10" from-port="9" to-layer="13" to-port="2" />
320
+ <edge from-layer="10" from-port="6" to-layer="15" to-port="1" />
321
+ <edge from-layer="10" from-port="5" to-layer="15" to-port="0" />
322
+ <edge from-layer="11" from-port="0" to-layer="13" to-port="3" />
323
+ <edge from-layer="12" from-port="0" to-layer="13" to-port="4" />
324
+ <edge from-layer="13" from-port="6" to-layer="14" to-port="1" />
325
+ <edge from-layer="13" from-port="7" to-layer="14" to-port="2" />
326
+ <edge from-layer="13" from-port="5" to-layer="14" to-port="0" />
327
+ <edge from-layer="14" from-port="3" to-layer="15" to-port="2" />
328
+ <edge from-layer="14" from-port="4" to-layer="15" to-port="3" />
329
+ <edge from-layer="14" from-port="5" to-layer="16" to-port="2" />
330
+ <edge from-layer="15" from-port="4" to-layer="16" to-port="0" />
331
+ <edge from-layer="15" from-port="5" to-layer="16" to-port="1" />
332
+ <edge from-layer="16" from-port="3" to-layer="17" to-port="0" />
333
+ <edge from-layer="16" from-port="4" to-layer="17" to-port="1" />
334
+ <edge from-layer="16" from-port="5" to-layer="17" to-port="2" />
335
+ <edge from-layer="17" from-port="3" to-layer="18" to-port="0" />
336
+ </edges>
337
+ <rt_info>
338
+ <add_attention_mask value="True" />
339
+ <add_prefix_space />
340
+ <add_special_tokens value="True" />
341
+ <bos_token_id value="2" />
342
+ <chat_template value="{{ bos_token }} &#10;{%- if messages[0]['role'] == 'system' -%}&#10; {%- if messages[0]['content'] is string -%}&#10; {%- set first_user_prefix = messages[0]['content'] + '&#10;&#10;' -%}&#10; {%- else -%}&#10; {%- set first_user_prefix = messages[0]['content'][0]['text'] + '&#10;&#10;' -%}&#10; {%- endif -%}&#10; {%- set loop_messages = messages[1:] -%}&#10;{%- else -%}&#10; {%- set first_user_prefix = &quot;&quot; -%}&#10; {%- set loop_messages = messages -%}&#10;{%- endif -%}&#10;{%- for message in loop_messages -%}&#10; {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}&#10; {{ raise_exception(&quot;Conversation roles must alternate user/assistant/user/assistant/...&quot;) }}&#10; {%- endif -%}&#10; {%- if (message['role'] == 'assistant') -%}&#10; {%- set role = &quot;model&quot; -%}&#10; {%- else -%}&#10; {%- set role = message['role'] -%}&#10; {%- endif -%}&#10; {{ '&lt;start_of_turn>' + role + '&#10;' + (first_user_prefix if loop.first else &quot;&quot;) }}&#10; {%- if message['content'] is string -%}&#10; {{ message['content'] | trim }}&#10; {%- elif message['content'] is iterable -%}&#10; {%- for item in message['content'] -%}&#10; {%- if item['type'] == 'image' -%}&#10; {{ '&lt;start_of_image>' }}&#10; {%- elif item['type'] == 'text' -%}&#10; {{ item['text'] | trim }}&#10; {%- endif -%}&#10; {%- endfor -%}&#10; {%- else -%}&#10; {{ raise_exception(&quot;Invalid content type&quot;) }}&#10; {%- endif -%}&#10; {{ '&lt;end_of_turn>&#10;' }}&#10;{%- endfor -%}&#10;{%- if add_generation_prompt -%}&#10; {{'&lt;start_of_turn>model&#10;'}}&#10;{%- endif -%}&#10;" />
343
+ <clean_up_tokenization_spaces />
344
+ <detokenizer_input_type value="i64" />
345
+ <eos_token_id value="1" />
346
+ <handle_special_tokens_with_re />
347
+ <number_of_inputs value="1" />
348
+ <openvino_tokenizers_version value="2025.0.0.0" />
349
+ <openvino_version value="2025.0.0" />
350
+ <original_tokenizer_class value="&lt;class 'transformers.models.gemma.tokenization_gemma_fast.GemmaTokenizerFast'>" />
351
+ <pad_token_id value="0" />
352
+ <sentencepiece_version value="0.2.0" />
353
+ <skip_special_tokens value="True" />
354
+ <streaming_detokenizer value="False" />
355
+ <tiktoken_version value="0.7.0" />
356
+ <tokenizer_output_type value="i64" />
357
+ <tokenizers_version value="0.21.1" />
358
+ <transformers_version value="4.50.2" />
359
+ <use_max_padding value="False" />
360
+ <use_sentencepiece_backend value="False" />
361
+ <utf8_replace_mode value="replace" />
362
+ <with_detokenizer value="True" />
363
+ </rt_info>
364
+ </net>
openvino_language_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d44ecfd0c308f98f18f19363c73e73af646df05008c5600c10e660c200ecbded
3
+ size 11775828438
openvino_language_model.xml ADDED
The diff for this file is too large to render. See raw diff
 
openvino_text_embeddings_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d56e1b338ac6ecd8339c867032a07b6559717aabdaf3c1278a4ee6fc2003f9de
3
+ size 1007403144
openvino_text_embeddings_model.xml ADDED
@@ -0,0 +1,207 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <?xml version="1.0"?>
2
+ <net name="Model6" version="11">
3
+ <layers>
4
+ <layer id="0" name="input" type="Parameter" version="opset1">
5
+ <data shape="?,?" element_type="i64" />
6
+ <output>
7
+ <port id="0" precision="I64" names="input">
8
+ <dim>-1</dim>
9
+ <dim>-1</dim>
10
+ </port>
11
+ </output>
12
+ </layer>
13
+ <layer id="1" name="self.weight" type="Const" version="opset1">
14
+ <data element_type="i8" shape="262208, 3840" offset="0" size="1006878720" />
15
+ <output>
16
+ <port id="0" precision="I8">
17
+ <dim>262208</dim>
18
+ <dim>3840</dim>
