Hugging Face's logo

macabdul9
/
hubert-large-ll60k-librispeech-clean-100h-demo-dist

Automatic Speech Recognition
Transformers
TensorBoard
Safetensors
hubert
Generated from Trainer
Model card Files Metrics Community
macabdul9 commited on
Commit
5ee5005
·
verified ·
1 Parent(s): f2eae1a

Model save

Browse files
Files changed (1) hide show
  1. README.md +64 -9
README.md CHANGED
@@ -17,9 +17,9 @@ should probably proofread and complete it, then remove this comment. -->
17
 
18
  This model is a fine-tuned version of [facebook/hubert-large-ll60k](https://huggingface.co/facebook/hubert-large-ll60k) on the None dataset.
19
  It achieves the following results on the evaluation set:
20
- - Loss: 2.9209
21
- - Wer: 1.0
22
- - Cer: 1.0
23
 
24
  ## Model description
25
 
@@ -39,24 +39,79 @@ More information needed
39
 
40
  The following hyperparameters were used during training:
41
  - learning_rate: 0.0003
42
- - train_batch_size: 64
43
  - eval_batch_size: 8
44
  - seed: 42
45
  - distributed_type: multi-GPU
46
  - num_devices: 8
47
- - total_train_batch_size: 512
48
  - total_eval_batch_size: 64
49
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
50
  - lr_scheduler_type: linear
51
  - lr_scheduler_warmup_steps: 500
52
- - num_epochs: 3.0
53
  - mixed_precision_training: Native AMP
54
 
55
  ### Training results
56
 
57
- | Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
58
- |:-------------:|:-----:|:----:|:---------------:|:---:|:---:|
59
- | 2.9166 | 1.79 | 100 | 2.9209 | 1.0 | 1.0 |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
60
 
61
 
62
  ### Framework versions
 
17
 
18
  This model is a fine-tuned version of [facebook/hubert-large-ll60k](https://huggingface.co/facebook/hubert-large-ll60k) on the None dataset.
19
  It achieves the following results on the evaluation set:
20
+ - Loss: 0.2143
21
+ - Wer: 0.0995
22
+ - Cer: 0.0316
23
 
24
  ## Model description
25
 
 
39
 
40
  The following hyperparameters were used during training:
41
  - learning_rate: 0.0003
42
+ - train_batch_size: 32
43
  - eval_batch_size: 8
44
  - seed: 42
45
  - distributed_type: multi-GPU
46
  - num_devices: 8
47
+ - total_train_batch_size: 256
48
  - total_eval_batch_size: 64
49
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
50
  - lr_scheduler_type: linear
51
  - lr_scheduler_warmup_steps: 500
52
+ - num_epochs: 50.0
53
  - mixed_precision_training: Native AMP
54
 
