Hugging Face's logo

DzmitryShchamialiou
/
whisper-tiny-int4

ONNX
Model card Files
xet
Community
DimQ1 commited on
Commit
bd2ba7f
·
1 Parent(s): 8af7f22

added model

Browse files
Files changed (3) hide show
  1. whisper_gpu_int4_gpu-cuda_metrics.json +7 -0
  2. whisper_gpu_int4_gpu-cuda_model.json +212 -0
  3. whisper_gpu_int4_gpu-cuda_model.onnx +3 -0
whisper_gpu_int4_gpu-cuda_metrics.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "latency-avg": {
3
+ "value": 404.89314,
4
+ "priority": 1,
5
+ "higher_is_better": false
6
+ }
7
+ }
whisper_gpu_int4_gpu-cuda_model.json ADDED
@@ -0,0 +1,212 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "ONNXModel",
3
+ "config": {
4
+ "model_path": "E:\\Work\\Dimq1\\source\\OpenAI\\Olive\\examples\\whisper\\models\\conversion-onnx_dynamic_quantization-insert_beam_search-prepost\\whisper_gpu_int4_gpu-cuda_model.onnx",
5
+ "onnx_file_name": null,
6
+ "inference_settings": null,
7
+ "use_ort_extensions": true,
8
+ "model_attributes": {
9
+ "vocab_size": 51865,
10
+ "num_mel_bins": 80,
11
+ "d_model": 384,
12
+ "encoder_layers": 4,
13
+ "encoder_attention_heads": 6,
14
+ "decoder_layers": 4,
15
+ "decoder_attention_heads": 6,
16
+ "decoder_ffn_dim": 1536,
17
+ "encoder_ffn_dim": 1536,
18
+ "dropout": 0.0,
19
+ "attention_dropout": 0.0,
20
+ "activation_dropout": 0.0,
21
+ "activation_function": "gelu",
22
+ "init_std": 0.02,
23
+ "encoder_layerdrop": 0.0,
24
+ "decoder_layerdrop": 0.0,
25
+ "use_cache": true,
26
+ "num_hidden_layers": 4,
27
+ "scale_embedding": false,
28
+ "max_source_positions": 1500,
29
+ "max_target_positions": 448,
30
+ "classifier_proj_size": 256,
31
+ "use_weighted_layer_sum": false,
32
+ "apply_spec_augment": false,
33
+ "mask_time_prob": 0.05,
34
+ "mask_time_length": 10,
35
+ "mask_time_min_masks": 2,
36
+ "mask_feature_prob": 0.0,
37
+ "mask_feature_length": 10,
38
+ "mask_feature_min_masks": 0,
39
+ "median_filter_width": 7,
40
+ "return_dict": true,
41
+ "output_hidden_states": false,
42
+ "output_attentions": false,
43
+ "torchscript": false,
44
+ "torch_dtype": "float32",
45
+ "use_bfloat16": false,
46
+ "tf_legacy_loss": false,
47
+ "pruned_heads": {},
48
+ "tie_word_embeddings": true,
49
+ "is_encoder_decoder": true,
50
+ "is_decoder": false,
51
+ "cross_attention_hidden_size": null,
52
+ "add_cross_attention": false,
53
+ "tie_encoder_decoder": false,
54
+ "max_length": 448,
55
+ "min_length": 0,
56
+ "do_sample": false,
57
+ "early_stopping": false,
58
+ "num_beams": 1,
59
+ "num_beam_groups": 1,
60
+ "diversity_penalty": 0.0,
61
+ "temperature": 1.0,
62
+ "top_k": 50,
63
+ "top_p": 1.0,
64
+ "typical_p": 1.0,
65
+ "repetition_penalty": 1.0,
66
+ "length_penalty": 1.0,
67
+ "no_repeat_ngram_size": 0,
68
+ "encoder_no_repeat_ngram_size": 0,
69
+ "bad_words_ids": null,
70
+ "num_return_sequences": 1,
71
+ "chunk_size_feed_forward": 0,
72
+ "output_scores": false,
73
+ "return_dict_in_generate": false,
74
+ "forced_bos_token_id": null,
75
+ "forced_eos_token_id": null,
76
+ "remove_invalid_values": false,
77
+ "exponential_decay_length_penalty": null,
78
+ "suppress_tokens": [
79
+ 1,
80
+ 2,
81
+ 7,
82
+ 8,
83
+ 9,
84
+ 10,
85
+ 14,
86
+ 25,
87
+ 26,
88
+ 27,
89
+ 28,
90
+ 29,
91
+ 31,
92
+ 58,
93
+ 59,
94
+ 60,
95
+ 61,
96
+ 62,
97
+ 63,
98
+ 90,
99
+ 91,
100
+ 92,
101
+ 93,
102
+ 359,
103
+ 503,
104
+ 522,
105
+ 542,
106
+ 873,
107
+ 893,
108
+ 902,
109
+ 918,
110
+ 922,
111
+ 931,
112
+ 1350,
113
+ 1853,
114
+ 1982,
115
+ 2460,
116
+ 2627,
117
+ 3246,
118
+ 3253,
119
+ 3268,
120
+ 3536,
121
+ 3846,
122
+ 3961,
123
+ 4183,
124
+ 4667,
125
+ 6585,
126
+ 6647,
127
+ 7273,
128
+ 9061,
129
+ 9383,
130
+ 10428,
131
+ 10929,
132
+ 11938,
133
+ 12033,
134
+ 12331,
135
+ 12562,
136
+ 13793,
137
+ 14157,
138
+ 14635,
139
+ 15265,
140
+ 15618,
141
+ 16553,
142
+ 16604,
143
+ 18362,
144
+ 18956,
145
+ 20075,
146
+ 21675,
147
+ 22520,
148
+ 26130,
149
+ 26161,
150
+ 26435,
151
+ 28279,
152
+ 29464,
153
+ 31650,
154
+ 32302,
155
+ 32470,
156
+ 36865,
157
+ 42863,
158
+ 47425,
159
+ 49870,
160
+ 50254,
161
+ 50258,
162
+ 50358,
163
+ 50359,
164
+ 50360,
165
+ 50361,
166
+ 50362
167
+ ],
168
+ "begin_suppress_tokens": [
169
+ 220,
170
+ 50257
171
+ ],
172
+ "architectures": [
173
+ "WhisperForConditionalGeneration"
174
+ ],
175
+ "finetuning_task": null,
176
+ "id2label": {
177
+ "0": "LABEL_0",
178
+ "1": "LABEL_1"
179
+ },
180
+ "label2id": {
181
+ "LABEL_0": 0,
182
+ "LABEL_1": 1
183
+ },
184
+ "tokenizer_class": null,
185
+ "prefix": null,
186
+ "bos_token_id": 50257,
187
+ "pad_token_id": 50257,
188
+ "eos_token_id": 50257,
189
+ "sep_token_id": null,
190
+ "decoder_start_token_id": 50258,
191
+ "task_specific_params": null,
192
+ "problem_type": null,
193
+ "_name_or_path": "openai/whisper-tiny",
194
+ "transformers_version": "4.35.2",
195
+ "forced_decoder_ids": [
196
+ [
197
+ 1,
198
+ 50259
199
+ ],
200
+ [
201
+ 2,
202
+ 50359
203
+ ],
204
+ [
205
+ 3,
206
+ 50363
207
+ ]
208
+ ],
209
+ "model_type": "whisper"
210
+ }
211
+ }
212
+ }
whisper_gpu_int4_gpu-cuda_model.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c6b5eb5028a05fe420d5342e41f72b0fe106f4f8b80630febe0f7960adc33090
3
+ size 112489363