Hugging Face's logo

shreyasdesaisuperU
/
whisper-fine-tune

Automatic Speech Recognition
Transformers
TensorBoard
Safetensors
whisper
Generated from Trainer
Inference Endpoints
Model card Files Metrics Community
shreyasdesaisuperU commited on
Commit
e6a2612
•
1 Parent(s): 53d73b9

End of training

Browse files
Files changed (1) hide show
  1. README.md +27 -47
README.md CHANGED
@@ -18,8 +18,8 @@ should probably proofread and complete it, then remove this comment. -->
18
 
19
  This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the None dataset.
20
  It achieves the following results on the evaluation set:
21
- - Loss: 6.3374
22
- - Wer: 57.2770
23
 
24
  ## Model description
25
 
@@ -42,58 +42,38 @@ The following hyperparameters were used during training:
42
  - train_batch_size: 2
43
  - eval_batch_size: 2
44
  - seed: 42
45
- - gradient_accumulation_steps: 128
46
- - total_train_batch_size: 256
47
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
48
  - lr_scheduler_type: linear
49
  - lr_scheduler_warmup_steps: 500
50
- - training_steps: 4000
51
  - mixed_precision_training: Native AMP
52
 
53
  ### Training results
54
 
55
- | Training Loss | Epoch | Step | Validation Loss | Wer |
56
- |:-------------:|:------:|:----:|:---------------:|:-------:|
57
- | 0.0 | 100.0 | 100 | 2.1912 | 59.6244 |
58
- | 0.0 | 200.0 | 200 | 2.3111 | 58.2160 |
59
- | 0.0 | 300.0 | 300 | 2.3886 | 58.2160 |
60
- | 0.0 | 400.0 | 400 | 2.5319 | 60.5634 |
61
- | 0.0 | 500.0 | 500 | 2.7160 | 60.0939 |
62
- | 0.0 | 600.0 | 600 | 2.9609 | 61.0329 |
63
- | 0.0 | 700.0 | 700 | 3.2141 | 61.0329 |
64
- | 0.0 | 800.0 | 800 | 3.4591 | 61.5023 |
65
- | 0.0 | 900.0 | 900 | 3.7213 | 62.4413 |
66
- | 0.0 | 1000.0 | 1000 | 3.9804 | 61.0329 |
67
- | 0.0 | 1100.0 | 1100 | 4.2234 | 78.4038 |
68
- | 0.0 | 1200.0 | 1200 | 4.4138 | 63.3803 |
69
- | 0.0 | 1300.0 | 1300 | 4.5889 | 77.9343 |
70
- | 0.0 | 1400.0 | 1400 | 4.7946 | 70.8920 |
71
- | 0.0 | 1500.0 | 1500 | 4.9337 | 65.7277 |
72
- | 0.0 | 1600.0 | 1600 | 5.0758 | 56.8075 |
73
- | 0.0 | 1700.0 | 1700 | 5.2692 | 56.8075 |
74
- | 0.0 | 1800.0 | 1800 | 5.4087 | 56.8075 |
75
- | 0.0 | 1900.0 | 1900 | 5.5500 | 56.8075 |
76
- | 0.0 | 2000.0 | 2000 | 5.6783 | 56.8075 |
77
- | 0.0 | 2100.0 | 2100 | 5.6287 | 56.8075 |
78
- | 0.0 | 2200.0 | 2200 | 5.6852 | 56.3380 |
79
- | 0.0 | 2300.0 | 2300 | 5.7374 | 56.3380 |
80
- | 0.0 | 2400.0 | 2400 | 5.8023 | 56.3380 |
81
- | 0.0 | 2500.0 | 2500 | 5.8672 | 57.2770 |
82
- | 0.0 | 2600.0 | 2600 | 5.9427 | 57.2770 |
83
- | 0.0 | 2700.0 | 2700 | 5.9891 | 57.2770 |
84
- | 0.0 | 2800.0 | 2800 | 6.0490 | 57.2770 |
85
- | 0.0 | 2900.0 | 2900 | 6.0639 | 57.2770 |
86
- | 0.0 | 3000.0 | 3000 | 6.1095 | 57.2770 |
87
- | 0.0 | 3100.0 | 3100 | 6.1477 | 57.2770 |
88
- | 0.0 | 3200.0 | 3200 | 6.2039 | 57.2770 |
89
- | 0.0 | 3300.0 | 3300 | 6.2346 | 57.2770 |
90
- | 0.0 | 3400.0 | 3400 | 6.2567 | 57.2770 |
91
- | 0.0 | 3500.0 | 3500 | 6.2841 | 57.2770 |
92
- | 0.0 | 3600.0 | 3600 | 6.3028 | 57.2770 |
93
- | 0.0 | 3700.0 | 3700 | 6.3029 | 57.2770 |
94
- | 0.0 | 3800.0 | 3800 | 6.3294 | 57.2770 |
95
- | 0.0 | 3900.0 | 3900 | 6.3346 | 57.2770 |
96
- | 0.0 | 4000.0 | 4000 | 6.3374 | 57.2770 |
97
 
