Hugging Face's logo

Mihaj
/
w2v2-uralic-mbart-large-karelian

Automatic Speech Recognition
Transformers
TensorBoard
Safetensors
speech-encoder-decoder
Inference Endpoints
Model card Files Metrics Community
Mihaj commited on
Commit
1000216
•
1 Parent(s): a5723f1

Training in progress, step 100

Browse files
Files changed (17) hide show
  1. config.json +283 -0
  2. model.safetensors +3 -0
  3. runs/May07_14-42-02_maxim-ubuntu3080/events.out.tfevents.1715082162.maxim-ubuntu3080.72127.0 +3 -0
  4. runs/May07_14-51-34_maxim-ubuntu3080/events.out.tfevents.1715082778.maxim-ubuntu3080.72127.1 +3 -0
  5. runs/May07_14-57-21_maxim-ubuntu3080/events.out.tfevents.1715083129.maxim-ubuntu3080.74410.0 +3 -0
  6. runs/May07_15-52-30_maxim-ubuntu3080/events.out.tfevents.1715086437.maxim-ubuntu3080.79527.0 +3 -0
  7. runs/May07_15-59-26_maxim-ubuntu3080/events.out.tfevents.1715086858.maxim-ubuntu3080.79527.1 +3 -0
  8. runs/May07_16-37-30_maxim-ubuntu3080/events.out.tfevents.1715089139.maxim-ubuntu3080.5821.0 +3 -0
  9. runs/May13_16-02-28_maxim-ubuntu3080/events.out.tfevents.1715605516.maxim-ubuntu3080.362133.0 +3 -0
  10. runs/May13_16-18-02_maxim-ubuntu3080/events.out.tfevents.1715606427.maxim-ubuntu3080.364996.0 +3 -0
  11. runs/May13_16-26-45_maxim-ubuntu3080/events.out.tfevents.1715606962.maxim-ubuntu3080.364996.1 +3 -0
  12. runs/May13_16-32-23_maxim-ubuntu3080/events.out.tfevents.1715607300.maxim-ubuntu3080.364996.2 +3 -0
  13. runs/May13_16-38-08_maxim-ubuntu3080/events.out.tfevents.1715607641.maxim-ubuntu3080.364996.3 +3 -0
  14. runs/May13_16-45-11_maxim-ubuntu3080/events.out.tfevents.1715607991.maxim-ubuntu3080.366559.0 +3 -0
  15. runs/May13_16-51-38_maxim-ubuntu3080/events.out.tfevents.1715608380.maxim-ubuntu3080.366898.0 +3 -0
  16. runs/May13_17-03-11_maxim-ubuntu3080/events.out.tfevents.1715609090.maxim-ubuntu3080.367475.0 +3 -0
  17. training_args.bin +3 -0
config.json ADDED
@@ -0,0 +1,283 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "SpeechEncoderDecoderModel"
4
+ ],
5
+ "decoder": {
6
+ "_name_or_path": "facebook/mbart-large-50",
7
+ "_num_labels": 3,
8
+ "activation_dropout": 0.0,
9
+ "activation_function": "gelu",
10
+ "add_bias_logits": false,
11
+ "add_cross_attention": true,
12
+ "add_final_layer_norm": true,
13
+ "architectures": [
14
+ "MBartForConditionalGeneration"
15
+ ],
16
+ "attention_dropout": 0.0,
17
+ "bad_words_ids": null,
18
+ "begin_suppress_tokens": null,
19
+ "bos_token_id": 0,
20
+ "chunk_size_feed_forward": 0,
21
+ "classif_dropout": 0.0,
22
+ "classifier_dropout": 0.0,
23
+ "cross_attention_hidden_size": null,
24
+ "d_model": 1024,
25
+ "decoder_attention_heads": 16,
26
+ "decoder_ffn_dim": 4096,
27
+ "decoder_layerdrop": 0.0,
28
+ "decoder_layers": 12,
29
+ "decoder_start_token_id": 2,
30
+ "diversity_penalty": 0.0,
31
+ "do_sample": false,
32
+ "dropout": 0.1,
33
+ "early_stopping": true,
34
+ "encoder_attention_heads": 16,
35
+ "encoder_ffn_dim": 4096,
36
+ "encoder_layerdrop": 0.0,
37
+ "encoder_layers": 12,
38
+ "encoder_no_repeat_ngram_size": 0,
39
+ "eos_token_id": 2,
40
+ "exponential_decay_length_penalty": null,
41
+ "finetuning_task": null,
42
+ "forced_bos_token_id": null,
43
+ "forced_eos_token_id": 2,
44