19
+ </port>
20
+ </output>
21
+ </layer>
22
+ <layer id="2" name="Convert_1923719" type="Convert" version="opset1">
23
+ <data destination_type="f16" />
24
+ <input>
25
+ <port id="0" precision="I8">
26
+ <dim>262208</dim>
27
+ <dim>3840</dim>
28
+ </port>
29
+ </input>
30
+ <output>
31
+ <port id="1" precision="FP16">
32
+ <dim>262208</dim>
33
+ <dim>3840</dim>
34
+ </port>
35
+ </output>
36
+ </layer>
37
+ <layer id="3" name="self.weight/scale" type="Const" version="opset1">
38
+ <data element_type="f16" shape="262208, 1" offset="1006878720" size="524416" />
39
+ <output>
40
+ <port id="0" precision="FP16">
41
+ <dim>262208</dim>
42
+ <dim>1</dim>
43
+ </port>
44
+ </output>
45
+ </layer>
46
+ <layer id="4" name="self.weight/fq_weights_0" type="Multiply" version="opset1">
47
+ <data auto_broadcast="numpy" />
48
+ <input>
49
+ <port id="0" precision="FP16">
50
+ <dim>262208</dim>
51
+ <dim>3840</dim>
52
+ </port>
53
+ <port id="1" precision="FP16">
54
+ <dim>262208</dim>
55
+ <dim>1</dim>
56
+ </port>
57
+ </input>
58
+ <output>
59
+ <port id="2" precision="FP16">
60
+ <dim>262208</dim>
61
+ <dim>3840</dim>
62
+ </port>
63
+ </output>
64
+ </layer>
65
+ <layer id="5" name="self.weight/fq_weights_0/convert" type="Convert" version="opset1">
66
+ <data destination_type="f32" />
67
+ <input>
68
+ <port id="0" precision="FP16">
69
+ <dim>262208</dim>
70
+ <dim>3840</dim>
71
+ </port>
72
+ </input>
73
+ <output>
74
+ <port id="1" precision="FP32">
75
+ <dim>262208</dim>
76
+ <dim>3840</dim>
77
+ </port>
78
+ </output>
79
+ </layer>
80
+ <layer id="6" name="aten::embedding/Convert" type="Convert" version="opset1">
81
+ <data destination_type="i32" />
82
+ <input>
83
+ <port id="0" precision="I64">
84
+ <dim>-1</dim>
85
+ <dim>-1</dim>
86
+ </port>
87
+ </input>
88
+ <output>
89
+ <port id="1" precision="I32">
90
+ <dim>-1</dim>
91
+ <dim>-1</dim>
92
+ </port>
93
+ </output>
94
+ </layer>
95
+ <layer id="7" name="aten::embedding/Constant" type="Const" version="opset1">
96
+ <data element_type="i32" shape="" offset="1007403136" size="4" />
97
+ <output>
98
+ <port id="0" precision="I32" />
99
+ </output>
100
+ </layer>
101
+ <layer id="8" name="aten::embedding/Gather" type="Gather" version="opset8">
102
+ <data batch_dims="0" />
103
+ <input>
104
+ <port id="0" precision="FP32">
105
+ <dim>262208</dim>
106
+ <dim>3840</dim>
107
+ </port>
108
+ <port id="1" precision="I32">
109
+ <dim>-1</dim>
110
+ <dim>-1</dim>
111
+ </port>
112
+ <port id="2" precision="I32" />
113
+ </input>
114
+ <output>
115
+ <port id="3" precision="FP32" names="6">
116
+ <dim>-1</dim>
117
+ <dim>-1</dim>
118
+ <dim>3840</dim>
119
+ </port>
120
+ </output>
121
+ </layer>
122
+ <layer id="9" name="Constant_229469" type="Const" version="opset1">
123
+ <data element_type="f32" shape="1, 1, 1" offset="1007403140" size="4" />
124
+ <output>
125
+ <port id="0" precision="FP32">
126
+ <dim>1</dim>
127
+ <dim>1</dim>
128
+ <dim>1</dim>
129
+ </port>
130
+ </output>
131
+ </layer>
132
+ <layer id="10" name="aten::mul/Multiply" type="Multiply" version="opset1">
133
+ <data auto_broadcast="numpy" />
134
+ <input>
135
+ <port id="0" precision="FP32">
136
+ <dim>-1</dim>
137
+ <dim>-1</dim>
138
+ <dim>3840</dim>
139
+ </port>
140
+ <port id="1" precision="FP32">
141
+ <dim>1</dim>
142
+ <dim>1</dim>
143
+ <dim>1</dim>
144
+ </port>
145
+ </input>
146
+ <output>
147
+ <port id="2" precision="FP32" names="inputs_embeds">
148
+ <dim>-1</dim>
149
+ <dim>-1</dim>
150
+ <dim>3840</dim>
151
+ </port>
152
+ </output>
153
+ </layer>
154
+ <layer id="11" name="Result_227823" type="Result" version="opset1">
155
+ <input>
156
+ <port id="0" precision="FP32">
157
+ <dim>-1</dim>
158
+ <dim>-1</dim>
159
+ <dim>3840</dim>
160
+ </port>
161
+ </input>
162
+ </layer>
163
+ </layers>
164
+ <edges>
165
+ <edge from-layer="0" from-port="0" to-layer="6" to-port="0" />
166
+ <edge from-layer="1" from-port="0" to-layer="2" to-port="0" />
167
+ <edge from-layer="2" from-port="1" to-layer="4" to-port="0" />
168
+ <edge from-layer="3" from-port="0" to-layer="4" to-port="1" />
169
+ <edge from-layer="4" from-port="2" to-layer="5" to-port="0" />
170
+ <edge from-layer="5" from-port="1" to-layer="8" to-port="0" />
171
+ <edge from-layer="6" from-port="1" to-layer="8" to-port="1" />
172
+ <edge from-layer="7" from-port="0" to-layer="8" to-port="2" />
173
+ <edge from-layer="8" from-port="3" to-layer="10" to-port="0" />
174
+ <edge from-layer="9" from-port="0" to-layer="10" to-port="1" />
175
+ <edge from-layer="10" from-port="2" to-layer="11" to-port="0" />
176
+ </edges>
177
+ <rt_info>
178
+ <Runtime_version value="2025.0.0-17942-1f68be9f594-releases/2025/0" />
179
+ <conversion_parameters>
180
+ <framework value="pytorch" />
181
+ <is_python_object value="True" />
182
+ </conversion_parameters>
183
+ <nncf>
184
+ <friendly_names_were_updated value="True" />
185
+ <weight_compression>
186
