Hugging Face's logo

ctaguchi
/
ssc-ttj-mms-model

Automatic Speech Recognition
Transformers
Safetensors
wav2vec2
Generated from Trainer
Model card Files
xet
 Community
ctaguchi commited on
Commit
7a5a0a1
·
verified ·
1 Parent(s): a8df31d

Model save

Browse files
Files changed (2) hide show
  1. README.md +47 -46
  2. model.safetensors +1 -1
README.md CHANGED
@@ -18,9 +18,9 @@ should probably proofread and complete it, then remove this comment. -->
18
 
19
  This model is a fine-tuned version of [facebook/mms-1b-all](https://huggingface.co/facebook/mms-1b-all) on an unknown dataset.
20
  It achieves the following results on the evaluation set:
21
- - Loss: 0.4346
22
- - Cer: 0.1035
23
- - Wer: 0.4959
24
 
25
  ## Model description
26
 
@@ -40,60 +40,61 @@ More information needed
40
 
41
  The following hyperparameters were used during training:
42
  - learning_rate: 0.0003
43
- - train_batch_size: 4
44
- - eval_batch_size: 4
45
  - seed: 42
46
  - gradient_accumulation_steps: 2
47
- - total_train_batch_size: 8
48
  - optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
49
  - lr_scheduler_type: linear
50
  - lr_scheduler_warmup_steps: 100
51
- - num_epochs: 5
52
  - mixed_precision_training: Native AMP
53
 
54
  ### Training results
55
 
56
  | Training Loss | Epoch | Step | Validation Loss | Cer | Wer |
57
  |:-------------:|:------:|:----:|:---------------:|:------:|:------:|
58
- | 2.9582 | 0.1269 | 200 | 2.8129 | 0.9925 | 1.0 |
59
- | 1.6728 | 0.2539 | 400 | 1.6120 | 0.4421 | 1.0 |
60
- | 1.3783 | 0.3808 | 600 | 1.4257 | 0.3708 | 0.9599 |
61
- | 1.2419 | 0.5078 | 800 | 1.3681 | 0.3729 | 0.9827 |
62
- | 1.1128 | 0.6347 | 1000 | 1.2200 | 0.3766 | 0.9535 |
63
- | 1.0767 | 0.7617 | 1200 | 0.9316 | 0.2897 | 0.9282 |
64
- | 1.0574 | 0.8886 | 1400 | 0.9615 | 0.3024 | 0.9134 |
65
- | 0.9684 | 1.0152 | 1600 | 0.6766 | 0.1867 | 0.7338 |
