RafatK commited on
Commit
e8eb2b7
·
verified ·
1 Parent(s): 0d88111

Upload WhisperForConditionalGeneration

Browse files
Files changed (3) hide show
  1. config.json +61 -60
  2. generation_config.json +233 -233
  3. model.safetensors +3 -0
config.json CHANGED
@@ -1,60 +1,61 @@
1
- {
2
- "activation_dropout": 0.0,
3
- "activation_function": "gelu",
4
- "apply_spec_augment": true,
5
- "architectures": [
6
- "WhisperForConditionalGeneration"
7
- ],
8
- "attention_dropout": 0.0,
9
- "begin_suppress_tokens": null,
10
- "bos_token_id": 50257,
11
- "classifier_proj_size": 256,
12
- "d_model": 1280,
13
- "decoder_attention_heads": 20,
14
- "decoder_ffn_dim": 5120,
15
- "decoder_layerdrop": 0.0,
16
- "decoder_layers": 32,
17
- "decoder_start_token_id": 50258,
18
- "dropout": 0.1,
19
- "encoder_attention_heads": 20,
20
- "encoder_ffn_dim": 5120,
21
- "encoder_layerdrop": 0.0,
22
- "encoder_layers": 32,
23
- "eos_token_id": 50257,
24
- "forced_decoder_ids": [
25
- [
26
- 1,
27
- 50318
28
- ],
29
- [
30
- 2,
31
- 50359
32
- ],
33
- [
34
- 3,
35
- 50363
36
- ]
37
- ],
38
- "init_std": 0.02,
39
- "is_encoder_decoder": true,
40
- "mask_feature_length": 10,
41
- "mask_feature_min_masks": 0,
42
- "mask_feature_prob": 0.0,
43
- "mask_time_length": 10,
44
- "mask_time_min_masks": 2,
45
- "mask_time_prob": 0.05,
46
- "max_length": null,
47
- "max_source_positions": 1500,
48
- "max_target_positions": 448,
49
- "median_filter_width": 7,
50
- "model_type": "whisper",
51
- "num_hidden_layers": 32,
52
- "num_mel_bins": 80,
53
- "pad_token_id": 50257,
54
- "scale_embedding": false,
55
- "torch_dtype": "float32",
56
- "transformers_version": "4.51.3",
57
- "use_cache": true,
58
- "use_weighted_layer_sum": false,
59
- "vocab_size": 51865
60
- }
 
 
1
+ {
2
+ "_name_or_path": "Whisper_DECODIS_data",
3
+ "activation_dropout": 0.0,
4
+ "activation_function": "gelu",
5
+ "apply_spec_augment": true,
6
+ "architectures": [
7
+ "WhisperForConditionalGeneration"
8
+ ],
9
+ "attention_dropout": 0.0,
10
+ "begin_suppress_tokens": null,
11
+ "bos_token_id": 50257,
12
+ "classifier_proj_size": 256,
13
+ "d_model": 1280,
14
+ "decoder_attention_heads": 20,
15
+ "decoder_ffn_dim": 5120,
16
+ "decoder_layerdrop": 0.0,
17
+ "decoder_layers": 32,
18
+ "decoder_start_token_id": 50258,
19
+ "dropout": 0.1,
20
+ "encoder_attention_heads": 20,
21
+ "encoder_ffn_dim": 5120,
22
+ "encoder_layerdrop": 0.0,
23
+ "encoder_layers": 32,
24
+ "eos_token_id": 50257,
25
+ "forced_decoder_ids": [
26
+ [
27
+ 1,
28
+ 50318
29
+ ],
30
+ [
31
+ 2,
32
+ 50359
33
+ ],
34
+ [
35
+ 3,
36
+ 50363
37
+ ]
38
+ ],
39
+ "init_std": 0.02,
40
+ "is_encoder_decoder": true,
41
+ "mask_feature_length": 10,
42
+ "mask_feature_min_masks": 0,
43
+ "mask_feature_prob": 0.0,
44
+ "mask_time_length": 10,
