Hugging Face's logo

Echo9Zulu
/
gemma-3-4b-it-qat-int4_asym-ov

Image-Text-to-Text
OpenVINO
gemma3
OpenArc
OpenVINO
Optimum-Intel
conversational
Model card Files Community
Echo9Zulu commited on
Commit
88cc288
·
verified ·
1 Parent(s): bfcbe0b

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 +11 -0
  6. openvino_config.json +27 -0
  7. openvino_detokenizer.bin +3 -0
  8. openvino_detokenizer.xml +301 -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 +823 -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": 2560,
25
+ "initializer_range": 0.02,
26
+ "intermediate_size": 10240,
27
+ "max_position_embeddings": 131072,
28
+ "model_type": "gemma3_text",
29
+ "num_attention_heads": 8,
30
+ "num_hidden_layers": 34,
31
+ "num_key_value_heads": 4,
32
+ "query_pre_attn_scalar": 256,
33
+ "rms_norm_eps": 1e-06,
34
+ "rope_local_base_freq": 10000,
35
+ "rope_scaling": {
36
+ "factor": 8.0,
37
+ "rope_type": "linear"
38
+ },
39
+ "rope_theta": 1000000,
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 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "cache_implementation": "hybrid",
3
+ "do_sample": true,
4
+ "eos_token_id": [
5
+ 1,
6
+ 106
7
+ ],
8
+ "top_k": 64,
9
+ "top_p": 0.95,
10
+ "transformers_version": "4.50.2"
11
+ }
openvino_config.json ADDED
@@ -0,0 +1,27 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dtype": "int4",
3
+ "input_info": null,
4
+ "optimum_version": "1.25.0.dev0",
5
+ "quantization_config": {
6
+ "all_layers": null,
7
+ "backup_precision": null,
8
+ "bits": 4,
9
+ "dataset": null,
10
+ "dtype": "int4",
11
+ "gptq": null,
12
+ "group_size": 128,
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 sha256:4ead5eca718061fcc6f5dc3e4db2de14f8b297489ddf8512bf62c5afdc7cf050
3
+ size 4413574
openvino_detokenizer.xml ADDED
@@ -0,0 +1,301 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <?xml version="1.0"?>
2
+ <net name="detokenizer" version="11">
3
+ <layers>
4
+ <layer id="0" name="Parameter_2236332" type="Parameter" version="opset1">
5
+ <data shape="?,?" element_type="i64" />
6
+ <output>
7
+ <port id="0" precision="I64" names="Parameter_2236332">
8
+ <dim>-1</dim>
9
+ <dim>-1</dim>
10
+ </port>
11
+ </output>
12
+ </layer>
13
+ <layer id="1" name="Convert_2236502" 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_2236287" type="Const" version="opset1">
29
+ <data element_type="i32" shape="262145" offset="0" size="1048580" />
30
+ <output>
31
+ <port id="0" precision="I32">
32
+ <dim>262145</dim>
33
+ </port>
34
+ </output>
35
+ </layer>
36
+ <layer id="3" name="Constant_2236289" type="Const" version="opset1">
37
+ <data element_type="i32" shape="262145" offset="1048580" size="1048580" />
38
+ <output>
39
+ <port id="0" precision="I32">
40
+ <dim>262145</dim>
41
+ </port>
42
+ </output>
43
+ </layer>
44
+ <layer id="4" name="Constant_2236291" type="Const" version="opset1">
45
+ <data element_type="u8" shape="2316374" offset="2097160" size="2316374" />
46
+ <output>
47
+ <port id="0" precision="U8">
48
+ <dim>2316374</dim>
49
+ </port>
50
+ </output>
51
+ </layer>
52
+ <layer id="5" name="Slice_2236337" type="Const" version="opset1">
53
+ <data element_type="i32" shape="9" offset="4413534" size="36" />
54
+ <output>
55
+ <port id="0" precision="I32">
56
+ <dim>9</dim>
57
+ </port>
58
+ </output>
59
+ </layer>
60
+ <layer id="6" name="VocabDecoder_2236339" type="VocabDecoder" version="extension">
61
+ <data skip_tokens="" />
62
+ <input>
63
+ <port id="0" precision="I32">
64
+ <dim>-1</dim>
65
+ <dim>-1</dim>
66
+ </port>
67
+ <port id="1" precision="I32">
68
+ <dim>262145</dim>
69
+ </port>
70
+ <port id="2" precision="I32">
71
+ <dim>262145</dim>
72
+ </port>
73
+ <port id="3" precision="U8">
74
+ <dim>2316374</dim>
75
+ </port>
76
+ <port id="4" precision="I32">
77
+ <dim>9</dim>
78
+ </port>
79
+ </input>
80
+ <output>
81
+ <port id="5" precision="I32">
82
+ <dim>-1</dim>
83
+ </port>
84
+ <port id="6" precision="I32">
85
+ <dim>-1</dim>
86
+ </port>
87
+ <port id="7" precision="I32">
88
+ <dim>-1</dim>
89
+ </port>
90
+ <port id="8" precision="I32">
91
+ <dim>-1</dim>
92
+ </port>
93
+ <port id="9" precision="U8">
94
+ <dim>-1</dim>
95
+ </port>
96
+ </output>
97
+ </layer>
98
+ <layer id="7" name="Constant_2236341" type="Const" version="opset1">
99
+ <data element_type="u8" shape="3" offset="4413570" size="3" />
100
+ <output>
101
+ <port id="0" precision="U8">
102
+ <dim>3</dim>
103
+ </port>
104
+ </output>
105
+ </layer>
106
+ <layer id="8" name="Constant_2236343" type="Const" version="opset1">
107
+ <data element_type="u8" shape="1" offset="4413573" size="1" />
108
+ <output>
109
+ <port id="0" precision="U8">
110
+ <dim>1</dim>
111
+ </port>
112
+ </output>
113
+ </layer>
114
+ <layer id="9" name="RegexNormalization_2236344" type="RegexNormalization" version="extension">
115
+ <data global_replace="true" />
116
+ <input>
117
+ <port id="0" precision="I32">
118
+ <dim>-1</dim>
119
+ </port>
120
+ <port id="1" precision="I32">
121
+ <dim>-1</dim>
122
+ </port>
123
+ <port id="2" precision="U8">
124
+ <dim>-1</dim>
125
+ </port>
126
+ <port id="3" precision="U8">
127
+ <dim>3</dim>
128
+ </port>
129
+ <port id="4" precision="U8">
130
+ <dim>1</dim>
131
+ </port>
132
+ </input>
133
+ <output>
134
+ <port id="5" precision="I32">