66
- | 0.8262 | 1.1422 | 1800 | 0.5780 | 0.1623 | 0.6825 |
67
- | 0.9177 | 1.2691 | 2000 | 0.6566 | 0.1669 | 0.6957 |
68
- | 0.8826 | 1.3961 | 2200 | 0.5821 | 0.1519 | 0.6742 |
69
- | 0.8281 | 1.5230 | 2400 | 0.5865 | 0.1506 | 0.6633 |
70
- | 0.7907 | 1.6500 | 2600 | 0.5503 | 0.1518 | 0.6641 |
71
- | 0.7862 | 1.7769 | 2800 | 0.5906 | 0.1492 | 0.6341 |
72
- | 0.7854 | 1.9038 | 3000 | 0.5223 | 0.1407 | 0.6249 |
73
- | 0.7479 | 2.0305 | 3200 | 0.5756 | 0.1582 | 0.6702 |
74
- | 0.6776 | 2.1574 | 3400 | 0.5789 | 0.1537 | 0.6523 |
75
- | 0.6779 | 2.2844 | 3600 | 0.6547 | 0.1616 | 0.6864 |
76
- | 0.6453 | 2.4113 | 3800 | 0.5342 | 0.1626 | 0.6693 |
77
- | 0.6003 | 2.5382 | 4000 | 0.5079 | 0.1355 | 0.6090 |
78
- | 0.6347 | 2.6652 | 4200 | 0.5048 | 0.1388 | 0.6018 |
79
- | 0.656 | 2.7921 | 4400 | 0.5581 | 0.1502 | 0.6410 |
80
- | 0.5906 | 2.9191 | 4600 | 0.5285 | 0.1259 | 0.5744 |
81
- | 0.5332 | 3.0457 | 4800 | 0.5069 | 0.1204 | 0.5664 |
82
- | 0.5129 | 3.1726 | 5000 | 0.4789 | 0.1206 | 0.5610 |
83
- | 0.4885 | 3.2996 | 5200 | 0.4748 | 0.1191 | 0.5575 |
84
- | 0.5258 | 3.4265 | 5400 | 0.4410 | 0.1168 | 0.5353 |
85
- | 0.4376 | 3.5535 | 5600 | 0.4344 | 0.1121 | 0.5239 |
86
- | 0.4872 | 3.6804 | 5800 | 0.4540 | 0.1130 | 0.5276 |
87
- | 0.462 | 3.8074 | 6000 | 0.4446 | 0.1127 | 0.5292 |
88
- | 0.4182 | 3.9343 | 6200 | 0.4123 | 0.1110 | 0.5264 |
89
- | 0.3969 | 4.0609 | 6400 | 0.4578 | 0.1139 | 0.5275 |
90
- | 0.3511 | 4.1879 | 6600 | 0.4524 | 0.1094 | 0.5190 |
91
- | 0.3694 | 4.3148 | 6800 | 0.4534 | 0.1109 | 0.5207 |
92
- | 0.3668 | 4.4418 | 7000 | 0.4461 | 0.1080 | 0.5092 |
93
- | 0.3945 | 4.5687 | 7200 | 0.4364 | 0.1068 | 0.5060 |
94
- | 0.3177 | 4.6957 | 7400 | 0.4470 | 0.1048 | 0.4995 |
95
- | 0.3178 | 4.8226 | 7600 | 0.4404 | 0.1035 | 0.4981 |
96
- | 0.3649 | 4.9495 | 7800 | 0.4346 | 0.1035 | 0.4959 |
 
