Hugging Face's logo

kiranpantha
/
whisper-large-v3-nepali

Automatic Speech Recognition
Transformers
TensorBoard
Safetensors
Nepali
whisper
Generated from Trainer
Eval Results
Inference Endpoints
Model card Files Metrics Community
kiranpantha commited on
Commit
a3daa47
·
verified ·
1 Parent(s): a669782

End of training

Browse files
Files changed (4) hide show
  1. README.md +26 -26
  2. generation_config.json +1 -11
  3. model-00001-of-00002.safetensors +1 -1
  4. model-00002-of-00002.safetensors +1 -1
README.md CHANGED
@@ -11,7 +11,7 @@ datasets:
11
  metrics:
12
  - wer
13
  model-index:
14
- - name: Whisper Large v3 Nepali - Kiran Pantha
15
  results:
16
  - task:
17
  name: Automatic Speech Recognition
@@ -25,19 +25,19 @@ model-index:
25
  metrics:
26
  - name: Wer
27
  type: wer
28
- value: 20.48611111111111
29
  ---
30
 
31
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
32
  should probably proofread and complete it, then remove this comment. -->
33
 
34
- # Whisper Large v3 Nepali - Kiran Pantha
35
 
36
  This model is a fine-tuned version of [openai/whisper-large-v3](https://huggingface.co/openai/whisper-large-v3) on the OpenSLR54 dataset.
37
  It achieves the following results on the evaluation set:
38
- - Loss: 0.1507
39
- - Wer: 20.4861
40
- - Cer: 4.9839
41
 
42
  ## Model description
43
 
@@ -57,10 +57,10 @@ More information needed
57
 
58
  The following hyperparameters were used during training:
59
  - learning_rate: 1e-05
60
- - train_batch_size: 16
61
  - eval_batch_size: 8
62
  - seed: 42
63
- - optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
64
  - lr_scheduler_type: linear
65
  - lr_scheduler_warmup_steps: 500
66
  - training_steps: 5000
@@ -70,27 +70,27 @@ The following hyperparameters were used during training:
70
 
71
  | Training Loss | Epoch | Step | Cer | Validation Loss | Wer |
72
  |:-------------:|:------:|:----:|:-------:|:---------------:|:-------:|
73
- | 0.209 | 0.3597 | 300 | 10.9916 | 0.2021 | 41.1574 |
74
- | 0.1714 | 0.7194 | 600 | 8.9474 | 0.1755 | 35.7176 |
75
- | 0.101 | 1.0791 | 900 | 7.5565 | 0.1485 | 29.6991 |
76
- | 0.0902 | 1.4388 | 1200 | 7.2372 | 0.1396 | 28.2407 |
77
- | 0.0872 | 1.7986 | 1500 | 7.8024 | 0.1319 | 27.9861 |
78
- | 0.0453 | 2.1583 | 1800 | 6.3344 | 0.1374 | 26.2269 |
79
- | 0.0368 | 2.5180 | 2100 | 6.1766 | 0.1381 | 25.2315 |
80
- | 0.0472 | 2.8777 | 2400 | 5.8316 | 0.1316 | 24.1435 |
81
- | 0.0191 | 3.2374 | 2700 | 5.8059 | 0.1356 | 24.0278 |
82
- | 0.0185 | 3.5971 | 3000 | 5.5674 | 0.1376 | 23.125 |
83
- | 0.0182 | 3.9568 | 3300 | 5.5123 | 0.1360 | 23.0556 |
84
- | 0.0074 | 4.3165 | 3600 | 5.2077 | 0.1428 | 21.7130 |
85
- | 0.0086 | 4.6763 | 3900 | 5.1784 | 0.1433 | 21.2731 |
86
- | 0.0031 | 5.0360 | 4200 | 0.1421 | 21.1806 | 5.0279 |
87
- | 0.0024 | 5.3957 | 4500 | 0.1482 | 20.7870 | 4.9912 |
88
- | 0.0014 | 5.7554 | 4800 | 0.1507 | 20.4861 | 4.9839 |
89
 
90
 
91
  ### Framework versions
92
 
93
- - Transformers 4.46.3
94
  - Pytorch 2.5.1+cxx11.abi
95
  - Datasets 3.2.0
96
- - Tokenizers 0.20.3
 
11
  metrics:
12
  - wer
13
  model-index:
14
+ - name: Whisper Large v3 Turbo Nepali - Kiran Pantha
15
  results:
16
  - task:
17
  name: Automatic Speech Recognition
 
25
  metrics:
26
  - name: Wer
27
  type: wer
28
+ value: 18.72503840245776
29
  ---
30
 
31
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
32
  should probably proofread and complete it, then remove this comment. -->
33
 