55
  ### Training results
56
 
57
+ | Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
58
+ |:-------------:|:-----:|:----:|:---------------:|:------:|:------:|
59
+ | 2.911 | 0.89 | 100 | 2.9202 | 1.0 | 1.0 |
60
+ | 2.6638 | 1.79 | 200 | 2.6310 | 1.0 | 1.0 |
61
+ | 0.3898 | 2.68 | 300 | 0.3892 | 0.3366 | 0.0968 |
62
+ | 0.2156 | 3.57 | 400 | 0.2250 | 0.2090 | 0.0591 |
63
+ | 0.1517 | 4.46 | 500 | 0.1834 | 0.1695 | 0.0474 |
64
+ | 0.1059 | 5.36 | 600 | 0.1668 | 0.1502 | 0.0428 |
65
+ | 0.0825 | 6.25 | 700 | 0.1662 | 0.1406 | 0.0393 |
66
+ | 0.0679 | 7.14 | 800 | 0.1747 | 0.1357 | 0.0393 |
67
+ | 0.0602 | 8.04 | 900 | 0.1767 | 0.1334 | 0.0390 |
68
+ | 0.0587 | 8.93 | 1000 | 0.1708 | 0.1292 | 0.0376 |
69
+ | 0.0517 | 9.82 | 1100 | 0.1677 | 0.1255 | 0.0372 |
70
+ | 0.0413 | 10.71 | 1200 | 0.1771 | 0.1234 | 0.0361 |
71
+ | 0.0418 | 11.61 | 1300 | 0.1731 | 0.1229 | 0.0358 |
72
+ | 0.0424 | 12.5 | 1400 | 0.1796 | 0.1191 | 0.0348 |
73
+ | 0.0469 | 13.39 | 1500 | 0.1848 | 0.1207 | 0.0358 |
74
+ | 0.0414 | 14.29 | 1600 | 0.1863 | 0.1213 | 0.0367 |
75
+ | 0.0338 | 15.18 | 1700 | 0.1889 | 0.1177 | 0.0347 |
76
+ | 0.0334 | 16.07 | 1800 | 0.1900 | 0.1188 | 0.0360 |
77
+ | 0.0315 | 16.96 | 1900 | 0.1901 | 0.1158 | 0.0346 |
78
+ | 0.0317 | 17.86 | 2000 | 0.1790 | 0.1134 | 0.0341 |
79
+ | 0.0264 | 18.75 | 2100 | 0.1864 | 0.1159 | 0.0356 |
80
+ | 0.0271 | 19.64 | 2200 | 0.1861 | 0.1150 | 0.0341 |
81
+ | 0.0272 | 20.54 | 2300 | 0.1945 | 0.1129 | 0.0339 |
82
+ | 0.0278 | 21.43 | 2400 | 0.1950 | 0.1131 | 0.0343 |
83
+ | 0.0254 | 22.32 | 2500 | 0.2015 | 0.1097 | 0.0330 |
84
+ | 0.0204 | 23.21 | 2600 | 0.1952 | 0.1069 | 0.0326 |
85
+ | 0.0259 | 24.11 | 2700 | 0.1976 | 0.1103 | 0.0330 |
86
+ | 0.0325 | 25.0 | 2800 | 0.1958 | 0.1088 | 0.0328 |
87
+ | 0.0359 | 25.89 | 2900 | 0.1908 | 0.1105 | 0.0346 |
88
+ | 0.0265 | 26.79 | 3000 | 0.1991 | 0.1096 | 0.0337 |
89
+ | 0.0223 | 27.68 | 3100 | 0.1948 | 0.1107 | 0.0345 |
90
+ | 0.025 | 28.57 | 3200 | 0.2046 | 0.1077 | 0.0330 |
91
+ | 0.0242 | 29.46 | 3300 | 0.2055 | 0.1072 | 0.0335 |
92
+ | 0.0187 | 30.36 | 3400 | 0.1980 | 0.1021 | 0.0307 |
93
+ | 0.0219 | 31.25 | 3500 | 0.1998 | 0.1054 | 0.0322 |
94
+ | 0.0198 | 32.14 | 3600 | 0.2104 | 0.1048 | 0.0322 |
95
+ | 0.0181 | 33.04 | 3700 | 0.2093 | 0.1050 | 0.0325 |
96
+ | 0.0166 | 33.93 | 3800 | 0.2120 | 0.1032 | 0.0315 |
97
+ | 0.0212 | 34.82 | 3900 | 0.2021 | 0.1003 | 0.0300 |
98
+ | 0.0214 | 35.71 | 4000 | 0.2045 | 0.1033 | 0.0316 |
99
+ | 0.016 | 36.61 | 4100 | 0.2022 | 0.1000 | 0.0302 |
100
+ | 0.0169 | 37.5 | 4200 | 0.2060 | 0.0996 | 0.0299 |
101
+ | 0.0191 | 38.39 | 4300 | 0.2114 | 0.1006 | 0.0307 |
102
+ | 0.0218 | 39.29 | 4400 | 0.2066 | 0.1015 | 0.0314 |
103
+ | 0.0182 | 40.18 | 4500 | 0.2054 | 0.0988 | 0.0300 |
104
+ | 0.0185 | 41.07 | 4600 | 0.2050 | 0.0994 | 0.0303 |
105
+ | 0.0171 | 41.96 | 4700 | 0.2136 | 0.0994 | 0.0306 |
106
+ | 0.0171 | 42.86 | 4800 | 0.2062 | 0.1007 | 0.0318 |
107
+ | 0.0161 | 43.75 | 4900 | 0.2101 | 0.1013 | 0.0319 |
108
+ | 0.0168 | 44.64 | 5000 | 0.2111 | 0.0985 | 0.0306 |
109
+ | 0.015 | 45.54 | 5100 | 0.2110 | 0.1003 | 0.0318 |
110
+ | 0.0126 | 46.43 | 5200 | 0.2086 | 0.0999 | 0.0319 |
111
+ | 0.0153 | 47.32 | 5300 | 0.2095 | 0.0981 | 0.0310 |
112
+ | 0.0172 | 48.21 | 5400 | 0.2130 | 0.0985 | 0.0310 |
113
+ | 0.017 | 49.11 | 5500 | 0.2137 | 0.0994 | 0.0316 |
114
+ | 0.0152 | 50.0 | 5600 | 0.2143 | 0.0995 | 0.0316 |
115
 
116
 
117
  ### Framework versions