+ "gradient_checkpointing": false,
45
+ "id2label": {
46
+ "0": "LABEL_0",
47
+ "1": "LABEL_1",
48
+ "2": "LABEL_2"
49
+ },
50
+ "init_std": 0.02,
51
+ "is_decoder": true,
52
+ "is_encoder_decoder": false,
53
+ "label2id": {
54
+ "LABEL_0": 0,
55
+ "LABEL_1": 1,
56
+ "LABEL_2": 2
57
+ },
58
+ "length_penalty": 1.0,
59
+ "max_length": 200,
60
+ "max_position_embeddings": 1024,
61
+ "min_length": 0,
62
+ "model_type": "mbart",
63
+ "no_repeat_ngram_size": 0,
64
+ "normalize_before": true,
65
+ "normalize_embedding": true,
66
+ "num_beam_groups": 1,
67
+ "num_beams": 5,
68
+ "num_hidden_layers": 12,
69
+ "num_return_sequences": 1,
70
+ "output_attentions": false,
71
+ "output_hidden_states": false,
72
+ "output_past": true,
73
+ "output_scores": false,
74
+ "pad_token_id": 1,
75
+ "prefix": null,
76
+ "problem_type": null,
77
+ "pruned_heads": {},
78
+ "remove_invalid_values": false,
79
+ "repetition_penalty": 1.0,
80
+ "return_dict": true,
81
+ "return_dict_in_generate": false,
82
+ "scale_embedding": true,
83
+ "sep_token_id": null,
84
+ "static_position_embeddings": false,
85
+ "suppress_tokens": null,
86
+ "task_specific_params": null,
87
+ "temperature": 1.0,
88
+ "tf_legacy_loss": false,
89
+ "tie_encoder_decoder": false,
90
+ "tie_word_embeddings": true,
91
+ "tokenizer_class": "MBart50Tokenizer",
92
+ "top_k": 50,
93
+ "top_p": 1.0,
94
+ "torch_dtype": null,
95
+ "torchscript": false,
96
+ "typical_p": 1.0,
97
+ "use_bfloat16": false,
98
+ "use_cache": true,
99
+ "vocab_size": 250054
100
+ },
101
+ "decoder_start_token_id": 0,
102
+ "encoder": {
103
+ "_name_or_path": "facebook/wav2vec2-large-uralic-voxpopuli-v2",
104
+ "activation_dropout": 0.0,
105
+ "adapter_attn_dim": null,
106
+ "adapter_kernel_size": 3,
107
+ "adapter_stride": 2,
108
+ "add_adapter": true,
109
+ "add_cross_attention": false,
110
+ "apply_spec_augment": true,
111
+ "architectures": [
112
+ "Wav2Vec2ForPreTraining"
113
+ ],
114
+ "attention_dropout": 0.1,
115
+ "bad_words_ids": null,
116
+ "begin_suppress_tokens": null,
117
+ "bos_token_id": 1,
118
+ "chunk_size_feed_forward": 0,
119
+ "classifier_proj_size": 256,
120
+ "codevector_dim": 768,
121
+ "contrastive_logits_temperature": 0.1,
122
+ "conv_bias": true,
123
+ "conv_dim": [
124
+ 512,
125
+ 512,
126
+ 512,
127
+ 512,
128
+ 512,
129
+ 512,
130
+ 512
131
+ ],
132
+ "conv_kernel": [
133
+ 10,
134
+ 3,
135
+ 3,
136
+ 3,
137
+ 3,
138
+ 2,
139
+ 2
140
+ ],
141
+ "conv_stride": [
142
+ 5,
143
+ 2,
144
+ 2,
145
+ 2,
146
+ 2,
147
+ 2,
148
+ 2
149
+ ],
150
+ "cross_attention_hidden_size": null,
151
+ "ctc_loss_reduction": "sum",
152
+ "ctc_zero_infinity": false,
153
+ "decoder_start_token_id": null,
154
+ "diversity_loss_weight": 0.1,
155
+ "diversity_penalty": 0.0,
156
+ "do_sample": false,
157
+ "do_stable_layer_norm": true,
158
+ "early_stopping": false,
159
+ "encoder_no_repeat_ngram_size": 0,
160
+ "eos_token_id": 2,
161
+ "exponential_decay_length_penalty": null,
162
+ "feat_extract_activation": "gelu",
163
+ "feat_extract_dropout": 0.0,
164
+ "feat_extract_norm": "layer",
165
+ "feat_proj_dropout": 0.0,
166
+ "feat_quantizer_dropout": 0.0,
167
+ "final_dropout": 0.0,
168
+ "finetuning_task": null,
169
+ "forced_bos_token_id": null,
