Hugging Face's logo

istomin9192
/
whisper-small-sr

Automatic Speech Recognition
Transformers
Safetensors
Serbian
whisper
Eval Results (legacy)
Model card Files
xet
 Community
1
istomin9192 commited on
Commit
9ad9647
·
verified ·
1 Parent(s): dc34af0

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +13 -11
README.md CHANGED
@@ -26,7 +26,7 @@ model-index:
26
  metrics:
27
  - name: Wer
28
  type: wer
29
- value: 0.0709
30
  library_name: transformers
31
  ---
32
 
@@ -53,11 +53,11 @@ This model was fine-tuned on a **mixture of publicly available Serbian speech co
53
 
54
  ## Training procedure
55
 
56
- - Epochs: 8
57
- - Batch size: 32
58
  - Optimizer: AdamW
59
  - LR: 6e-5 with warmup (50 steps) + cosine decay to min_lr = 1e-7
60
- - Mixed precision: bfloat16
61
  - SpecAugment: frequency + time masking
62
  - Sampling: weighted sampling across datasets
63
 
@@ -65,13 +65,15 @@ This model was fine-tuned on a **mixture of publicly available Serbian speech co
65
 
66
  | Epoch | Train loss | CV WER |
67
  |------:|------------------:|-------:|
68
- | 1 | 0.331 | 0.1562 |
69
- | 2 | 0.338 | 0.1202 |
70
- | 3 | 0.241 | 0.1062 |
71
- | 4 | 0.187 | 0.0913 |
72
- | 5 | 0.150 | 0.0853 |
73
- | 6 | 0.122 | 0.0745 |
74
- | 7 | 0.106 | 0.0709 |
 
 
75
 
76
  ## Evaluation Metrics
77
 
 
26
  metrics:
27
  - name: Wer
28
  type: wer
29
+ value: 0.065924219787
30
  library_name: transformers
31
  ---
32
 
 
53
 
54
  ## Training procedure
55
 
56
+ - Epochs: 9
57
+ - Batch size: 32 / 20
58
  - Optimizer: AdamW
59
  - LR: 6e-5 with warmup (50 steps) + cosine decay to min_lr = 1e-7
60
+ - Mixed precision: bfloat16 (fp32 in the final epoch)
61
  - SpecAugment: frequency + time masking
62
  - Sampling: weighted sampling across datasets
63
 
 
65
 
66
  | Epoch | Train loss | CV WER |
67
  |------:|------------------:|-------:|
68
+ | 1 | 0.333 | 0.1614 |
69
+ | 2 | 0.344 | 0.1278 |
70
+ | 3 | 0.251 | 0.1112 |
71
+ | 4 | 0.202 | 0.1032 |
72
+ | 5 | 0.167 | 0.0934 |
73
+ | 6 | 0.138 | 0.0790 |
74
+ | 7 | 0.118 | 0.0740 |
75
+ | 8 | 0.103 | 0.0709 |
76
+ | 9 | 0.096 | 0.0659 |
77
 
78
  ## Evaluation Metrics
79