Hugging Face's logo

StofEzz
/
mascir_fr_hubert_version1000

Automatic Speech Recognition
Transformers
PyTorch
TensorBoard
hubert
Generated from Trainer
Inference Endpoints
Model card Files Metrics Community
1
StofEzz commited on
Commit
1f413c4
•
1 Parent(s): fab221a

update model card README.md

Browse files
Files changed (1) hide show
  1. README.md +28 -53
README.md CHANGED
@@ -17,8 +17,8 @@ should probably proofread and complete it, then remove this comment. -->
17
 
18
  This model is a fine-tuned version of [facebook/hubert-large-ls960-ft](https://huggingface.co/facebook/hubert-large-ls960-ft) on an unknown dataset.
19
  It achieves the following results on the evaluation set:
20
- - Loss: 1.0469
21
- - Wer: 0.5322
22
 
23
  ## Model description
24
 
@@ -44,62 +44,37 @@ The following hyperparameters were used during training:
44
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
45
  - lr_scheduler_type: linear
46
  - lr_scheduler_warmup_steps: 1000
47
- - num_epochs: 100
48
 
49
  ### Training results
50
 
51
  | Training Loss | Epoch | Step | Validation Loss | Wer |
52
  |:-------------:|:-----:|:-----:|:---------------:|:------:|
53
- | No log | 2.0 | 250 | 3.0909 | 1.0 |
54
- | 5.0589 | 4.0 | 500 | 2.9060 | 1.0 |
55
- | 5.0589 | 6.0 | 750 | 1.3938 | 0.9789 |
56
- | 1.8801 | 8.0 | 1000 | 0.9636 | 0.8422 |
57
- | 1.8801 | 10.0 | 1250 | 0.8361 | 0.7644 |
58
- | 0.8006 | 12.0 | 1500 | 0.8474 | 0.7444 |
59
- | 0.8006 | 14.0 | 1750 | 0.8360 | 0.7078 |
60
- | 0.5307 | 16.0 | 2000 | 0.8514 | 0.6944 |
61
- | 0.5307 | 18.0 | 2250 | 0.8770 | 0.6544 |
62
- | 0.3998 | 20.0 | 2500 | 0.8200 | 0.65 |
63
- | 0.3998 | 22.0 | 2750 | 0.8362 | 0.63 |
64
- | 0.3281 | 24.0 | 3000 | 0.8933 | 0.6144 |
65
- | 0.3281 | 26.0 | 3250 | 0.9355 | 0.62 |
66
- | 0.262 | 28.0 | 3500 | 0.9134 | 0.6222 |
67
- | 0.262 | 30.0 | 3750 | 0.9302 | 0.5989 |
68
- | 0.2256 | 32.0 | 4000 | 0.9307 | 0.5856 |
69
- | 0.2256 | 34.0 | 4250 | 0.9078 | 0.6011 |
70
- | 0.2011 | 36.0 | 4500 | 0.9647 | 0.5822 |
71
- | 0.2011 | 38.0 | 4750 | 0.9252 | 0.5844 |
72
- | 0.1819 | 40.0 | 5000 | 0.9917 | 0.5711 |
73
- | 0.1819 | 42.0 | 5250 | 0.9577 | 0.5678 |
74
- | 0.1706 | 44.0 | 5500 | 1.0094 | 0.5722 |
75
- | 0.1706 | 46.0 | 5750 | 0.9774 | 0.5722 |
76
- | 0.1504 | 48.0 | 6000 | 0.9702 | 0.5456 |
77
- | 0.1504 | 50.0 | 6250 | 0.9575 | 0.5756 |
78
- | 0.1509 | 52.0 | 6500 | 0.9855 | 0.5644 |
79
- | 0.1509 | 54.0 | 6750 | 0.9429 | 0.5411 |
80
- | 0.1292 | 56.0 | 7000 | 1.0471 | 0.5644 |
81
- | 0.1292 | 58.0 | 7250 | 1.0106 | 0.5589 |
82
- | 0.1217 | 60.0 | 7500 | 1.0118 | 0.5544 |
83
- | 0.1217 | 62.0 | 7750 | 1.0415 | 0.5478 |
84
- | 0.1187 | 64.0 | 8000 | 1.0047 | 0.5489 |
85
- | 0.1187 | 66.0 | 8250 | 1.0700 | 0.5644 |
86
- | 0.1075 | 68.0 | 8500 | 1.0357 | 0.5444 |
87
- | 0.1075 | 70.0 | 8750 | 0.9647 | 0.5444 |
88
- | 0.1009 | 72.0 | 9000 | 1.0392 | 0.5489 |
89
- | 0.1009 | 74.0 | 9250 | 1.0569 | 0.5433 |
90
- | 0.0997 | 76.0 | 9500 | 1.0266 | 0.5456 |
91
- | 0.0997 | 78.0 | 9750 | 1.0328 | 0.54 |
92
- | 0.101 | 80.0 | 10000 | 1.0338 | 0.5522 |
93
- | 0.101 | 82.0 | 10250 | 1.0422 | 0.5511 |
94
- | 0.088 | 84.0 | 10500 | 1.0233 | 0.55 |
95
- | 0.088 | 86.0 | 10750 | 1.0446 | 0.5522 |
96
- | 0.0922 | 88.0 | 11000 | 1.0558 | 0.5467 |
97
- | 0.0922 | 90.0 | 11250 | 1.0405 | 0.5433 |
98
- | 0.0863 | 92.0 | 11500 | 1.0336 | 0.5322 |
99
- | 0.0863 | 94.0 | 11750 | 1.0575 | 0.5356 |
100
- | 0.0845 | 96.0 | 12000 | 1.0449 | 0.5378 |
101
- | 0.0845 | 98.0 | 12250 | 1.0482 | 0.5344 |
102
- | 0.0818 | 100.0 | 12500 | 1.0469 | 0.5322 |
103
 
