waveletdeboshir commited on
Commit
856646d
1 Parent(s): 3eee831

Remove additional tokens

Browse files
README.md CHANGED
@@ -1,13 +1,6 @@
1
  ---
2
- license: apache-2.0
3
- library_name: transformers
4
- pipeline_tag: automatic-speech-recognition
5
- tags:
6
- - asr
7
- - Pytorch
8
- - pruned
9
- - audio
10
- - automatic-speech-recognition
11
  language:
12
  - en
13
  - zh
@@ -108,8 +101,15 @@ language:
108
  - ba
109
  - jw
110
  - su
111
- base_model:
112
- - openai/whisper-large-v3
 
 
 
 
 
 
 
113
  ---
114
 
115
  # Whisper-large-v3-no-numbers
 
1
  ---
2
+ base_model:
3
+ - openai/whisper-large-v3
 
 
 
 
 
 
 
4
  language:
5
  - en
6
  - zh
 
101
  - ba
102
  - jw
103
  - su
104
+ library_name: transformers
105
+ license: apache-2.0
106
+ pipeline_tag: automatic-speech-recognition
107
+ tags:
108
+ - asr
109
+ - Pytorch
110
+ - pruned
111
+ - audio
112
+ - automatic-speech-recognition
113
  ---
114
 
115
  # Whisper-large-v3-no-numbers
config.json CHANGED
@@ -9,34 +9,34 @@
9
  "attention_dropout": 0.0,
10
  "begin_suppress_tokens": [
11
  210,
12
- 49839
13
  ],
14
- "bos_token_id": 49839,
15
  "classifier_proj_size": 256,
16
  "d_model": 1280,
17
  "decoder_attention_heads": 20,
18
  "decoder_ffn_dim": 5120,
19
  "decoder_layerdrop": 0.0,
20
  "decoder_layers": 32,
21
- "decoder_start_token_id": 49840,
22
  "dropout": 0.0,
23
  "encoder_attention_heads": 20,
24
  "encoder_ffn_dim": 5120,
25
  "encoder_layerdrop": 0.0,
26
  "encoder_layers": 32,
27
- "eos_token_id": 49839,
28
  "forced_decoder_ids": [
29
  [
30
  1,
31
- 49841
32
  ],
33
  [
34
  2,
35
- 49941
36
  ],
37
  [
38
  3,
39
- 49945
40
  ]
41
  ],
42
  "init_std": 0.02,
@@ -54,12 +54,12 @@
54
  "model_type": "whisper",
55
  "num_hidden_layers": 32,
56
  "num_mel_bins": 128,
57
- "pad_token_id": 49838,
58
  "scale_embedding": false,
59
  "suppress_tokens": [],
60
  "torch_dtype": "float32",
61
  "transformers_version": "4.44.0",
62
  "use_cache": true,
63
  "use_weighted_layer_sum": false,
64
- "vocab_size": 51447
65
  }
 
9
  "attention_dropout": 0.0,
10
  "begin_suppress_tokens": [
11
  210,
12
+ 49826
13
  ],
14
+ "bos_token_id": 49826,
15
  "classifier_proj_size": 256,
16
  "d_model": 1280,
17
  "decoder_attention_heads": 20,
18
  "decoder_ffn_dim": 5120,
19
  "decoder_layerdrop": 0.0,
20
  "decoder_layers": 32,
21
+ "decoder_start_token_id": 49827,
22
  "dropout": 0.0,
23
  "encoder_attention_heads": 20,
24
  "encoder_ffn_dim": 5120,
25
  "encoder_layerdrop": 0.0,
26
  "encoder_layers": 32,
27
+ "eos_token_id": 49826,
28
  "forced_decoder_ids": [
29
  [
30
  1,
31
+ 49828
32
  ],
33
  [
34
  2,
35
+ 49928
36
  ],
37
  [
38
  3,
39
+ 49932
40
  ]
41
  ],
42
  "init_std": 0.02,
 