45
+ "mask_time_min_masks": 2,
46
+ "mask_time_prob": 0.05,
47
+ "max_length": null,
48
+ "max_source_positions": 1500,
49
+ "max_target_positions": 448,
50
+ "median_filter_width": 7,
51
+ "model_type": "whisper",
52
+ "num_hidden_layers": 32,
53
+ "num_mel_bins": 80,
54
+ "pad_token_id": 50257,
55
+ "scale_embedding": false,
56
+ "torch_dtype": "float16",
57
+ "transformers_version": "4.48.3",
58
+ "use_cache": true,
59
+ "use_weighted_layer_sum": false,
60
+ "vocab_size": 51865
61
+ }
generation_config.json CHANGED
@@ -1,233 +1,233 @@
1
- {
2
- "alignment_heads": [
3
- [
4
- 10,
5
- 12
6
- ],
7
- [
8
- 13,
9
- 17
10
- ],
11
- [
12
- 16,
13
- 11
14
- ],
15
- [
16
- 16,
17
- 12
18
- ],
19
- [
20
- 16,
21
- 13
22
- ],
23
- [
24
- 17,
25
- 15
26
- ],
27
- [
28
- 17,
29
- 16
30
- ],
31
- [
32
- 18,
33
- 4
34
- ],
35
- [
36
- 18,
37
- 11
38
- ],
39
- [
40
- 18,
41
- 19
42
- ],
43
- [
44
- 19,
45
- 11
46
- ],
47
- [
48
- 21,
49
- 2
50
- ],
51
- [
52
- 21,
53
- 3
54
- ],
55
- [
56
- 22,
57
- 3
58
- ],
59
- [
60
- 22,
61
- 9
62
- ],
63
- [
64
- 22,
65
- 12
66
- ],
67
- [
68
- 23,
69
- 5
70
- ],
71
- [
72
- 23,
73
- 7
74
- ],
75
- [
76
- 23,
77
- 13
78
- ],
79
- [
80
- 25,
81
- 5
82
- ],
83
- [
84
- 26,
85
- 1
86
- ],
87
- [
88
- 26,
89
- 12
90
- ],
91
- [
92
- 27,
93
- 15
94
- ]
95
- ],
96
- "begin_suppress_tokens": [
97
- 220,
98
- 50257
99
- ],
100
- "bos_token_id": 50257,
101
- "decoder_start_token_id": 50258,
102
- "eos_token_id": 50257,
103
- "forced_decoder_ids": [
104
- [
105
- 1,
106
- 50318
107
- ],
108
- [
109
- 2,
110
- 50359
111
- ],
112
- [
113
- 3,
114
- 50363
115
- ]
116
- ],
117
- "is_multilingual": true,
118
- "lang_to_id": {
119
- "<|af|>": 50327,
120
- "<|am|>": 50334,
121
- "<|ar|>": 50272,
122
- "<|as|>": 50350,
123
- "<|az|>": 50304,
124
- "<|ba|>": 50355,
125
- "<|be|>": 50330,
126
- "<|bg|>": 50292,
127
- "<|bn|>": 50302,
128
- "<|bo|>": 50347,
129
- "<|br|>": 50309,
130
- "<|bs|>": 50315,
131
- "<|ca|>": 50270,
132
- "<|cs|>": 50283,
133
- "<|cy|>": 50297,
134
- "<|da|>": 50285,
135
- "<|de|>": 50261,
136
- "<|el|>": 50281,
137
- "<|en|>": 50259,
138
- "<|es|>": 50262,
139
- "<|et|>": 50307,
140
- "<|eu|>": 50310,
141
- "<|fa|>": 50300,
142
- "<|fi|>": 50277,
143
- "<|fo|>": 50338,
144
- "<|fr|>": 50265,
145
- "<|gl|>": 50319,
146
- "<|gu|>": 50333,
147
- "<|haw|>": 50352,
148
- "<|ha|>": 50354,
149
- "<|he|>": 50279,
150
- "<|hi|>": 50276,
151
- "<|hr|>": 50291,
152
- "<|ht|>": 50339,
153
- "<|hu|>": 50286,
154
- "<|hy|>": 50312,
155
- "<|id|>": 50275,
156
- "<|is|>": 50311,
157
- "<|it|>": 50274,
158
- "<|ja|>": 50266,
159
- "<|jw|>": 50356,
160
- "<|ka|>": 50329,
161
- "<|kk|>": 50316,
162
- "<|km|>": 50323,
163
- "<|kn|>": 50306,
164
- "<|ko|>": 50264,
165
- "<|la|>": 50294,
166
- "<|lb|>": 50345,
167