34
+ # Whisper Large v3 Turbo Nepali - Kiran Pantha
35
 
36
  This model is a fine-tuned version of [openai/whisper-large-v3](https://huggingface.co/openai/whisper-large-v3) on the OpenSLR54 dataset.
37
  It achieves the following results on the evaluation set:
38
+ - Loss: 0.0876
39
+ - Wer: 18.7250
40
+ - Cer: 4.4861
41
 
42
  ## Model description
43
 
 
57
 
58
  The following hyperparameters were used during training:
59
  - learning_rate: 1e-05
60
+ - train_batch_size: 8
61
  - eval_batch_size: 8
62
  - seed: 42
63
+ - optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
64
  - lr_scheduler_type: linear
65
  - lr_scheduler_warmup_steps: 500
66
  - training_steps: 5000
 
70
 
71
  | Training Loss | Epoch | Step | Cer | Validation Loss | Wer |
72
  |:-------------:|:------:|:----:|:-------:|:---------------:|:-------:|
73
+ | 0.2266 | 0.1200 | 300 | 11.9034 | 0.2345 | 44.7619 |
74
+ | 0.208 | 0.2399 | 600 | 11.3157 | 0.2132 | 41.1060 |
75
+ | 0.185 | 0.3599 | 900 | 9.4204 | 0.1753 | 35.6068 |
76
+ | 0.1567 | 0.4798 | 1200 | 8.8596 | 0.1634 | 33.9324 |
77
+ | 0.1411 | 0.5998 | 1500 | 8.7004 | 0.1523 | 33.0568 |
78
+ | 0.1377 | 0.7197 | 1800 | 7.3120 | 0.1371 | 29.7849 |
79
+ | 0.1147 | 0.8397 | 2100 | 7.0010 | 0.1332 | 27.7112 |
80
+ | 0.1116 | 0.9596 | 2400 | 6.5798 | 0.1212 | 26.3287 |
81
+ | 0.0757 | 1.0796 | 2700 | 6.1268 | 0.1193 | 24.7773 |
82
+ | 0.0609 | 1.1995 | 3000 | 5.8991 | 0.1154 | 24.6237 |
83
+ | 0.0612 | 1.3195 | 3300 | 5.2599 | 0.1091 | 22.0737 |
84
+ | 0.0627 | 1.4394 | 3600 | 5.3579 | 0.1045 | 21.6283 |
85
+ | 0.0582 | 1.5594 | 3900 | 5.1938 | 0.0995 | 21.5054 |
86
+ | 0.0551 | 1.6793 | 4200 | 4.7947 | 0.0956 | 19.8771 |
87
+ | 0.052 | 1.7993 | 4500 | 4.5473 | 0.0897 | 19.1244 |
88
+ | 0.0438 | 1.9192 | 4800 | 0.0876 | 18.7250 | 4.4861 |
89
 
90
 
91
  ### Framework versions
92
 
93
+ - Transformers 4.47.1
94
  - Pytorch 2.5.1+cxx11.abi
95
  - Datasets 3.2.0
96
+ - Tokenizers 0.21.0
generation_config.json CHANGED
@@ -48,16 +48,6 @@
48
  "bos_token_id": 50257,
49
  "decoder_start_token_id": 50258,
50
  "eos_token_id": 50257,
51
- "forced_decoder_ids": [
52
- [
53
- 1,
54
- null
55
- ],
56
- [
57
- 2,
58
- 50360
59
- ]
60
- ],
61
  "is_multilingual": true,
62
  "lang_to_id": {
63
  "<|af|>": 50327,
@@ -263,5 +253,5 @@
263
  "transcribe": 50360,
264
  "translate": 50359
265
  },
266
- "transformers_version": "4.46.3"
267
  }
 
48
  "bos_token_id": 50257,
49
  "decoder_start_token_id": 50258,
50
  "eos_token_id": 50257,
 
 
 
 
 
 
 
 
 
 
51
  "is_multilingual": true,
52
  "lang_to_id": {
53
  "<|af|>": 50327,
 
253
  "transcribe": 50360,
254
  "translate": 50359
255
  },
256
+ "transformers_version": "4.47.1"
257
  }
model-00001-of-00002.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5bba7fa23d3a06d94ee319279e4e28ea4089eb212f82300fe72da6bfd14f5ed1
3
  size 4993448880
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6f883dcabd631321110701e05f61994998c3ca7ce015c9161ac9fe7d7e036b91
3
  size 4993448880
model-00002-of-00002.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9eaf523e50340add3b139a6f6a53c51c571c74852871e6b4e93a553268ce55a3
3
  size 1180663192
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:89b5bb57ec7c4e57d7cb9a440a3c2c772ddaafa7122c5869415540e3c1516703
3
  size 1180663192