170
+ "forced_eos_token_id": null,
171
+ "gradient_checkpointing": false,
172
+ "hidden_act": "gelu",
173
+ "hidden_dropout": 0.1,
174
+ "hidden_size": 1024,
175
+ "id2label": {
176
+ "0": "LABEL_0",
177
+ "1": "LABEL_1"
178
+ },
179
+ "initializer_range": 0.02,
180
+ "intermediate_size": 4096,
181
+ "is_decoder": false,
182
+ "is_encoder_decoder": false,
183
+ "label2id": {
184
+ "LABEL_0": 0,
185
+ "LABEL_1": 1
186
+ },
187
+ "layer_norm_eps": 1e-05,
188
+ "layerdrop": 0.0,
189
+ "length_penalty": 1.0,
190
+ "mask_channel_length": 10,
191
+ "mask_channel_min_space": 1,
192
+ "mask_channel_other": 0.0,
193
+ "mask_channel_prob": 0.0,
194
+ "mask_channel_selection": "static",
195
+ "mask_feature_length": 10,
196
+ "mask_feature_min_masks": 0,
197
+ "mask_feature_prob": 0.0,
198
+ "mask_time_length": 10,
199
+ "mask_time_min_masks": 2,
200
+ "mask_time_min_space": 1,
201
+ "mask_time_other": 0.0,
202
+ "mask_time_prob": 0.0,
203
+ "mask_time_selection": "static",
204
+ "max_length": 20,
205
+ "min_length": 0,
206
+ "model_type": "wav2vec2",
207
+ "no_repeat_ngram_size": 0,
208
+ "num_adapter_layers": 3,
209
+ "num_attention_heads": 16,
210
+ "num_beam_groups": 1,
211
+ "num_beams": 1,
212
+ "num_codevector_groups": 2,
213
+ "num_codevectors_per_group": 320,
214
+ "num_conv_pos_embedding_groups": 16,
215
+ "num_conv_pos_embeddings": 128,
216
+ "num_feat_extract_layers": 7,
217
+ "num_hidden_layers": 24,
218
+ "num_negatives": 100,
219
+ "num_return_sequences": 1,
220
+ "output_attentions": false,
221
+ "output_hidden_size": 1024,
222
+ "output_hidden_states": false,
223
+ "output_scores": false,
224
+ "pad_token_id": 0,
225
+ "prefix": null,
226
+ "problem_type": null,
227
+ "proj_codevector_dim": 768,
228
+ "pruned_heads": {},
229
+ "remove_invalid_values": false,
230
+ "repetition_penalty": 1.0,
231
+ "return_dict": true,
232
+ "return_dict_in_generate": false,
233
+ "sep_token_id": null,
234
+ "suppress_tokens": null,
235
+ "task_specific_params": null,
236
+ "tdnn_dilation": [
237
+ 1,
238
+ 2,
239
+ 3,
240
+ 1,
241
+ 1
242
+ ],
243
+ "tdnn_dim": [
244
+ 512,
245
+ 512,
246
+ 512,
247
+ 512,
248
+ 1500
249
+ ],
250
+ "tdnn_kernel": [
251
+ 5,
252
+ 3,
253
+ 3,
254
+ 1,
255
+ 1
256
+ ],
257
+ "temperature": 1.0,
258
+ "tf_legacy_loss": false,
259
+ "tie_encoder_decoder": false,
260
+ "tie_word_embeddings": true,
261
+ "tokenizer_class": null,
262
+ "top_k": 50,
263
+ "top_p": 1.0,
264
+ "torch_dtype": "float32",
265
+ "torchscript": false,
266
+ "typical_p": 1.0,
267
+ "use_bfloat16": false,
268
+ "use_weighted_layer_sum": false,
269
+ "vocab_size": 32,
270
+ "xvector_output_dim": 512
271
+ },
272
+ "eos_token_id": 2,
273
+ "is_encoder_decoder": true,
274
+ "max_length": 40,
275
+ "model_type": "speech-encoder-decoder",
276
+ "num_beams": 5,
277
+ "pad_token_id": 1,
278
+ "processor_class": "Wav2Vec2Processor",
279
+ "tie_word_embeddings": false,
280
+ "torch_dtype": "float16",
281
+ "transformers_version": "4.41.0.dev0",
282
+ "use_cache": false
283
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5348e4948e209a544914da6c10e9112e599a4786963b354b1172d2b7a3621e4a
3
+ size 2098183416