+ <advanced_parameters value="{'statistics_path': None, 'awq_params': {'subset_size': 32, 'percent_to_apply': 0.002, 'alpha_min': 0.0, 'alpha_max': 1.0, 'steps': 100}, 'scale_estimation_params': {'subset_size': 64, 'initial_steps': 5, 'scale_steps': 5, 'weight_penalty': -1.0}, 'gptq_params': {'damp_percent': 0.1, 'block_size': 128, 'subset_size': 128}, 'lora_correction_params': {'adapter_rank': 8, 'num_iterations': 3, 'apply_regularization': True, 'subset_size': 128, 'use_int8_adapters': True}}" />
187
+ <all_layers value="False" />
188
+ <awq value="False" />
189
+ <backup_mode value="int8_asym" />
190
+ <gptq value="False" />
191
+ <group_size value="-1" />
192
+ <ignored_scope value="[]" />
193
+ <lora_correction value="False" />
194
+ <mode value="int8_sym" />
195
+ <ratio value="1.0" />
196
+ <scale_estimation value="False" />
197
+ <sensitivity_metric value="weight_quantization_error" />
198
+ </weight_compression>
199
+ </nncf>
200
+ <optimum>
201
+ <optimum_intel_version value="1.23.0.dev0+a20051d" />
202
+ <optimum_version value="1.25.0.dev0" />
203
+ <pytorch_version value="2.5.1" />
204
+ <transformers_version value="4.50.2" />
205
+ </optimum>
206
+ </rt_info>
207
+ </net>
openvino_tokenizer.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5abb96c6a6d10be7fa0849ff8b3b416e555cc48c7a2041b6572c52491f932162
3
+ size 12340138
openvino_tokenizer.xml ADDED
@@ -0,0 +1,837 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <?xml version="1.0"?>
2
+ <net name="tokenizer" version="11">
3
+ <layers>
4
+ <layer id="0" name="Parameter_2718533" type="Parameter" version="opset1">
5
+ <data shape="?" element_type="string" />
6
+ <output>
7
+ <port id="0" precision="STRING" names="Parameter_2718533">
8
+ <dim>-1</dim>
9
+ </port>
10
+ </output>
11
+ </layer>
12
+ <layer id="1" name="Constant_2718636" type="Const" version="opset1">
13
+ <data element_type="i32" shape="" offset="0" size="4" />
14
+ <output>
15
+ <port id="0" precision="I32" />
16
+ </output>
17
+ </layer>
18
+ <layer id="2" name="Constant_2718637" type="Const" version="opset1">
19
+ <data element_type="i32" shape="" offset="4" size="4" />
20
+ <output>
21
+ <port id="0" precision="I32" />
22
+ </output>
23
+ </layer>
24
+ <layer id="3" name="Constant_2718638" type="Const" version="opset1">
25
+ <data element_type="i32" shape="1" offset="8" size="4" />
26
+ <output>
27
+ <port id="0" precision="I32">
28
+ <dim>1</dim>
29
+ </port>
30
+ </output>
31
+ </layer>
32
+ <layer id="4" name="Constant_2718539" type="Const" version="opset1">
33
+ <data element_type="i64" shape="" offset="12" size="8" />
34
+ <output>
35
+ <port id="0" precision="I64" />
36
+ </output>
37
+ </layer>
38
+ <layer id="5" name="StringTensorUnpack_2718534" type="StringTensorUnpack" version="extension">
39
+ <data mode="begins_ends" />
40
+ <input>
41
+ <port id="0" precision="STRING">
42
+ <dim>-1</dim>
43
+ </port>
44
+ </input>
45
+ <output>
46
+ <port id="1" precision="I32">
47
+ <dim>-1</dim>
48
+ </port>
49
+ <port id="2" precision="I32">
50
+ <dim>-1</dim>
51
+ </port>
52
+ <port id="3" precision="U8">
53
+ <dim>-1</dim>
54
+ </port>
55
+ </output>
56
+ </layer>
57
+ <layer id="6" name="ShapeOf_2718535" type="ShapeOf" version="opset3">
58
+ <data output_type="i64" />
59
+ <input>
60
+ <port id="0" precision="I32">
61
+ <dim>-1</dim>
62
+ </port>
63
+ </input>
64
+ <output>
65
+ <port id="1" precision="I64">
66
+ <dim>1</dim>
67
+ </port>
68
+ </output>
69
+ </layer>
70
+ <layer id="7" name="Constant_2718536" type="Const" version="opset1">
71
+ <data element_type="i64" shape="" offset="12" size="8" />
72
+ <output>
73
+ <port id="0" precision="I64" />
74
+ </output>
75
+ </layer>
76
+ <layer id="8" name="Constant_2718537" type="Const" version="opset1">
77
+ <data element_type="i64" shape="" offset="12" size="8" />
78
+ <output>
79
+ <port id="0" precision="I64" />
80
+ </output>
81
+ </layer>
82
+ <layer id="9" name="Gather_2718538" type="Gather" version="opset8">
83
+ <data batch_dims="0" />
84
+ <input>
85
+ <port id="0" precision="I64">
86
+ <dim>1</dim>
87
+ </port>
88
+ <port id="1" precision="I64" />
89
+ <port id="2" precision="I64" />
90
+ </input>
91
+ <output>
92
+ <port id="3" precision="I64" />
93
+ </output>
94
+ </layer>
95
+ <layer id="10" name="Constant_2718540" type="Const" version="opset1">
96
+ <data element_type="i64" shape="" offset="20" size="8" />
97
+ <output>
98
+ <port id="0" precision="I64" />
99
+ </output>
100
+ </layer>
101
+ <layer id="11" name="Range_2718541" type="Range" version="opset4">
102
+ <data output_type="i32" />
103
+ <input>
104
+ <port id="0" precision="I64" />
105
+ <port id="1" precision="I64" />
106
+ <port id="2" precision="I64" />
107
+ </input>
108
+ <output>
109
+ <port id="3" precision="I32">
110
+ <dim>-1</dim>
111
+ </port>
112
+ </output>
113
+ </layer>
114
+ <layer id="12" name="Constant_2718542" type="Const" version="opset1">
115
+ <data element_type="i64" shape="" offset="20" size="8" />
116
+ <output>
117
+ <port id="0" precision="I64" />
118
+ </output>
119
+ </layer>
120
+ <layer id="13" name="Constant_2718543" type="Const" version="opset1">
121
+ <data element_type="i64" shape="" offset="20" size="8" />