- "<|ln|>": 50353,
168
- "<|lo|>": 50336,
169
- "<|lt|>": 50293,
170
- "<|lv|>": 50301,
171
- "<|mg|>": 50349,
172
- "<|mi|>": 50295,
173
- "<|mk|>": 50308,
174
- "<|ml|>": 50296,
175
- "<|mn|>": 50314,
176
- "<|mr|>": 50320,
177
- "<|ms|>": 50282,
178
- "<|mt|>": 50343,
179
- "<|my|>": 50346,
180
- "<|ne|>": 50313,
181
- "<|nl|>": 50271,
182
- "<|nn|>": 50342,
183
- "<|no|>": 50288,
184
- "<|oc|>": 50328,
185
- "<|pa|>": 50321,
186
- "<|pl|>": 50269,
187
- "<|ps|>": 50340,
188
- "<|pt|>": 50267,
189
- "<|ro|>": 50284,
190
- "<|ru|>": 50263,
191
- "<|sa|>": 50344,
192
- "<|sd|>": 50332,
193
- "<|si|>": 50322,
194
- "<|sk|>": 50298,
195
- "<|sl|>": 50305,
196
- "<|sn|>": 50324,
197
- "<|so|>": 50326,
198
- "<|sq|>": 50317,
199
- "<|sr|>": 50303,
200
- "<|su|>": 50357,
201
- "<|sv|>": 50273,
202
- "<|sw|>": 50318,
203
- "<|ta|>": 50287,
204
- "<|te|>": 50299,
205
- "<|tg|>": 50331,
206
- "<|th|>": 50289,
207
- "<|tk|>": 50341,
208
- "<|tl|>": 50348,
209
- "<|tr|>": 50268,
210
- "<|tt|>": 50351,
211
- "<|uk|>": 50280,
212
- "<|ur|>": 50290,
213
- "<|uz|>": 50337,
214
- "<|vi|>": 50278,
215
- "<|yi|>": 50335,
216
- "<|yo|>": 50325,
217
- "<|zh|>": 50260
218
- },
219
- "language": "swahili",
220
- "max_initial_timestamp_index": 50,
221
- "max_length": 448,
222
- "no_timestamps_token_id": 50363,
223
- "pad_token_id": 50257,
224
- "prev_sot_token_id": 50361,
225
- "return_timestamps": false,
226
- "suppress_tokens": [],
227
- "task": "transcribe",
228
- "task_to_id": {
229
- "transcribe": 50359,
230
- "translate": 50358
231
- },
232
- "transformers_version": "4.51.3"
233
- }
 
1
+ {
2
+ "alignment_heads": [
3
+ [
4
+ 10,
5
+ 12
6
+ ],
7
+ [
8
+ 13,
9
+ 17
10
+ ],
11
+ [
12
+ 16,
13
+ 11
14
+ ],
15
+ [
16
+ 16,
17
+ 12
18
+ ],
19
+ [
20
+ 16,
21
+ 13
22
+ ],
23
+ [
24
+ 17,
25
+ 15
26
+ ],
27
+ [
28
+ 17,
29
+ 16
30
+ ],
31
+ [
32
+ 18,
33
+ 4
34
+ ],
35
+ [
36
+ 18,
37
+ 11
38
+ ],
39
+ [
40
+ 18,
41
+ 19
42
+ ],
43
+ [
44
+ 19,
45
+ 11
46
+ ],
47
+ [
48
+ 21,
49
+ 2
50
+ ],
51
+ [
52
+ 21,
53
+ 3
54
+ ],
55
+ [
56
+ 22,
57
+ 3
58
+ ],
59
+ [
60
+ 22,
61
+ 9
62
+ ],
63
+ [
64
+ 22,
65
+ 12
66
+ ],
67
+ [
68
+ 23,
69
+ 5
70
+ ],
71
+ [
72
+ 23,
73
+ 7
74
+ ],
75
+ [
76
+ 23,
77
+ 13
78
+ ],
79
+ [
80
+ 25,
81
+ 5
82
+ ],
83
+ [
84
+ 26,
85
+ 1
86
+ ],
87
+ [
88
+ 26,
89
+ 12
90
+ ],
91
+ [
92
+ 27,
93
+ 15
94
+ ]
95
+ ],
96
+ "begin_suppress_tokens": [
97
+ 220,
98
+ 50257
99
+ ],
100
+ "bos_token_id": 50257,
101
+ "decoder_start_token_id": 50258,
102
+ "eos_token_id": 50257,
103
+ "input_ids": [
104
+ [
105
+ 1,
106
+ 50318
107
+ ],
108
+ [
109
+ 2,
110
+ 50359
111
+ ],
112
+ [
113
+ 3,
114
+ 50363
115
+ ]
116
+ ],
117
+ "is_multilingual": true,
118
+ "lang_to_id": {
119
+ "<|af|>": 50327,
120
+ "<|am|>": 50334,
121
+ "<|ar|>": 50272,