135
+ <dim>-1</dim>
136
+ </port>
137
+ <port id="6" precision="I32">
138
+ <dim>-1</dim>
139
+ </port>
140
+ <port id="7" precision="U8">
141
+ <dim>-1</dim>
142
+ </port>
143
+ </output>
144
+ </layer>
145
+ <layer id="10" name="ByteFallback_2236345" type="ByteFallback" version="extension">
146
+ <input>
147
+ <port id="0" precision="I32">
148
+ <dim>-1</dim>
149
+ </port>
150
+ <port id="1" precision="I32">
151
+ <dim>-1</dim>
152
+ </port>
153
+ <port id="2" precision="U8">
154
+ <dim>-1</dim>
155
+ </port>
156
+ </input>
157
+ <output>
158
+ <port id="3" precision="I32">
159
+ <dim>-1</dim>
160
+ </port>
161
+ <port id="4" precision="I32">
162
+ <dim>-1</dim>
163
+ </port>
164
+ <port id="5" precision="U8">
165
+ <dim>-1</dim>
166
+ </port>
167
+ </output>
168
+ </layer>
169
+ <layer id="11" name="FuzeRagged_2236346" type="FuzeRagged" version="extension">
170
+ <input>
171
+ <port id="0" precision="I32">
172
+ <dim>-1</dim>
173
+ </port>
174
+ <port id="1" precision="I32">
175
+ <dim>-1</dim>
176
+ </port>
177
+ <port id="2" precision="I32">
178
+ <dim>-1</dim>
179
+ </port>
180
+ <port id="3" precision="I32">
181
+ <dim>-1</dim>
182
+ </port>
183
+ </input>
184
+ <output>
185
+ <port id="4" precision="I32">
186
+ <dim>-1</dim>
187
+ </port>
188
+ <port id="5" precision="I32">
189
+ <dim>-1</dim>
190
+ </port>
191
+ </output>
192
+ </layer>
193
+ <layer id="12" name="UTF8Validate_2236347" type="UTF8Validate" version="extension">
194
+ <data replace_mode="true" />
195
+ <input>
196
+ <port id="0" precision="I32">
197
+ <dim>-1</dim>
198
+ </port>
199
+ <port id="1" precision="I32">
200
+ <dim>-1</dim>
201
+ </port>
202
+ <port id="2" precision="U8">
203
+ <dim>-1</dim>
204
+ </port>
205
+ </input>
206
+ <output>
207
+ <port id="3" precision="I32">
208
+ <dim>-1</dim>
209
+ </port>
210
+ <port id="4" precision="I32">
211
+ <dim>-1</dim>
212
+ </port>
213
+ <port id="5" precision="U8">
214
+ <dim>-1</dim>
215
+ </port>
216
+ </output>
217
+ </layer>
218
+ <layer id="13" name="StringTensorPack_2236348" type="StringTensorPack" version="opset15">
219
+ <input>
220
+ <port id="0" precision="I32">
221
+ <dim>-1</dim>
222
+ </port>
223
+ <port id="1" precision="I32">
224
+ <dim>-1</dim>
225
+ </port>
226
+ <port id="2" precision="U8">
227
+ <dim>-1</dim>
228
+ </port>
229
+ </input>
230
+ <output>
231
+ <port id="3" precision="STRING" names="Result_2236349,string_output">
232
+ <dim>-1</dim>
233
+ </port>
234
+ </output>
235
+ </layer>
236
+ <layer id="14" name="Result_2236349" type="Result" version="opset1" output_names="Result_2236349,string_output">
237
+ <input>
238
+ <port id="0" precision="STRING">
239
+ <dim>-1</dim>
240
+ </port>
241
+ </input>
242
+ </layer>
243
+ </layers>
244
+ <edges>
245
+ <edge from-layer="0" from-port="0" to-layer="1" to-port="0" />
246
+ <edge from-layer="1" from-port="1" to-layer="6" to-port="0" />
247
+ <edge from-layer="2" from-port="0" to-layer="6" to-port="1" />
248
+ <edge from-layer="3" from-port="0" to-layer="6" to-port="2" />
249
+ <edge from-layer="4" from-port="0" to-layer="6" to-port="3" />
250
+ <edge from-layer="5" from-port="0" to-layer="6" to-port="4" />
251
+ <edge from-layer="6" from-port="7" to-layer="9" to-port="0" />
252
+ <edge from-layer="6" from-port="8" to-layer="9" to-port="1" />
253
+ <edge from-layer="6" from-port="9" to-layer="9" to-port="2" />
254
+ <edge from-layer="6" from-port="6" to-layer="11" to-port="1" />
255
+ <edge from-layer="6" from-port="5" to-layer="11" to-port="0" />
256
+ <edge from-layer="7" from-port="0" to-layer="9" to-port="3" />
257
+ <edge from-layer="8" from-port="0" to-layer="9" to-port="4" />
258
+ <edge from-layer="9" from-port="6" to-layer="10" to-port="1" />
259
+ <edge from-layer="9" from-port="7" to-layer="10" to-port="2" />
260
+ <edge from-layer="9" from-port="5" to-layer="10" to-port="0" />
261
+ <edge from-layer="10" from-port="3" to-layer="11" to-port="2" />
262
+ <edge from-layer="10" from-port="4" to-layer="11" to-port="3" />
263
+ <edge from-layer="10" from-port="5" to-layer="12" to-port="2" />
264
+ <edge from-layer="11" from-port="4" to-layer="12" to-port="0" />
265
+ <edge from-layer="11" from-port="5" to-layer="12" to-port="1" />
266
+ <edge from-layer="12" from-port="3" to-layer="13" to-port="0" />
267
+ <edge from-layer="12" from-port="4" to-layer="13" to-port="1" />
268
+ <edge from-layer="12" from-port="5" to-layer="13" to-port="2" />
269
+ <edge from-layer="13" from-port="3" to-layer="14" to-port="0" />
270
+ </edges>
271
+ <rt_info>
272
+ <add_attention_mask value="True" />
273
+ <add_prefix_space />
274
+ <add_special_tokens value="True" />
275
+ <bos_token_id value="2" />
276
+ <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;" />
277
+ <clean_up_tokenization_spaces />
278
+ <detokenizer_input_type value="i64" />
279
+ <eos_token_id value="1" />
280
+ <handle_special_tokens_with_re />
281
+ <max_length />
282
+ <number_of_inputs value="1" />
283
+ <openvino_tokenizers_version value="2025.1.0.0-523-710ddf14de8" />
284
+ <openvino_version value="2025.1.0-18503-6fec06580ab-releases/2025/1" />
285