122
+ <output>
123
+ <port id="0" precision="I64" />
124
+ </output>
125
+ </layer>
126
+ <layer id="14" name="Add_2718544" type="Add" version="opset1">
127
+ <data auto_broadcast="numpy" />
128
+ <input>
129
+ <port id="0" precision="I64" />
130
+ <port id="1" precision="I64" />
131
+ </input>
132
+ <output>
133
+ <port id="2" precision="I64" />
134
+ </output>
135
+ </layer>
136
+ <layer id="15" name="Constant_2718545" type="Const" version="opset1">
137
+ <data element_type="i64" shape="" offset="20" size="8" />
138
+ <output>
139
+ <port id="0" precision="I64" />
140
+ </output>
141
+ </layer>
142
+ <layer id="16" name="Range_2718546" type="Range" version="opset4">
143
+ <data output_type="i32" />
144
+ <input>
145
+ <port id="0" precision="I64" />
146
+ <port id="1" precision="I64" />
147
+ <port id="2" precision="I64" />
148
+ </input>
149
+ <output>
150
+ <port id="3" precision="I32">
151
+ <dim>-1</dim>
152
+ </port>
153
+ </output>
154
+ </layer>
155
+ <layer id="17" name="Constant_2718608" type="Const" version="opset1">
156
+ <data element_type="u8" shape="110204" offset="28" size="110204" />
157
+ <output>
158
+ <port id="0" precision="U8">
159
+ <dim>110204</dim>
160
+ </port>
161
+ </output>
162
+ </layer>
163
+ <layer id="18" name="SpecialTokensSplit_2718609" type="SpecialTokensSplit" version="extension">
164
+ <input>
165
+ <port id="0" precision="I32">
166
+ <dim>-1</dim>
167
+ </port>
168
+ <port id="1" precision="I32">
169
+ <dim>-1</dim>
170
+ </port>
171
+ <port id="2" precision="I32">
172
+ <dim>-1</dim>
173
+ </port>
174
+ <port id="3" precision="I32">
175
+ <dim>-1</dim>
176
+ </port>
177
+ <port id="4" precision="U8">
178
+ <dim>-1</dim>
179
+ </port>
180
+ <port id="5" precision="U8">
181
+ <dim>110204</dim>
182
+ </port>
183
+ </input>
184
+ <output>
185
+ <port id="6" precision="I32">
186
+ <dim>-1</dim>
187
+ </port>
188
+ <port id="7" precision="I32">
189
+ <dim>-1</dim>
190
+ </port>
191
+ <port id="8" precision="I32">
192
+ <dim>-1</dim>
193
+ </port>
194
+ <port id="9" precision="I32">
195
+ <dim>-1</dim>
196
+ </port>
197
+ <port id="10" precision="U8">
198
+ <dim>-1</dim>
199
+ </port>
200
+ <port id="11" precision="BOOL">
201
+ <dim>-1</dim>
202
+ </port>
203
+ </output>
204
+ </layer>
205
+ <layer id="19" name="Constant_2718611" type="Const" version="opset1">
206
+ <data element_type="u8" shape="1" offset="110232" size="1" />
207
+ <output>
208
+ <port id="0" precision="U8">
209
+ <dim>1</dim>
210
+ </port>
211
+ </output>
212
+ </layer>
213
+ <layer id="20" name="Constant_2718613" type="Const" version="opset1">
214
+ <data element_type="u8" shape="3" offset="110233" size="3" />
215
+ <output>
216
+ <port id="0" precision="U8">
217
+ <dim>3</dim>
218
+ </port>
219
+ </output>
220
+ </layer>
221
+ <layer id="21" name="RegexNormalization_2718614" type="RegexNormalization" version="extension">
222
+ <data global_replace="true" />
223
+ <input>
224
+ <port id="0" precision="I32">
225
+ <dim>-1</dim>
226
+ </port>
227
+ <port id="1" precision="I32">
228
+ <dim>-1</dim>
229
+ </port>
230
+ <port id="2" precision="U8">
231
+ <dim>-1</dim>
232
+ </port>
233
+ <port id="3" precision="BOOL">
234
+ <dim>-1</dim>
235
+ </port>
236
+ <port id="4" precision="U8">
237
+ <dim>1</dim>
238
+ </port>
239
+ <port id="5" precision="U8">
240
+ <dim>3</dim>
241
+ </port>
242
+ </input>
243
+ <output>
244
+ <port id="6" precision="I32">
245
+ <dim>-1</dim>
246
+ </port>
247
+ <port id="7" precision="I32">
248
+ <dim>-1</dim>
249
+ </port>
250
+ <port id="8" precision="U8">
251
+ <dim>-1</dim>
252
+ </port>
253
+ <port id="9" precision="BOOL">
254
+ <dim>-1</dim>
255
+ </port>
256
+ </output>
257
+ </layer>
258
+ <layer id="22" name="Constant_2718616" type="Const" version="opset1">
259
+ <data element_type="u8" shape="1" offset="110232" size="1" />
260
+ <output>
261
+ <port id="0" precision="U8">
262
+ <dim>1</dim>
263
+ </port>
264
+ </output>
265
+ </layer>
266
+ <layer id="23" name="RegexSplit_2718617" type="RegexSplit" version="extension">
267
+ <data behaviour="mergedwithprevious" invert="false" max_splits="-1" />
268
+ <input>
269
+ <port id="0" precision="I32">
270
+ <dim>-1</dim>
271
+ </port>
272
+ <port id="1" precision="I32">
273
+ <dim>-1</dim>
274
+ </port>
275
+ <port id="2" precision="I32">
276
+ <dim>-1</dim>
277
+ </port>
278
+ <port id="3" precision="I32">
279
+ <dim>-1</dim>
280
+ </port>
281
+ <port id="4" precision="U8">
282
+ <dim>-1</dim>
283
+ </port>
284
+ <port id="5" precision="BOOL">
285
+ <dim>-1</dim>
286
+ </port>
287
+ <port id="6" precision="U8">
288
+ <dim>1</dim>
289
+ </port>
290
+ </input>
291
+ <output>
292
+ <port id="7" precision="I32">
293
+ <dim>-1</dim>
294
+ </port>
295
+ <port id="8" precision="I32">
296
+ <dim>-1</dim>
297
+ </port>
298
+ <port id="9" precision="I32">
299
+ <dim>-1</dim>
300
+ </port>
301
+ <port id="10" precision="I32">
302
+ <dim>-1</dim>
303
+ </port>
304
+ <port id="11" precision="U8">
305
+ <dim>-1</dim>
306
+ </port>
307
+ <port id="12" precision="BOOL">