54
  "model_type": "whisper",
55
  "num_hidden_layers": 32,
56
  "num_mel_bins": 128,
57
+ "pad_token_id": 49825,
58
  "scale_embedding": false,
59
  "suppress_tokens": [],
60
  "torch_dtype": "float32",
61
  "transformers_version": "4.44.0",
62
  "use_cache": true,
63
  "use_weighted_layer_sum": false,
64
+ "vocab_size": 51434
65
  }
generation_config.json CHANGED
@@ -43,11 +43,11 @@
43
  ],
44
  "begin_suppress_tokens": [
45
  210,
46
- 49839
47
  ],
48
- "bos_token_id": 49839,
49
- "decoder_start_token_id": 49840,
50
- "eos_token_id": 49839,
51
  "forced_decoder_ids": [
52
  [
53
  1,
@@ -55,122 +55,122 @@
55
  ],
56
  [
57
  2,
58
- 49941
59
  ]
60
  ],
61
  "is_multilingual": true,
62
  "lang_to_id": {
63
- "<|af|>": 49909,
64
- "<|am|>": 49916,
65
- "<|ar|>": 49854,
66
- "<|as|>": 49932,
67
- "<|az|>": 49886,
68
- "<|ba|>": 49937,
69
- "<|be|>": 49912,
70
- "<|bg|>": 49874,
71
- "<|bn|>": 49884,
72
- "<|bo|>": 49929,
73
- "<|br|>": 49891,
74
- "<|bs|>": 49897,
75
- "<|ca|>": 49852,
76
- "<|cs|>": 49865,
77
- "<|cy|>": 49879,
78
- "<|da|>": 49867,
79
- "<|de|>": 49843,
80
- "<|el|>": 49863,
81
- "<|en|>": 49841,
82
- "<|es|>": 49844,
83
- "<|et|>": 49889,
84
- "<|eu|>": 49892,
85
- "<|fa|>": 49882,
86
- "<|fi|>": 49859,
87
- "<|fo|>": 49920,
88
- "<|fr|>": 49847,
89
- "<|gl|>": 49901,
90
- "<|gu|>": 49915,
91
- "<|haw|>": 49934,
92
- "<|ha|>": 49936,
93
- "<|he|>": 49861,
94
- "<|hi|>": 49858,
95
- "<|hr|>": 49873,
96
- "<|ht|>": 49921,
97
- "<|hu|>": 49868,
98
- "<|hy|>": 49894,
99
- "<|id|>": 49857,
100
- "<|is|>": 49893,
101
- "<|it|>": 49856,
102
- "<|ja|>": 49848,
103
- "<|jw|>": 49938,
104
- "<|ka|>": 49911,
105
- "<|kk|>": 49898,
106
- "<|km|>": 49905,
107
- "<|kn|>": 49888,
108
- "<|ko|>": 49846,
109
- "<|la|>": 49876,
110
- "<|lb|>": 49927,
111
- "<|ln|>": 49935,
112
- "<|lo|>": 49918,
113
- "<|lt|>": 49875,
114
- "<|lv|>": 49883,
115
- "<|mg|>": 49931,
116
- "<|mi|>": 49877,
117
- "<|mk|>": 49890,
118
- "<|ml|>": 49878,
119
- "<|mn|>": 49896,
120
- "<|mr|>": 49902,
121
- "<|ms|>": 49864,
122
- "<|mt|>": 49925,
123
- "<|my|>": 49928,
124
- "<|ne|>": 49895,
125
- "<|nl|>": 49853,
126
- "<|nn|>": 49924,
127
- "<|no|>": 49870,
128
- "<|oc|>": 49910,
129
- "<|pa|>": 49903,
130
- "<|pl|>": 49851,
131
- "<|ps|>": 49922,
132
- "<|pt|>": 49849,
133
- "<|ro|>": 49866,
134
- "<|ru|>": 49845,
135
- "<|sa|>": 49926,
136
- "<|sd|>": 49914,
137
- "<|si|>": 49904,
138
- "<|sk|>": 49880,
139
- "<|sl|>": 49887,
140
- "<|sn|>": 49906,
141
- "<|so|>": 49908,
142
- "<|sq|>": 49899,
143
- "<|sr|>": 49885,
144
- "<|su|>": 49939,
145
- "<|sv|>": 49855,
146
- "<|sw|>": 49900,
147
- "<|ta|>": 49869,
148
- "<|te|>": 49881,
149
- "<|tg|>": 49913,
150
- "<|th|>": 49871,
151
- "<|tk|>": 49923,
152
- "<|tl|>": 49930,
153
- "<|tr|>": 49850,
154
- "<|tt|>": 49933,
155
- "<|uk|>": 49862,
156
- "<|ur|>": 49872,
157
- "<|uz|>": 49919,
158
- "<|vi|>": 49860,
159
- "<|yi|>": 49917,
160
- "<|yo|>": 49907,
161
- "<|yue|>": 49940,
162
- "<|zh|>": 49842
163
  },
164
  "max_initial_timestamp_index": 50,
165
  "max_length": 448,
166
- "no_timestamps_token_id": 49946,
167
- "pad_token_id": 49839,
168
- "prev_sot_token_id": 49944,
169
  "return_timestamps": false,
170
  "suppress_tokens": [],
171
  "task_to_id": {
172
- "transcribe": 49942,
173
- "translate": 49941
174
  },
175
  "transformers_version": "4.44.0"
176
  }
 
43
  ],
44
  "begin_suppress_tokens": [
45
  210,
46
+ 49826
47
  ],
48
+ "bos_token_id": 49826,
49
+ "decoder_start_token_id": 49827,
50
+ "eos_token_id": 49826,
51
  "forced_decoder_ids": [
52
  [
53
  1,
 