+ <original_post_processor_template value="{&quot;type&quot;: &quot;TemplateProcessing&quot;, &quot;single&quot;: [{&quot;SpecialToken&quot;: {&quot;id&quot;: &quot;&lt;bos>&quot;, &quot;type_id&quot;: 0}}, {&quot;Sequence&quot;: {&quot;id&quot;: &quot;A&quot;, &quot;type_id&quot;: 0}}], &quot;pair&quot;: [{&quot;SpecialToken&quot;: {&quot;id&quot;: &quot;&lt;bos>&quot;, &quot;type_id&quot;: 0}}, {&quot;Sequence&quot;: {&quot;id&quot;: &quot;A&quot;, &quot;type_id&quot;: 0}}, {&quot;SpecialToken&quot;: {&quot;id&quot;: &quot;&lt;bos>&quot;, &quot;type_id&quot;: 1}}, {&quot;Sequence&quot;: {&quot;id&quot;: &quot;B&quot;, &quot;type_id&quot;: 1}}], &quot;special_tokens&quot;: {&quot;&lt;bos>&quot;: {&quot;id&quot;: &quot;&lt;bos>&quot;, &quot;ids&quot;: [2], &quot;tokens&quot;: [&quot;&lt;bos>&quot;]}}}" />
286
+ <original_tokenizer_class value="&lt;class 'transformers.models.gemma.tokenization_gemma_fast.GemmaTokenizerFast'>" />
287
+ <pad_token_id value="0" />
288
+ <processed_post_processor_template value="{&quot;single&quot;: {&quot;ids&quot;: [2, -1], &quot;type_ids&quot;: [0, 0]}, &quot;pair&quot;: {&quot;ids&quot;: [2, -1, 2, -2], &quot;type_ids&quot;: [0, 0, 1, 1]}}" />
289
+ <sentencepiece_version value="0.2.0" />
290
+ <skip_special_tokens value="True" />
291
+ <streaming_detokenizer value="False" />
292
+ <tiktoken_version value="0.7.0" />
293
+ <tokenizer_output_type value="i64" />
294
+ <tokenizers_version value="0.21.1" />
295
+ <transformers_version value="4.50.2" />
296
+ <use_max_padding value="False" />
297
+ <use_sentencepiece_backend value="False" />
298
+ <utf8_replace_mode value="replace" />
299
+ <with_detokenizer value="True" />
300
+ </rt_info>
301
+ </net>
openvino_language_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2bfe2919f988fde2366c3cf050e1980b689e945bd0073dbd5e2aeea4cd89c997
3
+ size 2340503078
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 sha256:0321f5a3042459ff17aafc49c465bf3cf8c3bbe640e63a6ba97f7cf7a1d21603
3
+ size 671776904
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, 2560" offset="0" size="671252480" />
15
+ <output>
16
+ <port id="0" precision="I8">
17
+ <dim>262208</dim>
18
+ <dim>2560</dim>
19
+ </port>
20
+ </output>
21
+ </layer>
22
+ <layer id="2" name="Convert_1423363" type="Convert" version="opset1">
23
+ <data destination_type="f16" />
24
+ <input>
25
+ <port id="0" precision="I8">
26
+ <dim>262208</dim>
27
+ <dim>2560</dim>
28
+ </port>
29
+ </input>
30
+ <output>
31
+ <port id="1" precision="FP16">
32
+ <dim>262208</dim>
33
+ <dim>2560</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="671252480" 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>2560</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>2560</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>2560</dim>
71
+ </port>
72
+ </input>
73
+ <output>
74
+ <port id="1" precision="FP32">
75
+ <dim>262208</dim>
76
+ <dim>2560</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="671776896" 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>2560</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>2560</dim>
119
+ </port>
120
+ </output>
121
+ </layer>
122
+ <layer id="9" name="Constant_157790" type="Const" version="opset1">
123
+ <data element_type="f32" shape="1, 1, 1" offset="671776900" 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>2560</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>2560</dim>
151
+ </port>
152
+ </output>
153
+ </layer>
154
+ <layer id="11" name="Result_155976" type="Result" version="opset1" output_names="inputs_embeds">
155
+ <input>
156
+ <port id="0" precision="FP32">
157
+ <dim>-1</dim>
158
+ <dim>-1</dim>
159
+ <dim>2560</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="2025.1.0-18503-6fec06580ab-releases/2025/1" />
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="2.6.0+cpu" />
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 sha256:428aacbdb05a817d2ecfda471c2e9f8a15739a7d6718e86fcff7b85c80fb3151
3
+ size 17533095
openvino_tokenizer.xml ADDED
@@ -0,0 +1,823 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <?xml version="1.0"?>
2
+ <net name="tokenizer" version="11">
3
+ <layers>
4
+ <layer id="0" name="Parameter_2236201" type="Parameter" version="opset1">
5
+ <data shape="?" element_type="string" />
6
+ <output>
7
+ <port id="0" precision="STRING" names="Parameter_2236201">
8
+ <dim>-1</dim>
9
+ </port>
10
+ </output>
11
+ </layer>
12
+ <layer id="1" name="Constant_2236316" 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_2236317" 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_2236318" 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_2236207" 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_2236202" type="StringTensorUnpack" version="opset15">
39
+ <input>
40
+ <port id="0" precision="STRING">
41
+ <dim>-1</dim>
42
+ </port>
43
+ </input>
44
+ <output>
45
+ <port id="1" precision="I32">
46
+ <dim>-1</dim>
47
+ </port>
48
+ <port id="2" precision="I32">
49
+ <dim>-1</dim>
50
+ </port>
51
+ <port id="3" precision="U8">
52
+ <dim>-1</dim>
53
+ </port>
54
+ </output>
55
+ </layer>
56
+ <layer id="6" name="ShapeOf_2236203" type="ShapeOf" version="opset3">
57
+ <data output_type="i64" />
58
+ <input>
59
+ <port id="0" precision="I32">
60
+ <dim>-1</dim>
61
+ </port>
62
+ </input>
63
+ <output>
64
+ <port id="1" precision="I64">
65
+ <dim>1</dim>
66
+ </port>
67
+ </output>
68
+ </layer>
69
+ <layer id="7" name="Constant_2236204" type="Const" version="opset1">