308
+ <dim>-1</dim>
309
+ </port>
310
+ </output>
311
+ </layer>
312
+ <layer id="24" name="Constant_2718619" type="Const" version="opset1">
313
+ <data element_type="u8" shape="3364962" offset="110236" size="3364962" />
314
+ <output>
315
+ <port id="0" precision="U8">
316
+ <dim>3364962</dim>
317
+ </port>
318
+ </output>
319
+ </layer>
320
+ <layer id="25" name="StringTensorUnpack_2718620" type="StringTensorUnpack" version="extension">
321
+ <data mode="begins_ends" />
322
+ <input>
323
+ <port id="0" precision="U8">
324
+ <dim>3364962</dim>
325
+ </port>
326
+ </input>
327
+ <output>
328
+ <port id="1" precision="I32">
329
+ <dim>-1</dim>
330
+ </port>
331
+ <port id="2" precision="I32">
332
+ <dim>-1</dim>
333
+ </port>
334
+ <port id="3" precision="U8">
335
+ <dim>-1</dim>
336
+ </port>
337
+ </output>
338
+ </layer>
339
+ <layer id="26" name="Constant_2718625" type="Const" version="opset1">
340
+ <data element_type="u8" shape="4773972" offset="3475198" size="4773972" />
341
+ <output>
342
+ <port id="0" precision="U8">
343
+ <dim>4773972</dim>
344
+ </port>
345
+ </output>
346
+ </layer>
347
+ <layer id="27" name="StringTensorUnpack_2718626" type="StringTensorUnpack" version="extension">
348
+ <data mode="begins_ends" />
349
+ <input>
350
+ <port id="0" precision="U8">
351
+ <dim>4773972</dim>
352
+ </port>
353
+ </input>
354
+ <output>
355
+ <port id="1" precision="I32">
356
+ <dim>-1</dim>
357
+ </port>
358
+ <port id="2" precision="I32">
359
+ <dim>-1</dim>
360
+ </port>
361
+ <port id="3" precision="U8">
362
+ <dim>-1</dim>
363
+ </port>
364
+ </output>
365
+ </layer>
366
+ <layer id="28" name="Constant_2718628" type="Const" version="opset1">
367
+ <data element_type="u8" shape="3962857" offset="8249170" size="3962857" />
368
+ <output>
369
+ <port id="0" precision="U8">
370
+ <dim>3962857</dim>
371
+ </port>
372
+ </output>
373
+ </layer>
374
+ <layer id="29" name="StringTensorUnpack_2718629" type="StringTensorUnpack" version="extension">
375
+ <data mode="begins_ends" />
376
+ <input>
377
+ <port id="0" precision="U8">
378
+ <dim>3962857</dim>
379
+ </port>
380
+ </input>
381
+ <output>
382
+ <port id="1" precision="I32">
383
+ <dim>-1</dim>
384
+ </port>
385
+ <port id="2" precision="I32">
386
+ <dim>-1</dim>
387
+ </port>
388
+ <port id="3" precision="U8">
389
+ <dim>-1</dim>
390
+ </port>
391
+ </output>
392
+ </layer>
393
+ <layer id="30" name="Constant_2718622" type="Const" version="opset1">
394
+ <data element_type="u8" shape="102451" offset="12212027" size="102451" />
395
+ <output>
396
+ <port id="0" precision="U8">
397
+ <dim>102451</dim>
398
+ </port>
399
+ </output>
400
+ </layer>
401
+ <layer id="31" name="StringTensorUnpack_2718623" type="StringTensorUnpack" version="extension">
402
+ <data mode="begins_ends" />
403
+ <input>
404
+ <port id="0" precision="U8">
405
+ <dim>102451</dim>
406
+ </port>
407
+ </input>
408
+ <output>
409
+ <port id="1" precision="I32">
410
+ <dim>-1</dim>
411
+ </port>
412
+ <port id="2" precision="I32">
413
+ <dim>-1</dim>
414
+ </port>
415
+ <port id="3" precision="U8">
416
+ <dim>-1</dim>
417
+ </port>
418
+ </output>
419
+ </layer>
420
+ <layer id="32" name="Constant_2718630" type="Const" version="opset1">
421
+ <data element_type="i32" shape="6414" offset="12314478" size="25656" />
422
+ <output>
423
+ <port id="0" precision="I32">
424
+ <dim>6414</dim>
425
+ </port>
426
+ </output>
427
+ </layer>
428
+ <layer id="33" name="BPETokenizer_2718631" type="BPETokenizer" version="extension">
429
+ <data unk_token="&lt;unk>" fuse_unk="true" suffix_indicator="" end_suffix="" byte_fallback="true" cache_capacity="52428" />
430
+ <input>
431
+ <port id="0" precision="I32">
432
+ <dim>-1</dim>
433
+ </port>
434
+ <port id="1" precision="I32">
435
+ <dim>-1</dim>
436
+ </port>
437
+ <port id="2" precision="I32">
438
+ <dim>-1</dim>
439
+ </port>
440
+ <port id="3" precision="I32">
441
+ <dim>-1</dim>
442
+ </port>
443
+ <port id="4" precision="U8">
444
+ <dim>-1</dim>
445
+ </port>
446
+ <port id="5" precision="I32">
447
+ <dim>-1</dim>
448
+ </port>
449
+ <port id="6" precision="I32">
450
+ <dim>-1</dim>
451
+ </port>
452
+ <port id="7" precision="U8">
453
+ <dim>-1</dim>
454
+ </port>
455
+ <port id="8" precision="I32">
456
+ <dim>-1</dim>
457
+ </port>
458
+ <port id="9" precision="I32">
459
+ <dim>-1</dim>
460
+ </port>
461
+ <port id="10" precision="U8">
462
+ <dim>-1</dim>
463
+ </port>
464
+ <port id="11" precision="I32">
465
+ <dim>-1</dim>
466
+ </port>
467
+ <port id="12" precision="I32">
468
+ <dim>-1</dim>
469
+ </port>
470
+ <port id="13" precision="U8">
471
+ <dim>-1</dim>
472
+ </port>
473
+ <port id="14" precision="I32">
474
+ <dim>-1</dim>
475
+ </port>
476
+ <port id="15" precision="I32">
477
+ <dim>-1</dim>
478
+ </port>
479
+ <port id="16" precision="U8">
480
+ <dim>-1</dim>
481
+ </port>
482
+ <port id="17" precision="I32">
483
+ <dim>6414</dim>
484
+ </port>
485
+ </input>
486
+ <output>
487
+ <port id="18" precision="I32">
488
+ <dim>-1</dim>
489
+ </port>
490