122
+ "<|as|>": 50350,
123
+ "<|az|>": 50304,
124
+ "<|ba|>": 50355,
125
+ "<|be|>": 50330,
126
+ "<|bg|>": 50292,
127
+ "<|bn|>": 50302,
128
+ "<|bo|>": 50347,
129
+ "<|br|>": 50309,
130
+ "<|bs|>": 50315,
131
+ "<|ca|>": 50270,
132
+ "<|cs|>": 50283,
133
+ "<|cy|>": 50297,
134
+ "<|da|>": 50285,
135
+ "<|de|>": 50261,
136
+ "<|el|>": 50281,
137
+ "<|en|>": 50259,
138
+ "<|es|>": 50262,
139
+ "<|et|>": 50307,
140
+ "<|eu|>": 50310,
141
+ "<|fa|>": 50300,
142
+ "<|fi|>": 50277,
143
+ "<|fo|>": 50338,
144
+ "<|fr|>": 50265,
145
+ "<|gl|>": 50319,
146
+ "<|gu|>": 50333,
147
+ "<|haw|>": 50352,
148
+ "<|ha|>": 50354,
149
+ "<|he|>": 50279,
150
+ "<|hi|>": 50276,
151
+ "<|hr|>": 50291,
152
+ "<|ht|>": 50339,
153
+ "<|hu|>": 50286,
154
+ "<|hy|>": 50312,
155
+ "<|id|>": 50275,
156
+ "<|is|>": 50311,
157
+ "<|it|>": 50274,
158
+ "<|ja|>": 50266,
159
+ "<|jw|>": 50356,
160
+ "<|ka|>": 50329,
161
+ "<|kk|>": 50316,
162
+ "<|km|>": 50323,
163
+ "<|kn|>": 50306,
164
+ "<|ko|>": 50264,
165
+ "<|la|>": 50294,
166
+ "<|lb|>": 50345,
167
+ "<|ln|>": 50353,
168
+ "<|lo|>": 50336,
169
+ "<|lt|>": 50293,
170
+ "<|lv|>": 50301,
171
+ "<|mg|>": 50349,
172
+ "<|mi|>": 50295,
173
+ "<|mk|>": 50308,
174
+ "<|ml|>": 50296,
175
+ "<|mn|>": 50314,
176
+ "<|mr|>": 50320,
177
+ "<|ms|>": 50282,
178
+ "<|mt|>": 50343,
179
+ "<|my|>": 50346,
180
+ "<|ne|>": 50313,
181
+ "<|nl|>": 50271,
182
+ "<|nn|>": 50342,
183
+ "<|no|>": 50288,
184
+ "<|oc|>": 50328,
185
+ "<|pa|>": 50321,
186
+ "<|pl|>": 50269,
187
+ "<|ps|>": 50340,
188
+ "<|pt|>": 50267,
189
+ "<|ro|>": 50284,
190
+ "<|ru|>": 50263,
191
+ "<|sa|>": 50344,
192
+ "<|sd|>": 50332,
193
+ "<|si|>": 50322,
194
+ "<|sk|>": 50298,
195
+ "<|sl|>": 50305,
196
+ "<|sn|>": 50324,
197
+ "<|so|>": 50326,
198
+ "<|sq|>": 50317,
199
+ "<|sr|>": 50303,
200
+ "<|su|>": 50357,
201
+ "<|sv|>": 50273,
202
+ "<|sw|>": 50318,
203
+ "<|ta|>": 50287,
204
+ "<|te|>": 50299,
205
+ "<|tg|>": 50331,
206
+ "<|th|>": 50289,
207
+ "<|tk|>": 50341,
208
+ "<|tl|>": 50348,
209
+ "<|tr|>": 50268,
210
+ "<|tt|>": 50351,
211
+ "<|uk|>": 50280,
212
+ "<|ur|>": 50290,
213
+ "<|uz|>": 50337,
214
+ "<|vi|>": 50278,
215
+ "<|yi|>": 50335,
216
+ "<|yo|>": 50325,
217
+ "<|zh|>": 50260
218
+ },
219
+ "language": "swahili",
220
+ "max_initial_timestamp_index": 50,
221
+ "max_length": 448,
222
+ "no_timestamps_token_id": 50363,
223
+ "pad_token_id": 50257,
224
+ "prev_sot_token_id": 50361,
225
+ "return_timestamps": false,
226
+ "suppress_tokens": [],
227
+ "task": "transcribe",
228
+ "task_to_id": {
229
+ "transcribe": 50359,
230
+ "translate": 50358
231
+ },
232
+ "transformers_version": "4.48.3"
233
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61c8cbbc1f070af8b5ad056f36cdbb79fb90ab9f643dc6809f9b2494dd73ad39
3
+ size 3086759768