70
+ <data element_type="i64" shape="" offset="12" size="8" />
71
+ <output>
72
+ <port id="0" precision="I64" />
73
+ </output>
74
+ </layer>
75
+ <layer id="8" name="Constant_2236205" type="Const" version="opset1">
76
+ <data element_type="i64" shape="" offset="12" size="8" />
77
+ <output>
78
+ <port id="0" precision="I64" />
79
+ </output>
80
+ </layer>
81
+ <layer id="9" name="Gather_2236206" type="Gather" version="opset8">
82
+ <data batch_dims="0" />
83
+ <input>
84
+ <port id="0" precision="I64">
85
+ <dim>1</dim>
86
+ </port>
87
+ <port id="1" precision="I64" />
88
+ <port id="2" precision="I64" />
89
+ </input>
90
+ <output>
91
+ <port id="3" precision="I64" />
92
+ </output>
93
+ </layer>
94
+ <layer id="10" name="Constant_2236208" type="Const" version="opset1">
95
+ <data element_type="i64" shape="" offset="20" size="8" />
96
+ <output>
97
+ <port id="0" precision="I64" />
98
+ </output>
99
+ </layer>
100
+ <layer id="11" name="Range_2236209" type="Range" version="opset4">
101
+ <data output_type="i32" />
102
+ <input>
103
+ <port id="0" precision="I64" />
104
+ <port id="1" precision="I64" />
105
+ <port id="2" precision="I64" />
106
+ </input>
107
+ <output>
108
+ <port id="3" precision="I32">
109
+ <dim>-1</dim>
110
+ </port>
111
+ </output>
112
+ </layer>
113
+ <layer id="12" name="Constant_2236210" type="Const" version="opset1">
114
+ <data element_type="i64" shape="" offset="20" size="8" />
115
+ <output>
116
+ <port id="0" precision="I64" />
117
+ </output>
118
+ </layer>
119
+ <layer id="13" name="Constant_2236211" type="Const" version="opset1">
120
+ <data element_type="i64" shape="" offset="20" size="8" />
121
+ <output>
122
+ <port id="0" precision="I64" />
123
+ </output>
124
+ </layer>
125
+ <layer id="14" name="Add_2236212" type="Add" version="opset1">
126
+ <data auto_broadcast="numpy" />
127
+ <input>
128
+ <port id="0" precision="I64" />
129
+ <port id="1" precision="I64" />
130
+ </input>
131
+ <output>
132
+ <port id="2" precision="I64" />
133
+ </output>
134
+ </layer>
135
+ <layer id="15" name="Constant_2236213" type="Const" version="opset1">
136
+ <data element_type="i64" shape="" offset="20" size="8" />
137
+ <output>
138
+ <port id="0" precision="I64" />
139
+ </output>
140
+ </layer>
141
+ <layer id="16" name="Range_2236214" type="Range" version="opset4">
142
+ <data output_type="i32" />
143
+ <input>
144
+ <port id="0" precision="I64" />
145
+ <port id="1" precision="I64" />
146
+ <port id="2" precision="I64" />
147
+ </input>
148
+ <output>
149
+ <port id="3" precision="I32">
150
+ <dim>-1</dim>
151
+ </port>
152
+ </output>
153
+ </layer>
154
+ <layer id="17" name="Constant_2236276" type="Const" version="opset1">
155
+ <data element_type="u8" shape="109709" offset="28" size="109709" />
156
+ <output>
157
+ <port id="0" precision="U8">
158
+ <dim>109709</dim>
159
+ </port>
160
+ </output>
161
+ </layer>
162
+ <layer id="18" name="SpecialTokensSplit_2236277" type="SpecialTokensSplit" version="extension">
163
+ <input>
164
+ <port id="0" precision="I32">
165
+ <dim>-1</dim>
166
+ </port>
167
+ <port id="1" precision="I32">
168
+ <dim>-1</dim>
169
+ </port>
170
+ <port id="2" precision="I32">
171
+ <dim>-1</dim>
172
+ </port>
173
+ <port id="3" precision="I32">
174
+ <dim>-1</dim>
175
+ </port>
176
+ <port id="4" precision="U8">
177
+ <dim>-1</dim>
178
+ </port>
179
+ <port id="5" precision="U8">
180
+ <dim>109709</dim>
181
+ </port>
182
+ </input>
183
+ <output>
184
+ <port id="6" precision="I32">
185
+ <dim>-1</dim>
186
+ </port>
187
+ <port id="7" precision="I32">
188
+ <dim>-1</dim>
189
+ </port>
190
+ <port id="8" precision="I32">
191
+ <dim>-1</dim>
192
+ </port>
193
+ <port id="9" precision="I32">
194
+ <dim>-1</dim>
195
+ </port>
196
+ <port id="10" precision="U8">
197
+ <dim>-1</dim>
198
+ </port>
199
+ <port id="11" precision="BOOL">
200
+ <dim>-1</dim>
201
+ </port>
202
+ </output>
203
+ </layer>
204
+ <layer id="19" name="Constant_2236279" type="Const" version="opset1">
205
+ <data element_type="u8" shape="1" offset="109737" size="1" />
206
+ <output>
207
+ <port id="0" precision="U8">
208
+ <dim>1</dim>
209
+ </port>
210
+ </output>
211
+ </layer>
212
+ <layer id="20" name="Constant_2236281" type="Const" version="opset1">
213
+ <data element_type="u8" shape="3" offset="109738" size="3" />
214
+ <output>
215
+ <port id="0" precision="U8">
216
+ <dim>3</dim>
217
+ </port>
218
+ </output>
219
+ </layer>
220
+ <layer id="21" name="RegexNormalization_2236282" type="RegexNormalization" version="extension">
221
+ <data global_replace="true" />
222
+ <input>
223
+ <port id="0" precision="I32">
224
+ <dim>-1</dim>
225
+ </port>
226
+ <port id="1" precision="I32">
227
+ <dim>-1</dim>
228
+ </port>
229
+ <port id="2" precision="U8">
230
+ <dim>-1</dim>
231
+ </port>
232
+ <port id="3" precision="BOOL">
233
+ <dim>-1</dim>
234
+ </port>
235
+ <port id="4" precision="U8">
236
+ <dim>1</dim>
237
+ </port>
238
+ <port id="5" precision="U8">
239
+ <dim>3</dim>
240
+ </port>
241
+ </input>
242
+ <output>
243
+ <port id="6" precision="I32">
244
+ <dim>-1</dim>
245
+ </port>
246
+ <port id="7" precision="I32">
247
+ <dim>-1</dim>
248
+ </port>
249
+ <port id="8" precision="U8">
250
+ <dim>-1</dim>
251
+ </port>
252