98
 
99
  ### Framework versions
 
18
 
19
  This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the None dataset.
20
  It achieves the following results on the evaluation set:
21
+ - Loss: 3.2716
22
+ - Wer: 96.2441
23
 
24
  ## Model description
25
 
 
42
  - train_batch_size: 2
43
  - eval_batch_size: 2
44
  - seed: 42
45
+ - gradient_accumulation_steps: 8
46
+ - total_train_batch_size: 16
47
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
48
  - lr_scheduler_type: linear
49
  - lr_scheduler_warmup_steps: 500
50
+ - training_steps: 1000
51
  - mixed_precision_training: Native AMP
52
 
53
  ### Training results
54
 
55
+ | Training Loss | Epoch | Step | Validation Loss | Wer |
56
+ |:-------------:|:------:|:----:|:---------------:|:--------:|
57
+ | 0.0 | 50.0 | 50 | 3.7169 | 94.3662 |
58
+ | 0.0 | 100.0 | 100 | 3.7773 | 102.3474 |
59
+ | 0.0 | 150.0 | 150 | 3.8105 | 102.3474 |
60
+ | 0.0 | 200.0 | 200 | 3.9007 | 101.8779 |
61
+ | 0.0028 | 250.0 | 250 | 3.2200 | 90.6103 |
62
+ | 0.0 | 300.0 | 300 | 3.1754 | 93.8967 |
63
+ | 0.0 | 350.0 | 350 | 3.1945 | 96.2441 |
64
+ | 0.0 | 400.0 | 400 | 3.2104 | 96.2441 |
65
+ | 0.0 | 450.0 | 450 | 3.2225 | 96.2441 |
66
+ | 0.0 | 500.0 | 500 | 3.2327 | 96.2441 |
67
+ | 0.0 | 550.0 | 550 | 3.2437 | 96.2441 |
68
+ | 0.0 | 600.0 | 600 | 3.2501 | 96.2441 |
69
+ | 0.0 | 650.0 | 650 | 3.2550 | 96.2441 |
70
+ | 0.0 | 700.0 | 700 | 3.2601 | 96.2441 |
71
+ | 0.0 | 750.0 | 750 | 3.2634 | 96.2441 |
72
+ | 0.0 | 800.0 | 800 | 3.2663 | 96.2441 |
73
+ | 0.0 | 850.0 | 850 | 3.2691 | 96.2441 |
74
+ | 0.0 | 900.0 | 900 | 3.2717 | 96.2441 |
75
+ | 0.0 | 950.0 | 950 | 3.2719 | 96.2441 |
76
+ | 0.0 | 1000.0 | 1000 | 3.2716 | 96.2441 |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
77
 
78
 
79
  ### Framework versions