+ <port id="19" precision="I32">
491
+ <dim>-1</dim>
492
+ </port>
493
+ <port id="20" precision="I32">
494
+ <dim>-1</dim>
495
+ </port>
496
+ </output>
497
+ </layer>
498
+ <layer id="34" name="Subtract_2718632" type="Subtract" version="opset1">
499
+ <data auto_broadcast="numpy" />
500
+ <input>
501
+ <port id="0" precision="I32">
502
+ <dim>-1</dim>
503
+ </port>
504
+ <port id="1" precision="I32">
505
+ <dim>-1</dim>
506
+ </port>
507
+ </input>
508
+ <output>
509
+ <port id="2" precision="I32">
510
+ <dim>-1</dim>
511
+ </port>
512
+ </output>
513
+ </layer>
514
+ <layer id="35" name="Constant_2718633" type="Const" version="opset1">
515
+ <data element_type="i32" shape="" offset="12340134" size="4" />
516
+ <output>
517
+ <port id="0" precision="I32" />
518
+ </output>
519
+ </layer>
520
+ <layer id="36" name="Minimum_2718634" type="Minimum" version="opset1">
521
+ <data auto_broadcast="numpy" />
522
+ <input>
523
+ <port id="0" precision="I32">
524
+ <dim>-1</dim>
525
+ </port>
526
+ <port id="1" precision="I32" />
527
+ </input>
528
+ <output>
529
+ <port id="2" precision="I32">
530
+ <dim>-1</dim>
531
+ </port>
532
+ </output>
533
+ </layer>
534
+ <layer id="37" name="Subtract_2718635" type="Subtract" version="opset1">
535
+ <data auto_broadcast="numpy" />
536
+ <input>
537
+ <port id="0" precision="I32">
538
+ <dim>-1</dim>
539
+ </port>
540
+ <port id="1" precision="I32">
541
+ <dim>-1</dim>
542
+ </port>
543
+ </input>
544
+ <output>
545
+ <port id="2" precision="I32">
546
+ <dim>-1</dim>
547
+ </port>
548
+ </output>
549
+ </layer>
550
+ <layer id="38" name="Constant_2718639" type="Const" version="opset1">
551
+ <data element_type="i32" shape="2" offset="12" size="8" />
552
+ <output>
553
+ <port id="0" precision="I32">
554
+ <dim>2</dim>
555
+ </port>
556
+ </output>
557
+ </layer>
558
+ <layer id="39" name="CombineSegments_2718640" type="CombineSegments" version="extension">
559
+ <input>
560
+ <port id="0" precision="I32" />
561
+ <port id="1" precision="I32" />
562
+ <port id="2" precision="I32">
563
+ <dim>1</dim>
564
+ </port>
565
+ <port id="3" precision="I32">
566
+ <dim>-1</dim>
567
+ </port>
568
+ <port id="4" precision="I32">
569
+ <dim>-1</dim>
570
+ </port>
571
+ <port id="5" precision="I32">
572
+ <dim>-1</dim>
573
+ </port>
574
+ <port id="6" precision="I32">
575
+ <dim>2</dim>
576
+ </port>
577
+ </input>
578
+ <output>
579
+ <port id="7" precision="I32">
580
+ <dim>-1</dim>
581
+ </port>
582
+ <port id="8" precision="I32">
583
+ <dim>-1</dim>
584
+ </port>
585
+ <port id="9" precision="I32">
586
+ <dim>-1</dim>
587
+ </port>
588
+ <port id="10" precision="I32">
589
+ <dim>-1</dim>
590
+ </port>
591
+ <port id="11" precision="I32">
592
+ <dim>-1</dim>
593
+ </port>
594
+ <port id="12" precision="I32">
595
+ <dim>-1</dim>
596
+ </port>
597
+ </output>
598
+ </layer>
599
+ <layer id="40" name="Subtract_2718641" type="Subtract" version="opset1">
600
+ <data auto_broadcast="numpy" />
601
+ <input>
602
+ <port id="0" precision="I32">
603
+ <dim>-1</dim>
604
+ </port>
605
+ <port id="1" precision="I32">
606
+ <dim>-1</dim>
607
+ </port>
608
+ </input>
609
+ <output>
610
+ <port id="2" precision="I32">
611
+ <dim>-1</dim>
612
+ </port>
613
+ </output>
614
+ </layer>
615
+ <layer id="41" name="Constant_2718642" type="Const" version="opset1">
616
+ <data element_type="i32" shape="" offset="0" size="4" />
617
+ <output>
618
+ <port id="0" precision="I32" />
619
+ </output>
620
+ </layer>
621
+ <layer id="42" name="ReduceMax_2718643" type="ReduceMax" version="opset1">
622
+ <data keep_dims="false" />
623
+ <input>
624
+ <port id="0" precision="I32">
625
+ <dim>-1</dim>
626
+ </port>
627
+ <port id="1" precision="I32" />
628
+ </input>
629
+ <output>
630
+ <port id="2" precision="I32" />
631
+ </output>
632
+ </layer>
633
+ <layer id="43" name="Constant_2718644" type="Const" version="opset1">
634
+ <data element_type="i32" shape="" offset="0" size="4" />
635
+ <output>
636
+ <port id="0" precision="I32" />
637
+ </output>
638
+ </layer>
639
+ <layer id="44" name="RaggedToDense_2718645" type="RaggedToDense" version="extension">
640
+ <data pad_right="false" />
641
+ <input>
642
+ <port id="0" precision="I32">
643
+ <dim>-1</dim>
644
+ </port>
645
+ <port id="1" precision="I32">
646
+ <dim>-1</dim>
647
+ </port>
648
+ <port id="2" precision="I32">
649
+ <dim>-1</dim>
650
+ </port>
651
+ <port id="3" precision="I32" />
652
+ <port id="4" precision="I32" />
653
+ </input>
654
+ <output>
655
+ <port id="5" precision="I32">
656
+ <dim>-1</dim>
657
+ <dim>-1</dim>
658
+ </port>
659
+ <port id="6" precision="BOOL">
660
+ <dim>-1</dim>
661
+ <dim>-1</dim>
662
+ </port>
663
+ </output>
664
+ </layer>
665
+ <layer id="45" name="Convert_2718646" type="Convert" version="opset1">
666
+ <data destination_type="i32" />
667
+ <input>
668
+ <port id="0" precision="BOOL">
669
+ <dim>-1</dim>
670
+ <dim>-1</dim>
671
+ </port>
672
+ </input>
673
+ <output>
674
+ <port id="1" precision="I32">
675
+ <dim>-1</dim>
676
+ <dim>-1</dim>
677
+ </port>
678
+ </output>
679