+ <port id="9" precision="BOOL">
253
+ <dim>-1</dim>
254
+ </port>
255
+ </output>
256
+ </layer>
257
+ <layer id="22" name="Constant_2236284" type="Const" version="opset1">
258
+ <data element_type="u8" shape="1" offset="109737" size="1" />
259
+ <output>
260
+ <port id="0" precision="U8">
261
+ <dim>1</dim>
262
+ </port>
263
+ </output>
264
+ </layer>
265
+ <layer id="23" name="RegexSplit_2236285" type="RegexSplit" version="extension">
266
+ <data behaviour="mergedwithprevious" invert="false" max_splits="-1" />
267
+ <input>
268
+ <port id="0" precision="I32">
269
+ <dim>-1</dim>
270
+ </port>
271
+ <port id="1" precision="I32">
272
+ <dim>-1</dim>
273
+ </port>
274
+ <port id="2" precision="I32">
275
+ <dim>-1</dim>
276
+ </port>
277
+ <port id="3" precision="I32">
278
+ <dim>-1</dim>
279
+ </port>
280
+ <port id="4" precision="U8">
281
+ <dim>-1</dim>
282
+ </port>
283
+ <port id="5" precision="BOOL">
284
+ <dim>-1</dim>
285
+ </port>
286
+ <port id="6" precision="U8">
287
+ <dim>1</dim>
288
+ </port>
289
+ </input>
290
+ <output>
291
+ <port id="7" precision="I32">
292
+ <dim>-1</dim>
293
+ </port>
294
+ <port id="8" precision="I32">
295
+ <dim>-1</dim>
296
+ </port>
297
+ <port id="9" precision="I32">
298
+ <dim>-1</dim>
299
+ </port>
300
+ <port id="10" precision="I32">
301
+ <dim>-1</dim>
302
+ </port>
303
+ <port id="11" precision="U8">
304
+ <dim>-1</dim>
305
+ </port>
306
+ <port id="12" precision="BOOL">
307
+ <dim>-1</dim>
308
+ </port>
309
+ </output>
310
+ </layer>
311
+ <layer id="24" name="Constant_2236287" type="Const" version="opset1">
312
+ <data element_type="i32" shape="262145" offset="109741" size="1048580" />
313
+ <output>
314
+ <port id="0" precision="I32">
315
+ <dim>262145</dim>
316
+ </port>
317
+ </output>
318
+ </layer>
319
+ <layer id="25" name="Constant_2236289" type="Const" version="opset1">
320
+ <data element_type="i32" shape="262145" offset="1158321" size="1048580" />
321
+ <output>
322
+ <port id="0" precision="I32">
323
+ <dim>262145</dim>
324
+ </port>
325
+ </output>
326
+ </layer>
327
+ <layer id="26" name="Constant_2236291" type="Const" version="opset1">
328
+ <data element_type="u8" shape="2316374" offset="2206901" size="2316374" />
329
+ <output>
330
+ <port id="0" precision="U8">
331
+ <dim>2316374</dim>
332
+ </port>
333
+ </output>
334
+ </layer>
335
+ <layer id="27" name="Constant_2236299" type="Const" version="opset1">
336
+ <data element_type="i32" shape="514906" offset="4523275" size="2059624" />
337
+ <output>
338
+ <port id="0" precision="I32">
339
+ <dim>514906</dim>
340
+ </port>
341
+ </output>
342
+ </layer>
343
+ <layer id="28" name="Constant_2236301" type="Const" version="opset1">
344
+ <data element_type="i32" shape="514906" offset="6582899" size="2059624" />
345
+ <output>
346
+ <port id="0" precision="I32">
347
+ <dim>514906</dim>
348
+ </port>
349
+ </output>
350
+ </layer>
351
+ <layer id="29" name="Constant_2236303" type="Const" version="opset1">
352
+ <data element_type="u8" shape="2714340" offset="8642523" size="2714340" />
353
+ <output>
354
+ <port id="0" precision="U8">
355
+ <dim>2714340</dim>
356
+ </port>
357
+ </output>
358
+ </layer>
359
+ <layer id="30" name="Constant_2236305" type="Const" version="opset1">
360
+ <data element_type="i32" shape="514906" offset="11356863" size="2059624" />
361
+ <output>
362
+ <port id="0" precision="I32">
363
+ <dim>514906</dim>
364
+ </port>
365
+ </output>
366
+ </layer>
367
+ <layer id="31" name="Constant_2236307" type="Const" version="opset1">
368
+ <data element_type="i32" shape="514906" offset="13416487" size="2059624" />
369
+ <output>
370
+ <port id="0" precision="I32">
371
+ <dim>514906</dim>
372
+ </port>
373
+ </output>
374
+ </layer>
375
+ <layer id="32" name="Constant_2236309" type="Const" version="opset1">
376
+ <data element_type="u8" shape="1903225" offset="15476111" size="1903225" />
377
+ <output>
378
+ <port id="0" precision="U8">
379
+ <dim>1903225</dim>
380
+ </port>
381
+ </output>
382
+ </layer>
383
+ <layer id="33" name="Constant_2236293" type="Const" version="opset1">
384
+ <data element_type="i32" shape="6414" offset="17379336" size="25656" />
385
+ <output>
386
+ <port id="0" precision="I32">
387
+ <dim>6414</dim>
388
+ </port>
389
+ </output>
390
+ </layer>
391
+ <layer id="34" name="Constant_2236295" type="Const" version="opset1">
392
+ <data element_type="i32" shape="6414" offset="17404992" size="25656" />
393
+ <output>
394
+ <port id="0" precision="I32">
395
+ <dim>6414</dim>
396
+ </port>
397
+ </output>
398
+ </layer>
399
+ <layer id="35" name="Constant_2236297" type="Const" version="opset1">
400
+ <data element_type="u8" shape="76787" offset="17430648" size="76787" />
401
+ <output>
402
+ <port id="0" precision="U8">
403
+ <dim>76787</dim>
404
+ </port>
405
+ </output>
406
+ </layer>
407
+ <layer id="36" name="Constant_2236310" type="Const" version="opset1">
408
+ <data element_type="i32" shape="6414" offset="17507435" size="25656" />
409
+ <output>
410
+ <port id="0" precision="I32">
411
+ <dim>6414</dim>
412
+ </port>
413
+ </output>
414
+ </layer>
415
+ <layer id="37" name="BPETokenizer_2236311" type="BPETokenizer" version="extension">
416
+ <data unk_token="&lt;unk>" fuse_unk="true" suffix_indicator="" end_suffix="" byte_fallback="true" cache_capacity="52428" />