runs/May07_14-42-02_maxim-ubuntu3080/events.out.tfevents.1715082162.maxim-ubuntu3080.72127.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:840a10d98901e76edf24b1f2b65140501d3254a77a26d034ccfbe285511fc71f
3
+ size 11200
runs/May07_14-51-34_maxim-ubuntu3080/events.out.tfevents.1715082778.maxim-ubuntu3080.72127.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a9346d21e65e0b05429694892e3321ebb703effc31ac90d1e5a30057d23503a2
3
+ size 11200
runs/May07_14-57-21_maxim-ubuntu3080/events.out.tfevents.1715083129.maxim-ubuntu3080.74410.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1c96148d01e761e0982d5a4d337f50bc45b5deea959e1d1a02b67d5c578ed156
3
+ size 11198
runs/May07_15-52-30_maxim-ubuntu3080/events.out.tfevents.1715086437.maxim-ubuntu3080.79527.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6d7287ac914b588f8571091a437938e73be73351f4f285424b728f75a38e7693
3
+ size 11199
runs/May07_15-59-26_maxim-ubuntu3080/events.out.tfevents.1715086858.maxim-ubuntu3080.79527.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:53a692d6eea1c2a598c9855dbb3c5b9d119b0f7ce8d1f1edc54afceac39448bf
3
+ size 8280
runs/May07_16-37-30_maxim-ubuntu3080/events.out.tfevents.1715089139.maxim-ubuntu3080.5821.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:45cd19c375551f8420be79ed37319336d0bb0d7918cff4195ed585ee6857dfb6
3
+ size 11407
runs/May13_16-02-28_maxim-ubuntu3080/events.out.tfevents.1715605516.maxim-ubuntu3080.362133.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:292c56763fbca975a1e14e7bb88416359c9fba0a86398cba88d679d49d0fff5f
3
+ size 11186
runs/May13_16-18-02_maxim-ubuntu3080/events.out.tfevents.1715606427.maxim-ubuntu3080.364996.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bd97b842e8766dd3af24d3026415942a40aac58fe52ab188641cd9bb4aaff38a
3
+ size 11266
runs/May13_16-26-45_maxim-ubuntu3080/events.out.tfevents.1715606962.maxim-ubuntu3080.364996.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:64c6678f8367a147bff0dccbad63cd4c87b9c175db236784d081811acb8d3b07
3
+ size 11266
runs/May13_16-32-23_maxim-ubuntu3080/events.out.tfevents.1715607300.maxim-ubuntu3080.364996.2 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7db9d5060cf77064f4c7d63858fc00e7da4d43f5956e3f0c0723700991c273f4
3
+ size 11266
runs/May13_16-38-08_maxim-ubuntu3080/events.out.tfevents.1715607641.maxim-ubuntu3080.364996.3 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d4c5c6f87d5055ef2ff8693e9ba7e9d0ce0c4ff6a16a8ae740f67518a529107
3
+ size 11266
runs/May13_16-45-11_maxim-ubuntu3080/events.out.tfevents.1715607991.maxim-ubuntu3080.366559.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f7fce0c193901d2cbb25f86be41c7c75b3d63f37d2873d114058502c5ecca45b
3
+ size 11266
runs/May13_16-51-38_maxim-ubuntu3080/events.out.tfevents.1715608380.maxim-ubuntu3080.366898.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:252d7a39570a3e29e69f985e714a78c6799578f9a73c686f3da4ce5d1c43e313
3
+ size 11847
runs/May13_17-03-11_maxim-ubuntu3080/events.out.tfevents.1715609090.maxim-ubuntu3080.367475.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7df2c121f0c9d721b6e18391cabb82a4e741bcf9c4574eb83b95edab2ac787e5
3
+ size 13708
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:724f1fd851b8aa7376aad0afd115a90087ece812cd107b601d48386025935335
3
+ size 6776