+ </layer>
680
+ <layer id="46" name="Convert_2718646.0" type="Convert" version="opset1">
681
+ <data destination_type="i64" />
682
+ <input>
683
+ <port id="0" precision="I32">
684
+ <dim>-1</dim>
685
+ <dim>-1</dim>
686
+ </port>
687
+ </input>
688
+ <output>
689
+ <port id="1" precision="I64" names="attention_mask">
690
+ <dim>-1</dim>
691
+ <dim>-1</dim>
692
+ </port>
693
+ </output>
694
+ </layer>
695
+ <layer id="48" name="RaggedToDense_2718645.0" type="Convert" version="opset1">
696
+ <data destination_type="i64" />
697
+ <input>
698
+ <port id="0" precision="I32">
699
+ <dim>-1</dim>
700
+ <dim>-1</dim>
701
+ </port>
702
+ </input>
703
+ <output>
704
+ <port id="1" precision="I64" names="input_ids">
705
+ <dim>-1</dim>
706
+ <dim>-1</dim>
707
+ </port>
708
+ </output>
709
+ </layer>
710
+ <layer id="49" name="Result_2718649" type="Result" version="opset1">
711
+ <input>
712
+ <port id="0" precision="I64">
713
+ <dim>-1</dim>
714
+ <dim>-1</dim>
715
+ </port>
716
+ </input>
717
+ </layer>
718
+ <layer id="47" name="Result_2718651" type="Result" version="opset1">
719
+ <input>
720
+ <port id="0" precision="I64">
721
+ <dim>-1</dim>
722
+ <dim>-1</dim>
723
+ </port>
724
+ </input>
725
+ </layer>
726
+ </layers>
727
+ <edges>
728
+ <edge from-layer="0" from-port="0" to-layer="5" to-port="0" />
729
+ <edge from-layer="1" from-port="0" to-layer="39" to-port="0" />
730
+ <edge from-layer="2" from-port="0" to-layer="39" to-port="1" />
731
+ <edge from-layer="3" from-port="0" to-layer="39" to-port="2" />
732
+ <edge from-layer="4" from-port="0" to-layer="11" to-port="0" />
733
+ <edge from-layer="5" from-port="1" to-layer="6" to-port="0" />
734
+ <edge from-layer="5" from-port="3" to-layer="18" to-port="4" />
735
+ <edge from-layer="5" from-port="2" to-layer="18" to-port="3" />
736
+ <edge from-layer="5" from-port="1" to-layer="18" to-port="2" />
737
+ <edge from-layer="6" from-port="1" to-layer="9" to-port="0" />
738
+ <edge from-layer="7" from-port="0" to-layer="9" to-port="1" />
739
+ <edge from-layer="8" from-port="0" to-layer="9" to-port="2" />
740
+ <edge from-layer="9" from-port="3" to-layer="14" to-port="0" />
741
+ <edge from-layer="9" from-port="3" to-layer="11" to-port="1" />
742
+ <edge from-layer="10" from-port="0" to-layer="11" to-port="2" />
743
+ <edge from-layer="11" from-port="3" to-layer="18" to-port="0" />
744
+ <edge from-layer="12" from-port="0" to-layer="16" to-port="0" />
745
+ <edge from-layer="13" from-port="0" to-layer="14" to-port="1" />
746
+ <edge from-layer="14" from-port="2" to-layer="16" to-port="1" />
747
+ <edge from-layer="15" from-port="0" to-layer="16" to-port="2" />
748
+ <edge from-layer="16" from-port="3" to-layer="18" to-port="1" />
749
+ <edge from-layer="17" from-port="0" to-layer="18" to-port="5" />
750
+ <edge from-layer="18" from-port="8" to-layer="21" to-port="0" />
751
+ <edge from-layer="18" from-port="7" to-layer="23" to-port="1" />
752
+ <edge from-layer="18" from-port="6" to-layer="23" to-port="0" />
753
+ <edge from-layer="18" from-port="11" to-layer="21" to-port="3" />
754
+ <edge from-layer="18" from-port="10" to-layer="21" to-port="2" />
755
+ <edge from-layer="18" from-port="9" to-layer="21" to-port="1" />
756
+ <edge from-layer="19" from-port="0" to-layer="21" to-port="4" />
757
+ <edge from-layer="20" from-port="0" to-layer="21" to-port="5" />
758
+ <edge from-layer="21" from-port="6" to-layer="23" to-port="2" />
759
+ <edge from-layer="21" from-port="7" to-layer="23" to-port="3" />
760
+ <edge from-layer="21" from-port="8" to-layer="23" to-port="4" />
761
+ <edge from-layer="21" from-port="9" to-layer="23" to-port="5" />
762
+ <edge from-layer="22" from-port="0" to-layer="23" to-port="6" />
763
+ <edge from-layer="23" from-port="11" to-layer="33" to-port="4" />
764
+ <edge from-layer="23" from-port="10" to-layer="33" to-port="3" />
765
+ <edge from-layer="23" from-port="9" to-layer="33" to-port="2" />
766
+ <edge from-layer="23" from-port="7" to-layer="33" to-port="0" />
767
+ <edge from-layer="23" from-port="8" to-layer="33" to-port="1" />
768
+ <edge from-layer="24" from-port="0" to-layer="25" to-port="0" />
769
+ <edge from-layer="25" from-port="1" to-layer="33" to-port="5" />
770
+ <edge from-layer="25" from-port="2" to-layer="33" to-port="6" />
771
+ <edge from-layer="25" from-port="3" to-layer="33" to-port="7" />
772
+ <edge from-layer="26" from-port="0" to-layer="27" to-port="0" />
773
+ <edge from-layer="27" from-port="1" to-layer="33" to-port="8" />
774
+ <edge from-layer="27" from-port="2" to-layer="33" to-port="9" />
775
+ <edge from-layer="27" from-port="3" to-layer="33" to-port="10" />
776
+ <edge from-layer="28" from-port="0" to-layer="29" to-port="0" />
777
+ <edge from-layer="29" from-port="1" to-layer="33" to-port="11" />
778
+ <edge from-layer="29" from-port="2" to-layer="33" to-port="12" />
779
+ <edge from-layer="29" from-port="3" to-layer="33" to-port="13" />
780
+ <edge from-layer="30" from-port="0" to-layer="31" to-port="0" />
781