417
+ <input>
418
+ <port id="0" precision="I32">
419
+ <dim>-1</dim>
420
+ </port>
421
+ <port id="1" precision="I32">
422
+ <dim>-1</dim>
423
+ </port>
424
+ <port id="2" precision="I32">
425
+ <dim>-1</dim>
426
+ </port>
427
+ <port id="3" precision="I32">
428
+ <dim>-1</dim>
429
+ </port>
430
+ <port id="4" precision="U8">
431
+ <dim>-1</dim>
432
+ </port>
433
+ <port id="5" precision="I32">
434
+ <dim>262145</dim>
435
+ </port>
436
+ <port id="6" precision="I32">
437
+ <dim>262145</dim>
438
+ </port>
439
+ <port id="7" precision="U8">
440
+ <dim>2316374</dim>
441
+ </port>
442
+ <port id="8" precision="I32">
443
+ <dim>514906</dim>
444
+ </port>
445
+ <port id="9" precision="I32">
446
+ <dim>514906</dim>
447
+ </port>
448
+ <port id="10" precision="U8">
449
+ <dim>2714340</dim>
450
+ </port>
451
+ <port id="11" precision="I32">
452
+ <dim>514906</dim>
453
+ </port>
454
+ <port id="12" precision="I32">
455
+ <dim>514906</dim>
456
+ </port>
457
+ <port id="13" precision="U8">
458
+ <dim>1903225</dim>
459
+ </port>
460
+ <port id="14" precision="I32">
461
+ <dim>6414</dim>
462
+ </port>
463
+ <port id="15" precision="I32">
464
+ <dim>6414</dim>
465
+ </port>
466
+ <port id="16" precision="U8">
467
+ <dim>76787</dim>
468
+ </port>
469
+ <port id="17" precision="I32">
470
+ <dim>6414</dim>
471
+ </port>
472
+ </input>
473
+ <output>
474
+ <port id="18" precision="I32">
475
+ <dim>-1</dim>
476
+ </port>
477
+ <port id="19" precision="I32">
478
+ <dim>-1</dim>
479
+ </port>
480
+ <port id="20" precision="I32">
481
+ <dim>-1</dim>
482
+ </port>
483
+ </output>
484
+ </layer>
485
+ <layer id="38" name="Subtract_2236312" type="Subtract" version="opset1">
486
+ <data auto_broadcast="numpy" />
487
+ <input>
488
+ <port id="0" precision="I32">
489
+ <dim>-1</dim>
490
+ </port>
491
+ <port id="1" precision="I32">
492
+ <dim>-1</dim>
493
+ </port>
494
+ </input>
495
+ <output>
496
+ <port id="2" precision="I32">
497
+ <dim>-1</dim>
498
+ </port>
499
+ </output>
500
+ </layer>
501
+ <layer id="39" name="Constant_2236313" type="Const" version="opset1">
502
+ <data element_type="i32" shape="" offset="17533091" size="4" />
503
+ <output>
504
+ <port id="0" precision="I32" />
505
+ </output>
506
+ </layer>
507
+ <layer id="40" name="Minimum_2236314" type="Minimum" version="opset1">
508
+ <data auto_broadcast="numpy" />
509
+ <input>
510
+ <port id="0" precision="I32">
511
+ <dim>-1</dim>
512
+ </port>
513
+ <port id="1" precision="I32" />
514
+ </input>
515
+ <output>
516
+ <port id="2" precision="I32">
517
+ <dim>-1</dim>
518
+ </port>
519
+ </output>
520
+ </layer>
521
+ <layer id="41" name="Subtract_2236315" type="Subtract" version="opset1">
522
+ <data auto_broadcast="numpy" />
523
+ <input>
524
+ <port id="0" precision="I32">
525
+ <dim>-1</dim>
526
+ </port>
527
+ <port id="1" precision="I32">
528
+ <dim>-1</dim>
529
+ </port>
530
+ </input>
531
+ <output>
532
+ <port id="2" precision="I32">
533
+ <dim>-1</dim>
534
+ </port>
535
+ </output>
536
+ </layer>
537
+ <layer id="42" name="Constant_2236319" type="Const" version="opset1">
538
+ <data element_type="i32" shape="2" offset="12" size="8" />
539
+ <output>
540
+ <port id="0" precision="I32">
541
+ <dim>2</dim>
542
+ </port>
543
+ </output>
544
+ </layer>
545
+ <layer id="43" name="CombineSegments_2236320" type="CombineSegments" version="extension">
546
+ <input>
547
+ <port id="0" precision="I32" />
548
+ <port id="1" precision="I32" />
549
+ <port id="2" precision="I32">
550
+ <dim>1</dim>
551
+ </port>
552
+ <port id="3" precision="I32">
553
+ <dim>-1</dim>
554
+ </port>
555
+ <port id="4" precision="I32">
556
+ <dim>-1</dim>
557
+ </port>
558
+ <port id="5" precision="I32">
559
+ <dim>-1</dim>
560
+ </port>
561
+ <port id="6" precision="I32">
562
+ <dim>2</dim>
563
+ </port>
564
+ </input>
565
+ <output>
566
+ <port id="7" precision="I32">
567
+ <dim>-1</dim>
568
+ </port>
569
+ <port id="8" precision="I32">
570
+ <dim>-1</dim>
571
+ </port>
572
+ <port id="9" precision="I32">
573
+ <dim>-1</dim>
574
+ </port>
575
+ <port id="10" precision="I32">
576
+ <dim>-1</dim>
577
+ </port>
578
+ <port id="11" precision="I32">
579
+ <dim>-1</dim>
580
+ </port>
581
+ <port id="12" precision="I32">
582
+ <dim>-1</dim>
583
+ </port>
584
+ </output>
585
+ </layer>
586
+ <layer id="44" name="Subtract_2236321" type="Subtract" version="opset1">
587
+ <data auto_broadcast="numpy" />
588
+ <input>
589
+ <port id="0" precision="I32">
590
+ <dim>-1</dim>
591
+ </port>
592
+ <port id="1" precision="I32">
593
+ <dim>-1</dim>
594
+ </port>
595
+ </input>
596
+ <output>
597
+ <port id="2" precision="I32">
598
+ <dim>-1</dim>
599
+ </port>
600
+ </output>
601
+ </layer>
602
+ <layer id="45" name="Constant_2236322" type="Const" version="opset1">
603
+ <data element_type="i32" shape="" offset="0" size="4" />
604
+ <output>
605
+ <port id="0" precision="I32" />
606
+ </output>
607
+ </layer>
608
+ <layer id="46" name="ReduceMax_2236323" type="ReduceMax" version="opset1">
609
+ <data keep_dims="false" />
610
+ <input>
611
+ <port id="0" precision="I32">
612
+ <dim>-1</dim>
613
+ </port>
614
+ <port id="1" precision="I32" />
615
+ </input>
616
+ <output>
617
+ <port id="2" precision="I32" />
618
+ </output>
619
+ </layer>
620
