Hugging Face's logo

circulus
/
on-gemma2-2b-it-ov-asym-int4

OpenVINO
gemma2
Model card Files Community
rippertnt commited on
Commit
c02247e
·
verified ·
1 Parent(s): 3798428

Upload 4 files

Browse files
Files changed (4) hide show
  1. config.json +37 -0
  2. generation_config.json +12 -0
  3. openvino_tokenizer.xml +291 -0
  4. special_tokens_map.json +34 -0
config.json ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "unsloth/gemma-2-2b-it",
3
+ "architectures": [
4
+ "Gemma2ForCausalLM"
5
+ ],
6
+ "attention_bias": false,
7
+ "attention_dropout": 0.0,
8
+ "attn_logit_softcapping": 50.0,
9
+ "bos_token_id": 2,
10
+ "cache_implementation": "hybrid",
11
+ "eos_token_id": [
12
+ 1,
13
+ 107
14
+ ],
15
+ "final_logit_softcapping": 30.0,
16
+ "head_dim": 256,
17
+ "hidden_act": "gelu_pytorch_tanh",
18
+ "hidden_activation": "gelu_pytorch_tanh",
19
+ "hidden_size": 2304,
20
+ "initializer_range": 0.02,
21
+ "intermediate_size": 9216,
22
+ "max_position_embeddings": 8192,
23
+ "model_type": "gemma2",
24
+ "num_attention_heads": 8,
25
+ "num_hidden_layers": 26,
26
+ "num_key_value_heads": 4,
27
+ "pad_token_id": 0,
28
+ "query_pre_attn_scalar": 256,
29
+ "rms_norm_eps": 1e-06,
30
+ "rope_theta": 10000.0,
31
+ "sliding_window": 4096,
32
+ "torch_dtype": "bfloat16",
33
+ "transformers_version": "4.43.4",
34
+ "unsloth_version": "2024.8",
35
+ "use_cache": true,
36
+ "vocab_size": 256000
37
+ }
generation_config.json ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 2,
4
+ "cache_implementation": "hybrid",
5
+ "eos_token_id": [
6
+ 1,
7
+ 107
8
+ ],
9
+ "max_length": 8192,
10
+ "pad_token_id": 0,
11
+ "transformers_version": "4.43.4"
12
+ }
openvino_tokenizer.xml ADDED
@@ -0,0 +1,291 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <?xml version="1.0"?>
2
+ <net name="tokenizer" version="11">
3
+ <layers>
4
+ <layer id="0" name="string_input" type="Parameter" version="opset1">
5
+ <data shape="?" element_type="string" />
6
+ <output>
7
+ <port id="0" precision="STRING" names="string_input">
8
+ <dim>-1</dim>
9
+ </port>
10
+ </output>
11
+ </layer>
12
+ <layer id="1" name="Constant_197111" 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_197107" type="Const" version="opset1">
19
+ <data element_type="u8" shape="4241003" offset="4" size="4241003" />
20
+ <output>
21
+ <port id="0" precision="U8">
22
+ <dim>4241003</dim>
23
+ </port>
24
+ </output>
25
+ </layer>
26
+ <layer id="3" name="SentencepieceTokenizer_197110" type="SentencepieceTokenizer" version="extension">
27
+ <data nbest_size="0" alpha="0" add_bos="true" add_eos="false" reverse="true" />
28
+ <input>
29
+ <port id="0" precision="U8">
30
+ <dim>4241003</dim>
31
+ </port>
32
+ <port id="1" precision="STRING">
33
+ <dim>-1</dim>
34
+ </port>
35
+ </input>
36
+ <output>
37
+ <port id="2" precision="I64">
38
+ <dim>-1</dim>
39
+ <dim>2</dim>
40
+ </port>
41
+ <port id="3" precision="I32">
42
+ <dim>-1</dim>
43
+ </port>
44
+ <port id="4" precision="I64">
45
+ <dim>2</dim>
46
+ </port>
47
+ </output>
48
+ </layer>
49
+ <layer id="4" name="Broadcast_197112" type="Broadcast" version="opset3">
50
+ <data mode="numpy" />
51
+ <input>
52
+ <port id="0" precision="I32" />
53
+ <port id="1" precision="I64">
54
+ <dim>2</dim>
55
+ </port>
56
+ </input>
57
+ <output>
58
+ <port id="2" precision="I32">
59
+ <dim>-1</dim>
60
+ <dim>-1</dim>
61
+ </port>
62
+ </output>
63
+ </layer>
64
+ <layer id="5" name="Constant_197113" type="Const" version="opset1">
65
+ <data element_type="i32" shape="" offset="4241007" size="4" />
66
+ <output>
67
+ <port id="0" precision="I32" />
68
+ </output>
69
+ </layer>
70
+ <layer id="6" name="ShapeOf_197114" type="ShapeOf" version="opset3">
71
+ <data output_type="i64" />
72
+ <input>
73
+ <port id="0" precision="I32">
74
+ <dim>-1</dim>
75
+ </port>
76
+ </input>
77
+ <output>
78
+ <port id="1" precision="I64">
79
+ <dim>1</dim>
80
+ </port>
