| --- |
| license: apache-2.0 |
| tags: |
| - generated_from_trainer |
| metrics: |
| - sacrebleu |
| - wer |
| model-index: |
| - name: la-whisper-small-covost2 |
| results: [] |
| --- |
| |
| <!-- This model card has been generated automatically according to the information the Trainer had access to. You |
| should probably proofread and complete it, then remove this comment. --> |
|
|
| # la-whisper-small-covost2 |
|
|
| This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the None dataset. |
| It achieves the following results on the evaluation set: |
| - Loss: 1.5845 |
| - Sacrebleu: 2090.6716 |
| - Wer: 73.0006 |
|
|
| ## Model description |
|
|
| More information needed |
|
|
| ## Intended uses & limitations |
|
|
| More information needed |
|
|
| ## Training and evaluation data |
|
|
| More information needed |
|
|
| ## Training procedure |
|
|
| ### Training hyperparameters |
|
|
| The following hyperparameters were used during training: |
| - learning_rate: 3e-05 |
| - train_batch_size: 2 |
| - eval_batch_size: 2 |
| - seed: 42 |
| - gradient_accumulation_steps: 16 |
| - total_train_batch_size: 32 |
| - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
| - lr_scheduler_type: linear |
| - lr_scheduler_warmup_steps: 100 |
| - training_steps: 2000 |
| - mixed_precision_training: Native AMP |
|
|
| ### Training results |
|
|
| | Training Loss | Epoch | Step | Validation Loss | Sacrebleu | Wer | |
| |:-------------:|:-----:|:----:|:---------------:|:---------:|:--------:| |
| | 2.6943 | 0.11 | 50 | 2.1667 | 118.2640 | 686.6897 | |
| | 1.5505 | 0.23 | 100 | 1.6016 | 259.9307 | 165.6116 | |
| | 1.4093 | 0.34 | 150 | 1.5858 | 496.7335 | 197.0106 | |
| | 1.3209 | 0.45 | 200 | 1.5648 | 724.2491 | 121.8795 | |
| | 1.2941 | 0.56 | 250 | 1.5596 | 820.1241 | 161.7159 | |
| | 1.2078 | 0.68 | 300 | 1.5074 | 1022.0043 | 140.3875 | |
| | 1.1532 | 0.79 | 350 | 1.4972 | 174.8350 | 610.3716 | |
| | 1.0167 | 0.9 | 400 | 1.4551 | 1904.0921 | 82.7635 | |
| | 0.8842 | 1.01 | 450 | 1.4296 | 1883.6113 | 81.3906 | |
| | 0.5619 | 1.13 | 500 | 1.4333 | 1817.9440 | 84.9312 | |
| | 0.5523 | 1.24 | 550 | 1.4237 | 1517.1744 | 104.0918 | |
| | 0.4881 | 1.35 | 600 | 1.4413 | 1650.1807 | 97.2067 | |
| | 0.471 | 1.46 | 650 | 1.3961 | 1885.0014 | 82.2664 | |
| | 0.4412 | 1.58 | 700 | 1.3986 | 2145.9786 | 72.0469 | |
| | 0.4625 | 1.69 | 750 | 1.3885 | 1837.7812 | 87.4472 | |
| | 0.4195 | 1.8 | 800 | 1.4095 | 1909.2655 | 78.6920 | |
| | 0.4532 | 1.91 | 850 | 1.3891 | 1925.2238 | 82.0162 | |
| | 0.3201 | 2.03 | 900 | 1.4415 | 1919.2020 | 80.4437 | |
| | 0.1955 | 2.14 | 950 | 1.4410 | 1540.5046 | 101.0145 | |
| | 0.2111 | 2.25 | 1000 | 1.4345 | 1735.9648 | 90.9269 | |
| | 0.1981 | 2.36 | 1050 | 1.4597 | 1730.3250 | 91.5356 | |
| | 0.2052 | 2.48 | 1100 | 1.4439 | 2143.3630 | 72.4933 | |
| | 0.1886 | 2.59 | 1150 | 1.4702 | 1965.5005 | 77.7519 | |
| | 0.1918 | 2.7 | 1200 | 1.4518 | 2057.4517 | 75.4929 | |
| | 0.1755 | 2.81 | 1250 | 1.4788 | 1954.2237 | 78.2997 | |
| | 0.1769 | 2.93 | 1300 | 1.4588 | 1774.1464 | 91.9279 | |
| | 0.1104 | 3.04 | 1350 | 1.5281 | 1838.1999 | 84.7317 | |
| | 0.0718 | 3.15 | 1400 | 1.5133 | 2058.0955 | 76.0306 | |
| | 0.0855 | 3.26 | 1450 | 1.5271 | 1720.1072 | 89.1346 | |
| | 0.0717 | 3.38 | 1500 | 1.5289 | 2007.5163 | 75.9291 | |
| | 0.0707 | 3.49 | 1550 | 1.5366 | 2149.6478 | 71.9523 | |
| | 0.0704 | 3.6 | 1600 | 1.5355 | 2179.5147 | 69.8759 | |
| | 0.0676 | 3.71 | 1650 | 1.5393 | 2086.2197 | 73.2474 | |
| | 0.0748 | 3.83 | 1700 | 1.5398 | 1879.1610 | 80.7277 | |
| | 0.0695 | 3.94 | 1750 | 1.5351 | 2001.8476 | 78.8306 | |
| | 0.033 | 4.05 | 1800 | 1.5807 | 1892.0435 | 82.2630 | |
| | 0.0317 | 4.16 | 1850 | 1.5843 | 1967.1172 | 78.7765 | |
| | 0.0302 | 4.28 | 1900 | 1.5848 | 1969.6753 | 79.1248 | |
| | 0.0337 | 4.39 | 1950 | 1.5808 | 2062.9546 | 74.1537 | |
| | 0.0306 | 4.5 | 2000 | 1.5845 | 2090.6716 | 73.0006 | |
|
|
|
|
| ### Framework versions |
|
|
| - Transformers 4.28.0.dev0 |
| - Pytorch 2.0.0 |
| - Datasets 2.10.1 |
| - Tokenizers 0.13.2 |
|
|