+ <layer id="47" name="Constant_2236324" type="Const" version="opset1">
621
+ <data element_type="i32" shape="" offset="0" size="4" />
622
+ <output>
623
+ <port id="0" precision="I32" />
624
+ </output>
625
+ </layer>
626
+ <layer id="48" name="RaggedToDense_2236325" type="RaggedToDense" version="extension">
627
+ <data pad_right="false" m_pad_max_length="false" />
628
+ <input>
629
+ <port id="0" precision="I32">
630
+ <dim>-1</dim>
631
+ </port>
632
+ <port id="1" precision="I32">
633
+ <dim>-1</dim>
634
+ </port>
635
+ <port id="2" precision="I32">
636
+ <dim>-1</dim>
637
+ </port>
638
+ <port id="3" precision="I32" />
639
+ <port id="4" precision="I32" />
640
+ </input>
641
+ <output>
642
+ <port id="5" precision="I32">
643
+ <dim>-1</dim>
644
+ <dim>-1</dim>
645
+ </port>
646
+ <port id="6" precision="BOOL">
647
+ <dim>-1</dim>
648
+ <dim>-1</dim>
649
+ </port>
650
+ </output>
651
+ </layer>
652
+ <layer id="49" name="Convert_2236326" type="Convert" version="opset1">
653
+ <data destination_type="i32" />
654
+ <input>
655
+ <port id="0" precision="BOOL">
656
+ <dim>-1</dim>
657
+ <dim>-1</dim>
658
+ </port>
659
+ </input>
660
+ <output>
661
+ <port id="1" precision="I32">
662
+ <dim>-1</dim>
663
+ <dim>-1</dim>
664
+ </port>
665
+ </output>
666
+ </layer>
667
+ <layer id="50" name="Convert_2236326.0" type="Convert" version="opset1">
668
+ <data destination_type="i64" />
669
+ <input>
670
+ <port id="0" precision="I32">
671
+ <dim>-1</dim>
672
+ <dim>-1</dim>
673
+ </port>
674
+ </input>
675
+ <output>
676
+ <port id="1" precision="I64" names="attention_mask">
677
+ <dim>-1</dim>
678
+ <dim>-1</dim>
679
+ </port>
680
+ </output>
681
+ </layer>
682
+ <layer id="52" name="RaggedToDense_2236325.0" type="Convert" version="opset1">
683
+ <data destination_type="i64" />
684
+ <input>
685
+ <port id="0" precision="I32">
686
+ <dim>-1</dim>
687
+ <dim>-1</dim>
688
+ </port>
689
+ </input>
690
+ <output>
691
+ <port id="1" precision="I64" names="input_ids">
692
+ <dim>-1</dim>
693
+ <dim>-1</dim>
694
+ </port>
695
+ </output>
696
+ </layer>
697
+ <layer id="53" name="Result_2236329" type="Result" version="opset1" output_names="input_ids">
698
+ <input>
699
+ <port id="0" precision="I64">
700
+ <dim>-1</dim>
701
+ <dim>-1</dim>
702
+ </port>
703
+ </input>
704
+ </layer>
705
+ <layer id="51" name="Result_2236331" type="Result" version="opset1" output_names="attention_mask">
706
+ <input>
707
+ <port id="0" precision="I64">
708
+ <dim>-1</dim>
709
+ <dim>-1</dim>
710
+ </port>
711
+ </input>
712
+ </layer>
713
+ </layers>
714
+ <edges>
715
+ <edge from-layer="0" from-port="0" to-layer="5" to-port="0" />
716
+ <edge from-layer="1" from-port="0" to-layer="43" to-port="0" />
717
+ <edge from-layer="2" from-port="0" to-layer="43" to-port="1" />
718
+ <edge from-layer="3" from-port="0" to-layer="43" to-port="2" />
719
+ <edge from-layer="4" from-port="0" to-layer="11" to-port="0" />
720
+ <edge from-layer="5" from-port="1" to-layer="6" to-port="0" />
721
+ <edge from-layer="5" from-port="3" to-layer="18" to-port="4" />
722
+ <edge from-layer="5" from-port="2" to-layer="18" to-port="3" />
723
+ <edge from-layer="5" from-port="1" to-layer="18" to-port="2" />
724
+ <edge from-layer="6" from-port="1" to-layer="9" to-port="0" />
725
+ <edge from-layer="7" from-port="0" to-layer="9" to-port="1" />
726
+ <edge from-layer="8" from-port="0" to-layer="9" to-port="2" />
727
+ <edge from-layer="9" from-port="3" to-layer="14" to-port="0" />
728
+ <edge from-layer="9" from-port="3" to-layer="11" to-port="1" />
729
+ <edge from-layer="10" from-port="0" to-layer="11" to-port="2" />
730
+ <edge from-layer="11" from-port="3" to-layer="18" to-port="0" />
731
+ <edge from-layer="12" from-port="0" to-layer="16" to-port="0" />
732
+ <edge from-layer="13" from-port="0" to-layer="14" to-port="1" />
733
+ <edge from-layer="14" from-port="2" to-layer="16" to-port="1" />
734
+ <edge from-layer="15" from-port="0" to-layer="16" to-port="2" />
735
+ <edge from-layer="16" from-port="3" to-layer="18" to-port="1" />
736
+ <edge from-layer="17" from-port="0" to-layer="18" to-port="5" />
737
+ <edge from-layer="18" from-port="6" to-layer="23" to-port="0" />
738
+ <edge from-layer="18" from-port="7" to-layer="23" to-port="1" />
739
+ <edge from-layer="18" from-port="11" to-layer="21" to-port="3" />
740
+ <edge from-layer="18" from-port="8" to-layer="21" to-port="0" />
741
+ <edge from-layer="18" from-port="9" to-layer="21" to-port="1" />
742
+ <edge from-layer="18" from-port="10" to-layer="21" to-port="2" />
743
+ <edge from-layer="19" from-port="0" to-layer="21" to-port="4" />
744
+ <edge from-layer="20" from-port="0" to-layer="21" to-port="5" />
745
+ <edge from-layer="21" from-port="6" to-layer="23" to-port="2" />
746
+ <edge from-layer="21" from-port="7" to-layer="23" to-port="3" />
747
+ <edge from-layer="21" from-port="8" to-layer="23" to-port="4" />
748
+ <edge from-layer="21" from-port="9" to-layer="23" to-port="5" />
749
+ <edge from-layer="22" from-port="0" to-layer="23" to-port="6" />
750
+ <edge from-layer="23" from-port="8" to-layer="37" to-port="1" />
751
+ <edge from-layer="23" from-port="11" to-layer="37" to-port="4" />
752
+ <edge from-layer="23" from-port="10" to-layer="37" to-port="3" />
753
+ <edge from-layer="23" from-port="9" to-layer="37" to-port="2" />