97
 
98
 
99
  ### Framework versions
 
18
 
19
  This model is a fine-tuned version of [facebook/mms-1b-all](https://huggingface.co/facebook/mms-1b-all) on an unknown dataset.
20
  It achieves the following results on the evaluation set:
21
+ - Loss: 0.2494
22
+ - Cer: 0.0795
23
+ - Wer: 0.4288
24
 
25
  ## Model description
26
 
 
40
 
41
  The following hyperparameters were used during training:
42
  - learning_rate: 0.0003
43
+ - train_batch_size: 8
44
+ - eval_batch_size: 12
45
  - seed: 42
46
  - gradient_accumulation_steps: 2
47
+ - total_train_batch_size: 16
48
  - optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
49
  - lr_scheduler_type: linear
50
  - lr_scheduler_warmup_steps: 100
51
+ - num_epochs: 10
52
  - mixed_precision_training: Native AMP
53
 
54
  ### Training results
55
 
56
  | Training Loss | Epoch | Step | Validation Loss | Cer | Wer |
57
  |:-------------:|:------:|:----:|:---------------:|:------:|:------:|
58
+ | 1.0495 | 0.2463 | 200 | 0.3003 | 0.0858 | 0.4553 |
59
+ | 1.016 | 0.4926 | 400 | 0.2813 | 0.0833 | 0.4464 |
60
+ | 1.0172 | 0.7389 | 600 | 0.2781 | 0.0829 | 0.4442 |
61
+ | 1.0097 | 0.9852 | 800 | 0.2720 | 0.0822 | 0.4393 |
62
+ | 0.9674 | 1.2315 | 1000 | 0.2654 | 0.0817 | 0.4393 |
63
+ | 0.9503 | 1.4778 | 1200 | 0.2619 | 0.0817 | 0.4368 |
64
+ | 0.9384 | 1.7241 | 1400 | 0.2598 | 0.0809 | 0.4347 |
65
+ | 0.9441 | 1.9704 | 1600 | 0.2577 | 0.0808 | 0.4355 |
66
+ | 1.0321 | 2.2167 | 1800 | 0.2553 | 0.0805 | 0.4320 |
67
+ | 0.949 | 2.4631 | 2000 | 0.2550 | 0.0803 | 0.4326 |
68
+ | 0.9281 | 2.7094 | 2200 | 0.2542 | 0.0807 | 0.4319 |
69
+ | 0.9555 | 2.9557 | 2400 | 0.2541 | 0.0804 | 0.4311 |
70
+ | 0.9329 | 3.2020 | 2600 | 0.2535 | 0.0804 | 0.4319 |
71
+ | 0.9786 | 3.4483 | 2800 | 0.2548 | 0.0802 | 0.4314 |
72
+ | 0.9575 | 3.6946 | 3000 | 0.2531 | 0.0802 | 0.4317 |
73
+ | 0.9381 | 3.9409 | 3200 | 0.2519 | 0.0801 | 0.43 |
74
+ | 0.9561 | 4.1872 | 3400 | 0.2526 | 0.0798 | 0.4292 |
75
+ | 0.945 | 4.4335 | 3600 | 0.2530 | 0.0804 | 0.4324 |
76
+ | 0.9401 | 4.6798 | 3800 | 0.2518 | 0.0794 | 0.4265 |
77
+ | 0.926 | 4.9261 | 4000 | 0.2544 | 0.0806 | 0.4339 |
78
+ | 0.932 | 5.1724 | 4200 | 0.2509 | 0.0797 | 0.4293 |
79
+ | 0.949 | 5.4187 | 4400 | 0.2500 | 0.0797 | 0.4314 |
80
+ | 0.951 | 5.6650 | 4600 | 0.2505 | 0.0800 | 0.4332 |
81
+ | 0.9221 | 5.9113 | 4800 | 0.2488 | 0.0795 | 0.4295 |
82
+ | 0.9168 | 6.1576 | 5000 | 0.2491 | 0.0796 | 0.4293 |
83
+ | 0.9188 | 6.4039 | 5200 | 0.2503 | 0.0797 | 0.4314 |
84
+ | 0.8848 | 6.6502 | 5400 | 0.2489 | 0.0793 | 0.4294 |
85
+ | 0.9561 | 6.8966 | 5600 | 0.2480 | 0.0794 | 0.4284 |
86
+ | 0.9438 | 7.1429 | 5800 | 0.2506 | 0.0798 | 0.4305 |
87
+ | 0.9402 | 7.3892 | 6000 | 0.2508 | 0.0797 | 0.4292 |
88
+ | 0.9369 | 7.6355 | 6200 | 0.2507 | 0.0794 | 0.4278 |
89
+ | 0.9201 | 7.8818 | 6400 | 0.2504 | 0.0792 | 0.4281 |
90
+ | 0.9243 | 8.1281 | 6600 | 0.2500 | 0.0794 | 0.4284 |
91
+ | 0.9347 | 8.3744 | 6800 | 0.2484 | 0.0792 | 0.4265 |
92
+ | 0.9072 | 8.6207 | 7000 | 0.2494 | 0.0792 | 0.4275 |
93
+ | 0.9002 | 8.8670 | 7200 | 0.2493 | 0.0794 | 0.4282 |
94
+ | 0.888 | 9.1133 | 7400 | 0.2489 | 0.0793 | 0.4272 |
95
+ | 0.9198 | 9.3596 | 7600 | 0.2495 | 0.0797 | 0.4292 |
96
+ | 0.916 | 9.6059 | 7800 | 0.2495 | 0.0794 | 0.4281 |
97
+ | 0.9148 | 9.8522 | 8000 | 0.2494 | 0.0795 | 0.4288 |
98
 
99
 
100
  ### Framework versions
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8a8397e27e965a33b69e6849b5590f081aa7df6e1c053564dc15924cfe59ada5
3
  size 3858890924
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d6c16aca457b66b299048f79ce73c2afad448c6091d6975c6f074635548c5c8b
3
  size 3858890924