55
  ],
56
  [
57
  2,
58
+ 49928
59
  ]
60
  ],
61
  "is_multilingual": true,
62
  "lang_to_id": {
63
+ "<|af|>": 49896,
64
+ "<|am|>": 49903,
65
+ "<|ar|>": 49841,
66
+ "<|as|>": 49919,
67
+ "<|az|>": 49873,
68
+ "<|ba|>": 49924,
69
+ "<|be|>": 49899,
70
+ "<|bg|>": 49861,
71
+ "<|bn|>": 49871,
72
+ "<|bo|>": 49916,
73
+ "<|br|>": 49878,
74
+ "<|bs|>": 49884,
75
+ "<|ca|>": 49839,
76
+ "<|cs|>": 49852,
77
+ "<|cy|>": 49866,
78
+ "<|da|>": 49854,
79
+ "<|de|>": 49830,
80
+ "<|el|>": 49850,
81
+ "<|en|>": 49828,
82
+ "<|es|>": 49831,
83
+ "<|et|>": 49876,
84
+ "<|eu|>": 49879,
85
+ "<|fa|>": 49869,
86
+ "<|fi|>": 49846,
87
+ "<|fo|>": 49907,
88
+ "<|fr|>": 49834,
89
+ "<|gl|>": 49888,
90
+ "<|gu|>": 49902,
91
+ "<|haw|>": 49921,
92
+ "<|ha|>": 49923,
93
+ "<|he|>": 49848,
94
+ "<|hi|>": 49845,
95
+ "<|hr|>": 49860,
96
+ "<|ht|>": 49908,
97
+ "<|hu|>": 49855,
98
+ "<|hy|>": 49881,
99
+ "<|id|>": 49844,
100
+ "<|is|>": 49880,
101
+ "<|it|>": 49843,
102
+ "<|ja|>": 49835,
103
+ "<|jw|>": 49925,
104
+ "<|ka|>": 49898,
105
+ "<|kk|>": 49885,
106
+ "<|km|>": 49892,
107
+ "<|kn|>": 49875,
108
+ "<|ko|>": 49833,
109
+ "<|la|>": 49863,
110
+ "<|lb|>": 49914,
111
+ "<|ln|>": 49922,
112
+ "<|lo|>": 49905,
113
+ "<|lt|>": 49862,
114
+ "<|lv|>": 49870,
115
+ "<|mg|>": 49918,
116
+ "<|mi|>": 49864,
117
+ "<|mk|>": 49877,
118
+ "<|ml|>": 49865,
119
+ "<|mn|>": 49883,
120
+ "<|mr|>": 49889,
121
+ "<|ms|>": 49851,
122
+ "<|mt|>": 49912,
123
+ "<|my|>": 49915,
124
+ "<|ne|>": 49882,
125
+ "<|nl|>": 49840,
126
+ "<|nn|>": 49911,
127
+ "<|no|>": 49857,
128
+ "<|oc|>": 49897,
129
+ "<|pa|>": 49890,
130
+ "<|pl|>": 49838,
131
+ "<|ps|>": 49909,
132
+ "<|pt|>": 49836,
133
+ "<|ro|>": 49853,
134
+ "<|ru|>": 49832,
135
+ "<|sa|>": 49913,
136
+ "<|sd|>": 49901,
137
+ "<|si|>": 49891,
138
+ "<|sk|>": 49867,
139
+ "<|sl|>": 49874,
140
+ "<|sn|>": 49893,
141
+ "<|so|>": 49895,
142
+ "<|sq|>": 49886,
143
+ "<|sr|>": 49872,
144
+ "<|su|>": 49926,
145
+ "<|sv|>": 49842,
146
+ "<|sw|>": 49887,
147
+ "<|ta|>": 49856,
148
+ "<|te|>": 49868,
149
+ "<|tg|>": 49900,
150
+ "<|th|>": 49858,
151
+ "<|tk|>": 49910,
152
+ "<|tl|>": 49917,
153
+ "<|tr|>": 49837,
154
+ "<|tt|>": 49920,
155
+ "<|uk|>": 49849,
156
+ "<|ur|>": 49859,
157
+ "<|uz|>": 49906,
158
+ "<|vi|>": 49847,
159
+ "<|yi|>": 49904,
160
+ "<|yo|>": 49894,
161
+ "<|yue|>": 49927,
162
+ "<|zh|>": 49829
163
  },
164
  "max_initial_timestamp_index": 50,
165
  "max_length": 448,
166
+ "no_timestamps_token_id": 49933,
167
+ "pad_token_id": 49826,
168
+ "prev_sot_token_id": 49931,
169
  "return_timestamps": false,
170
  "suppress_tokens": [],
171
  "task_to_id": {
172
+ "transcribe": 49929,
173
+ "translate": 49928
174
  },
175
  "transformers_version": "4.44.0"
176
  }
model-00001-of-00002.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6269c5b070d7b142a9ad236f3aa540d51e3513de94aa4ce2fb4cc3cabbb2e147
3
- size 4991303600
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:39e3f1a1f4283821eb92f0185c06d78ec703a342f350a2095731dbb7c8b2d215
3
+ size 4991237040
model.safetensors.index.json CHANGED
@@ -1,6 +1,6 @@
1
  {
2
  "metadata": {
3
- "total_size": 6171816960
4
  },
5
  "weight_map": {
6
  "model.decoder.embed_positions.weight": "model-00001-of-00002.safetensors",
 
1
  {
2
  "metadata": {
3
+ "total_size": 6171750400
4
  },
5
  "weight_map": {
6
  "model.decoder.embed_positions.weight": "model-00001-of-00002.safetensors",