+ <edge from-layer="31" from-port="2" to-layer="33" to-port="15" />
782
+ <edge from-layer="31" from-port="3" to-layer="33" to-port="16" />
783
+ <edge from-layer="31" from-port="1" to-layer="33" to-port="14" />
784
+ <edge from-layer="32" from-port="0" to-layer="33" to-port="17" />
785
+ <edge from-layer="33" from-port="19" to-layer="34" to-port="0" />
786
+ <edge from-layer="33" from-port="18" to-layer="34" to-port="1" />
787
+ <edge from-layer="33" from-port="19" to-layer="37" to-port="0" />
788
+ <edge from-layer="33" from-port="20" to-layer="39" to-port="5" />
789
+ <edge from-layer="33" from-port="19" to-layer="39" to-port="4" />
790
+ <edge from-layer="34" from-port="2" to-layer="36" to-port="0" />
791
+ <edge from-layer="35" from-port="0" to-layer="36" to-port="1" />
792
+ <edge from-layer="36" from-port="2" to-layer="37" to-port="1" />
793
+ <edge from-layer="37" from-port="2" to-layer="39" to-port="3" />
794
+ <edge from-layer="38" from-port="0" to-layer="39" to-port="6" />
795
+ <edge from-layer="39" from-port="8" to-layer="40" to-port="0" />
796
+ <edge from-layer="39" from-port="7" to-layer="40" to-port="1" />
797
+ <edge from-layer="39" from-port="7" to-layer="44" to-port="0" />
798
+ <edge from-layer="39" from-port="8" to-layer="44" to-port="1" />
799
+ <edge from-layer="39" from-port="9" to-layer="44" to-port="2" />
800
+ <edge from-layer="40" from-port="2" to-layer="42" to-port="0" />
801
+ <edge from-layer="41" from-port="0" to-layer="42" to-port="1" />
802
+ <edge from-layer="42" from-port="2" to-layer="44" to-port="3" />
803
+ <edge from-layer="43" from-port="0" to-layer="44" to-port="4" />
804
+ <edge from-layer="44" from-port="6" to-layer="45" to-port="0" />
805
+ <edge from-layer="44" from-port="5" to-layer="48" to-port="0" />
806
+ <edge from-layer="45" from-port="1" to-layer="46" to-port="0" />
807
+ <edge from-layer="46" from-port="1" to-layer="47" to-port="0" />
808
+ <edge from-layer="48" from-port="1" to-layer="49" to-port="0" />
809
+ </edges>
810
+ <rt_info>
811
+ <add_attention_mask value="True" />
812
+ <add_prefix_space />
813
+ <add_special_tokens value="True" />
814
+ <bos_token_id value="2" />
815
+ <chat_template value="{{ bos_token }} &#10;{%- if messages[0]['role'] == 'system' -%}&#10; {%- if messages[0]['content'] is string -%}&#10; {%- set first_user_prefix = messages[0]['content'] + '&#10;&#10;' -%}&#10; {%- else -%}&#10; {%- set first_user_prefix = messages[0]['content'][0]['text'] + '&#10;&#10;' -%}&#10; {%- endif -%}&#10; {%- set loop_messages = messages[1:] -%}&#10;{%- else -%}&#10; {%- set first_user_prefix = &quot;&quot; -%}&#10; {%- set loop_messages = messages -%}&#10;{%- endif -%}&#10;{%- for message in loop_messages -%}&#10; {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}&#10; {{ raise_exception(&quot;Conversation roles must alternate user/assistant/user/assistant/...&quot;) }}&#10; {%- endif -%}&#10; {%- if (message['role'] == 'assistant') -%}&#10; {%- set role = &quot;model&quot; -%}&#10; {%- else -%}&#10; {%- set role = message['role'] -%}&#10; {%- endif -%}&#10; {{ '&lt;start_of_turn>' + role + '&#10;' + (first_user_prefix if loop.first else &quot;&quot;) }}&#10; {%- if message['content'] is string -%}&#10; {{ message['content'] | trim }}&#10; {%- elif message['content'] is iterable -%}&#10; {%- for item in message['content'] -%}&#10; {%- if item['type'] == 'image' -%}&#10; {{ '&lt;start_of_image>' }}&#10; {%- elif item['type'] == 'text' -%}&#10; {{ item['text'] | trim }}&#10; {%- endif -%}&#10; {%- endfor -%}&#10; {%- else -%}&#10; {{ raise_exception(&quot;Invalid content type&quot;) }}&#10; {%- endif -%}&#10; {{ '&lt;end_of_turn>&#10;' }}&#10;{%- endfor -%}&#10;{%- if add_generation_prompt -%}&#10; {{'&lt;start_of_turn>model&#10;'}}&#10;{%- endif -%}&#10;" />
816
+ <clean_up_tokenization_spaces />
817
+ <detokenizer_input_type value="i64" />
818
+ <eos_token_id value="1" />
819
+ <handle_special_tokens_with_re />
820
+ <number_of_inputs value="1" />
821
+ <openvino_tokenizers_version value="2025.0.0.0" />
822
+ <openvino_version value="2025.0.0" />
823
+ <original_tokenizer_class value="&lt;class 'transformers.models.gemma.tokenization_gemma_fast.GemmaTokenizerFast'>" />
824
+ <pad_token_id value="0" />
825
+ <sentencepiece_version value="0.2.0" />
826
+ <skip_special_tokens value="True" />
827
+ <streaming_detokenizer value="False" />
828
+ <tiktoken_version value="0.7.0" />
829
+ <tokenizer_output_type value="i64" />
830
+ <tokenizers_version value="0.21.1" />
831
+ <transformers_version value="4.50.2" />
832
+ <use_max_padding value="False" />
833
+ <use_sentencepiece_backend value="False" />
834
+ <utf8_replace_mode value="replace" />
835
+ <with_detokenizer value="True" />
836
+ </rt_info>
837
+ </net>
openvino_vision_embeddings_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:64606538736c69645e6b42d4945fc57cef15ad42756db5b947516002441158a5
3
+ size 427238272
openvino_vision_embeddings_model.xml 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": "<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