./whisper-large-cit-do0.25-wd0-lr1e-06
This model is a fine-tuned version of openai/whisper-large-v3 on the SF 200 dataset. It achieves the following results on the evaluation set:
- Loss: 0.6753
- Wer: 34.3249
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: 1e-06
- train_batch_size: 4
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- gradient_accumulation_steps: 4
- total_train_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- training_steps: 200
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
1.1266 | 0.8889 | 10 | 1.1143 | 48.9703 |
1.0858 | 1.7778 | 20 | 1.0078 | 39.5881 |
0.9333 | 2.6667 | 30 | 0.8696 | 39.1304 |
0.7546 | 3.5556 | 40 | 0.7925 | 34.0961 |
0.7029 | 4.4444 | 50 | 0.7202 | 34.7826 |
0.601 | 5.3333 | 60 | 0.6558 | 32.9519 |
0.5097 | 6.2222 | 70 | 0.6177 | 31.5789 |
0.415 | 7.1111 | 80 | 0.5913 | 33.4096 |
0.3794 | 8.0 | 90 | 0.5728 | 32.9519 |
0.3026 | 8.8889 | 100 | 0.5693 | 33.4096 |
0.2687 | 9.7778 | 110 | 0.5732 | 33.1808 |
0.2175 | 10.6667 | 120 | 0.5825 | 31.8078 |
0.1864 | 11.5556 | 130 | 0.5942 | 33.1808 |
0.155 | 12.4444 | 140 | 0.6060 | 31.1213 |
0.1266 | 13.3333 | 150 | 0.6182 | 33.1808 |
0.1212 | 14.2222 | 160 | 0.6328 | 33.8673 |
0.1043 | 15.1111 | 170 | 0.6499 | 33.8673 |
0.0894 | 16.0 | 180 | 0.6616 | 32.7231 |
0.084 | 16.8889 | 190 | 0.6724 | 33.8673 |
0.0787 | 17.7778 | 200 | 0.6753 | 34.3249 |
Framework versions
- Transformers 4.41.1
- Pytorch 1.13.1+cu117
- Datasets 2.19.1
- Tokenizers 0.19.1
- Downloads last month
- 1