81
+ </output>
82
+ </layer>
83
+ <layer id="7" name="Broadcast_197115" type="Broadcast" version="opset3">
84
+ <data mode="numpy" />
85
+ <input>
86
+ <port id="0" precision="I32" />
87
+ <port id="1" precision="I64">
88
+ <dim>1</dim>
89
+ </port>
90
+ </input>
91
+ <output>
92
+ <port id="2" precision="I32">
93
+ <dim>-1</dim>
94
+ </port>
95
+ </output>
96
+ </layer>
97
+ <layer id="8" name="ScatterNDUpdate_197119" type="ScatterNDUpdate" version="opset4">
98
+ <input>
99
+ <port id="0" precision="I32">
100
+ <dim>-1</dim>
101
+ <dim>-1</dim>
102
+ </port>
103
+ <port id="1" precision="I64">
104
+ <dim>-1</dim>
105
+ <dim>2</dim>
106
+ </port>
107
+ <port id="2" precision="I32">
108
+ <dim>-1</dim>
109
+ </port>
110
+ </input>
111
+ <output>
112
+ <port id="3" precision="I32">
113
+ <dim>-1</dim>
114
+ <dim>-1</dim>
115
+ </port>
116
+ </output>
117
+ </layer>
118
+ <layer id="9" name="Constant_197123" type="Const" version="opset1">
119
+ <data element_type="i32" shape="1" offset="4241011" size="4" />
120
+ <output>
121
+ <port id="0" precision="I32">
122
+ <dim>1</dim>
123
+ </port>
124
+ </output>
125
+ </layer>
126
+ <layer id="10" name="Reverse_197124" type="Reverse" version="opset1">
127
+ <data mode="index" />
128
+ <input>
129
+ <port id="0" precision="I32">
130
+ <dim>-1</dim>
131
+ <dim>-1</dim>
132
+ </port>
133
+ <port id="1" precision="I32">
134
+ <dim>1</dim>
135
+ </port>
136
+ </input>
137
+ <output>
138
+ <port id="2" precision="I32">
139
+ <dim>-1</dim>
140
+ <dim>-1</dim>
141
+ </port>
142
+ </output>
143
+ </layer>
144
+ <layer id="11" name="Reverse_197124" type="Convert" version="opset1">
145
+ <data destination_type="i64" />
146
+ <input>
147
+ <port id="0" precision="I32">
148
+ <dim>-1</dim>
149
+ <dim>-1</dim>
150
+ </port>
151
+ </input>
152
+ <output>
153
+ <port id="1" precision="I64" names="attention_mask">
154
+ <dim>-1</dim>
155
+ <dim>-1</dim>
156
+ </port>
157
+ </output>
158
+ </layer>
159
+ <layer id="13" name="Constant_197120" type="Const" version="opset1">
160
+ <data element_type="i32" shape="" offset="0" size="4" />
161
+ <output>
162
+ <port id="0" precision="I32" />
163
+ </output>
164
+ </layer>
165
+ <layer id="14" name="Broadcast_197121" type="Broadcast" version="opset3">
166
+ <data mode="bidirectional" />
167
+ <input>
168
+ <port id="0" precision="I32" />
169
+ <port id="1" precision="I64">
170
+ <dim>2</dim>
171
+ </port>
172
+ </input>
173
+ <output>
174
+ <port id="2" precision="I32">
175
+ <dim>-1</dim>
176
+ <dim>-1</dim>
177
+ </port>
178
+ </output>
179
+ </layer>
180
+ <layer id="15" name="ScatterNDUpdate_197122" type="ScatterNDUpdate" version="opset4">
181
+ <input>
182
+ <port id="0" precision="I32">
183
+ <dim>-1</dim>
184
+ <dim>-1</dim>
185
+ </port>
186
+ <port id="1" precision="I64">
187
+ <dim>-1</dim>
188
+ <dim>2</dim>
189
+ </port>
190
+ <port id="2" precision="I32">
191
+ <dim>-1</dim>
192
+ </port>
193
+ </input>
194
+ <output>
195
+ <port id="3" precision="I32">
196
+ <dim>-1</dim>
197
+ <dim>-1</dim>
198
+ </port>
199
+ </output>
200
+ </layer>
201
+ <layer id="16" name="Constant_197125" type="Const" version="opset1">
202
+ <data element_type="i32" shape="1" offset="4241011" size="4" />
203
+ <output>
204
+ <port id="0" precision="I32">
205
+ <dim>1</dim>
206
+ </port>
207
+ </output>
208
+ </layer>
209
+ <layer id="17" name="Reverse_197126" type="Reverse" version="opset1">
210
+ <data mode="index" />
211
+ <input>
212
+ <port id="0" precision="I32">
213
+ <dim>-1</dim>
214
+ <dim>-1</dim>
215
+ </port>
216
+ <port id="1" precision="I32">
217
+ <dim>1</dim>
218
+ </port>
219
+ </input>
220
+ <output>
221
+ <port id="2" precision="I32">
222
+ <dim>-1</dim>
223
+ <dim>-1</dim>
224
+ </port>
225
+ </output>
226
+ </layer>
227
+ <layer id="18" name="Reverse_197126" type="Convert" version="opset1">
228
+ <data destination_type="i64" />
229
+ <input>
230