754
+ <edge from-layer="23" from-port="7" to-layer="37" to-port="0" />
755
+ <edge from-layer="24" from-port="0" to-layer="37" to-port="5" />
756
+ <edge from-layer="25" from-port="0" to-layer="37" to-port="6" />
757
+ <edge from-layer="26" from-port="0" to-layer="37" to-port="7" />
758
+ <edge from-layer="27" from-port="0" to-layer="37" to-port="8" />
759
+ <edge from-layer="28" from-port="0" to-layer="37" to-port="9" />
760
+ <edge from-layer="29" from-port="0" to-layer="37" to-port="10" />
761
+ <edge from-layer="30" from-port="0" to-layer="37" to-port="11" />
762
+ <edge from-layer="31" from-port="0" to-layer="37" to-port="12" />
763
+ <edge from-layer="32" from-port="0" to-layer="37" to-port="13" />
764
+ <edge from-layer="33" from-port="0" to-layer="37" to-port="14" />
765
+ <edge from-layer="34" from-port="0" to-layer="37" to-port="15" />
766
+ <edge from-layer="35" from-port="0" to-layer="37" to-port="16" />
767
+ <edge from-layer="36" from-port="0" to-layer="37" to-port="17" />
768
+ <edge from-layer="37" from-port="20" to-layer="43" to-port="5" />
769
+ <edge from-layer="37" from-port="19" to-layer="43" to-port="4" />
770
+ <edge from-layer="37" from-port="19" to-layer="41" to-port="0" />
771
+ <edge from-layer="37" from-port="18" to-layer="38" to-port="1" />
772
+ <edge from-layer="37" from-port="19" to-layer="38" to-port="0" />
773
+ <edge from-layer="38" from-port="2" to-layer="40" to-port="0" />
774
+ <edge from-layer="39" from-port="0" to-layer="40" to-port="1" />
775
+ <edge from-layer="40" from-port="2" to-layer="41" to-port="1" />
776
+ <edge from-layer="41" from-port="2" to-layer="43" to-port="3" />
777
+ <edge from-layer="42" from-port="0" to-layer="43" to-port="6" />
778
+ <edge from-layer="43" from-port="8" to-layer="44" to-port="0" />
779
+ <edge from-layer="43" from-port="7" to-layer="44" to-port="1" />
780
+ <edge from-layer="43" from-port="7" to-layer="48" to-port="0" />
781
+ <edge from-layer="43" from-port="8" to-layer="48" to-port="1" />
782
+ <edge from-layer="43" from-port="9" to-layer="48" to-port="2" />
783
+ <edge from-layer="44" from-port="2" to-layer="46" to-port="0" />
784
+ <edge from-layer="45" from-port="0" to-layer="46" to-port="1" />
785
+ <edge from-layer="46" from-port="2" to-layer="48" to-port="3" />
786
+ <edge from-layer="47" from-port="0" to-layer="48" to-port="4" />
787
+ <edge from-layer="48" from-port="6" to-layer="49" to-port="0" />
788
+ <edge from-layer="48" from-port="5" to-layer="52" to-port="0" />
789
+ <edge from-layer="49" from-port="1" to-layer="50" to-port="0" />
790
+ <edge from-layer="50" from-port="1" to-layer="51" to-port="0" />
791
+ <edge from-layer="52" from-port="1" to-layer="53" to-port="0" />
792
+ </edges>
793
+ <rt_info>
794
+ <add_attention_mask value="True" />
795
+ <add_prefix_space />
796
+ <add_special_tokens value="True" />
797
+ <bos_token_id value="2" />
798
+ <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;" />
799
+ <clean_up_tokenization_spaces />
800
+ <detokenizer_input_type value="i64" />
801
+ <eos_token_id value="1" />
802
+ <handle_special_tokens_with_re />
803
+ <max_length />
804
+ <number_of_inputs value="1" />
805
+ <openvino_tokenizers_version value="2025.1.0.0-523-710ddf14de8" />
806
+ <openvino_version value="2025.1.0-18503-6fec06580ab-releases/2025/1" />
807
+ <original_post_processor_template value="{&quot;type&quot;: &quot;TemplateProcessing&quot;, &quot;single&quot;: [{&quot;SpecialToken&quot;: {&quot;id&quot;: &quot;&lt;bos>&quot;, &quot;type_id&quot;: 0}}, {&quot;Sequence&quot;: {&quot;id&quot;: &quot;A&quot;, &quot;type_id&quot;: 0}}], &quot;pair&quot;: [{&quot;SpecialToken&quot;: {&quot;id&quot;: &quot;&lt;bos>&quot;, &quot;type_id&quot;: 0}}, {&quot;Sequence&quot;: {&quot;id&quot;: &quot;A&quot;, &quot;type_id&quot;: 0}}, {&quot;SpecialToken&quot;: {&quot;id&quot;: &quot;&lt;bos>&quot;, &quot;type_id&quot;: 1}}, {&quot;Sequence&quot;: {&quot;id&quot;: &quot;B&quot;, &quot;type_id&quot;: 1}}], &quot;special_tokens&quot;: {&quot;&lt;bos>&quot;: {&quot;id&quot;: &quot;&lt;bos>&quot;, &quot;ids&quot;: [2], &quot;tokens&quot;: [&quot;&lt;bos>&quot;]}}}" />
808
+ <original_tokenizer_class value="&lt;class 'transformers.models.gemma.tokenization_gemma_fast.GemmaTokenizerFast'>" />
809
+ <pad_token_id value="0" />
810
+ <processed_post_processor_template value="{&quot;single&quot;: {&quot;ids&quot;: [2, -1], &quot;type_ids&quot;: [0, 0]}, &quot;pair&quot;: {&quot;ids&quot;: [2, -1, 2, -2], &quot;type_ids&quot;: [0, 0, 1, 1]}}" />
811
+ <sentencepiece_version value="0.2.0" />
812
+ <skip_special_tokens value="True" />
813
+ <streaming_detokenizer value="False" />
814
+ <tiktoken_version value="0.7.0" />
815
+ <tokenizer_output_type value="i64" />
816
+ <tokenizers_version value="0.21.1" />
817
+ <transformers_version value="4.50.2" />
818
+ <use_max_padding value="False" />
819
+ <use_sentencepiece_backend value="False" />
820
+ <utf8_replace_mode value="replace" />
821
+ <with_detokenizer value="True" />
822
+ </rt_info>
823
+ </net>
openvino_vision_embeddings_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:77ab5a87358697d20811e85f143a4033d0b3f050f8a1af5d070631ace1097541
3
+ size 425761152
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