104
 
105
  ### Framework versions
 
17
 
18
  This model is a fine-tuned version of [facebook/hubert-large-ls960-ft](https://huggingface.co/facebook/hubert-large-ls960-ft) on an unknown dataset.
19
  It achieves the following results on the evaluation set:
20
+ - Loss: 0.8026
21
+ - Wer: 0.5
22
 
23
  ## Model description
24
 
 
44
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
45
  - lr_scheduler_type: linear
46
  - lr_scheduler_warmup_steps: 1000
47
+ - num_epochs: 50
48
 
49
  ### Training results
50
 
51
  | Training Loss | Epoch | Step | Validation Loss | Wer |
52
  |:-------------:|:-----:|:-----:|:---------------:|:------:|
53
+ | 5.9386 | 2.0 | 500 | 2.9031 | 0.9856 |
54
+ | 2.0513 | 4.0 | 1000 | 1.0727 | 0.9144 |
55
+ | 1.0528 | 6.0 | 1500 | 0.7645 | 0.7567 |
56
+ | 0.7915 | 8.0 | 2000 | 0.6926 | 0.6744 |
57
+ | 0.6418 | 10.0 | 2500 | 0.6881 | 0.6633 |
58
+ | 0.5558 | 12.0 | 3000 | 0.6724 | 0.5978 |
59
+ | 0.4792 | 14.0 | 3500 | 0.6674 | 0.6011 |
60
+ | 0.4236 | 16.0 | 4000 | 0.6907 | 0.5778 |
61
+ | 0.3808 | 18.0 | 4500 | 0.7231 | 0.5444 |
62
+ | 0.3364 | 20.0 | 5000 | 0.7069 | 0.5456 |
63
+ | 0.3193 | 22.0 | 5500 | 0.7189 | 0.5456 |
64
+ | 0.2827 | 24.0 | 6000 | 0.7432 | 0.5322 |
65
+ | 0.2769 | 26.0 | 6500 | 0.7838 | 0.5656 |
66
+ | 0.2543 | 28.0 | 7000 | 0.8012 | 0.5333 |
67
+ | 0.2365 | 30.0 | 7500 | 0.8180 | 0.5178 |
68
+ | 0.2274 | 32.0 | 8000 | 0.7943 | 0.5233 |
69
+ | 0.2095 | 34.0 | 8500 | 0.7664 | 0.5222 |
70
+ | 0.2055 | 36.0 | 9000 | 0.7621 | 0.5122 |
71
+ | 0.2044 | 38.0 | 9500 | 0.7712 | 0.5056 |
72
+ | 0.1946 | 40.0 | 10000 | 0.7987 | 0.4989 |
73
+ | 0.1891 | 42.0 | 10500 | 0.7978 | 0.5044 |
74
+ | 0.1878 | 44.0 | 11000 | 0.7894 | 0.4967 |
75
+ | 0.1742 | 46.0 | 11500 | 0.7964 | 0.4944 |
76
+ | 0.1701 | 48.0 | 12000 | 0.7990 | 0.4956 |
77
+ | 0.163 | 50.0 | 12500 | 0.8026 | 0.5 |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
78
 
79
 
80
  ### Framework versions