+ <port id="0" precision="I32">
231
+ <dim>-1</dim>
232
+ <dim>-1</dim>
233
+ </port>
234
+ </input>
235
+ <output>
236
+ <port id="1" precision="I64" names="input_ids">
237
+ <dim>-1</dim>
238
+ <dim>-1</dim>
239
+ </port>
240
+ </output>
241
+ </layer>
242
+ <layer id="19" name="Result_197127" type="Result" version="opset1">
243
+ <input>
244
+ <port id="0" precision="I64">
245
+ <dim>-1</dim>
246
+ <dim>-1</dim>
247
+ </port>
248
+ </input>
249
+ </layer>
250
+ <layer id="12" name="Result_197128" type="Result" version="opset1">
251
+ <input>
252
+ <port id="0" precision="I64">
253
+ <dim>-1</dim>
254
+ <dim>-1</dim>
255
+ </port>
256
+ </input>
257
+ </layer>
258
+ </layers>
259
+ <edges>
260
+ <edge from-layer="0" from-port="0" to-layer="3" to-port="1" />
261
+ <edge from-layer="1" from-port="0" to-layer="4" to-port="0" />
262
+ <edge from-layer="2" from-port="0" to-layer="3" to-port="0" />
263
+ <edge from-layer="3" from-port="4" to-layer="4" to-port="1" />
264
+ <edge from-layer="3" from-port="3" to-layer="6" to-port="0" />
265
+ <edge from-layer="3" from-port="2" to-layer="8" to-port="1" />
266
+ <edge from-layer="3" from-port="4" to-layer="14" to-port="1" />
267
+ <edge from-layer="3" from-port="2" to-layer="15" to-port="1" />
268
+ <edge from-layer="3" from-port="3" to-layer="15" to-port="2" />
269
+ <edge from-layer="4" from-port="2" to-layer="8" to-port="0" />
270
+ <edge from-layer="5" from-port="0" to-layer="7" to-port="0" />
271
+ <edge from-layer="6" from-port="1" to-layer="7" to-port="1" />
272
+ <edge from-layer="7" from-port="2" to-layer="8" to-port="2" />
273
+ <edge from-layer="8" from-port="3" to-layer="10" to-port="0" />
274
+ <edge from-layer="9" from-port="0" to-layer="10" to-port="1" />
275
+ <edge from-layer="10" from-port="2" to-layer="11" to-port="0" />
276
+ <edge from-layer="11" from-port="1" to-layer="12" to-port="0" />
277
+ <edge from-layer="13" from-port="0" to-layer="14" to-port="0" />
278
+ <edge from-layer="14" from-port="2" to-layer="15" to-port="0" />
279
+ <edge from-layer="15" from-port="3" to-layer="17" to-port="0" />
280
+ <edge from-layer="16" from-port="0" to-layer="17" to-port="1" />
281
+ <edge from-layer="17" from-port="2" to-layer="18" to-port="0" />
282
+ <edge from-layer="18" from-port="1" to-layer="19" to-port="0" />
283
+ </edges>
284
+ <rt_info>
285
+ <bos_token_id value="2" />
286
+ <chat_template value="{{ bos_token }}{% if messages[0]['role'] == 'system' %}{{ raise_exception('System role not supported') }}{% endif %}{% for message in messages %}{% if (message['role'] == 'user') != (loop.index0 % 2 == 0) %}{{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }}{% endif %}{% if (message['role'] == 'assistant') %}{% set role = 'model' %}{% else %}{% set role = message['role'] %}{% endif %}{{ '&lt;start_of_turn>' + role + '&#10;' + message['content'] | trim + '&lt;end_of_turn>&#10;' }}{% endfor %}{% if add_generation_prompt %}{{'&lt;start_of_turn>model&#10;'}}{% endif %}" />
287
+ <eos_token_id value="1" />
288
+ <original_tokenizer_class value="&lt;class 'transformers.models.gemma.tokenization_gemma_fast.GemmaTokenizerFast'>" />
289
+ <pad_token_id value="0" />
290
+ </rt_info>
291
+ </net>
special_tokens_map.json ADDED
@@ -0,0 +1,34 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<start_of_turn>",
4
+ "<end_of_turn>"
5
+ ],
6
+ "bos_token": {
7
+ "content": "<bos>",
8
+ "lstrip": false,
9
+ "normalized": false,
10
+ "rstrip": false,
11
+ "single_word": false
12
+ },
13
+ "eos_token": {
14
+ "content": "<eos>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false
19
+ },
20
+ "pad_token": {
21
+ "content": "<pad>",
22
+ "lstrip": false,
23
+ "normalized": false,
24
+ "rstrip": false,
25
+ "single_word": false
26
+ },
27
+ "unk_token": {
28
+ "content": "<unk>",
29
+ "lstrip": false,
30
+ "normalized": false,
31
+ "rstrip": false,
32
+ "single_word": false
33
+ }
34
+ }