whisper-large-v3-pt-1000h
This model is a fine-tuned version of openai/whisper-large-v3 on the fsicoli/cv17-fleurs-coraa-mls-ted-alcaim-cf-cdc-lapsbm-lapsmail-sydney-lingualibre-voxforge-tatoeba default dataset. It achieves the following results on the evaluation set:
- Loss: 0.5576
- Wer: 0.1113
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: 5e-06
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- num_devices: 2
- total_train_batch_size: 32
- total_eval_batch_size: 16
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 10000
- training_steps: 82000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.2717 | 0.39 | 10000 | 0.4143 | 0.1341 |
| 0.2646 | 0.79 | 20000 | 0.4141 | 0.1284 |
| 0.2244 | 1.18 | 30000 | 0.5361 | 0.1253 |
| 0.2056 | 1.57 | 40000 | 0.4714 | 0.1223 |
| 0.2034 | 1.97 | 50000 | 0.4937 | 0.1195 |
| 0.1717 | 2.36 | 60000 | 0.5127 | 0.1178 |
| 0.1692 | 2.75 | 70000 | 0.6040 | 0.1146 |
| 0.121 | 3.15 | 80000 | 0.5361 | 0.1130 |
Framework versions
- Transformers 4.39.0.dev0
- Pytorch 2.2.1
- Datasets 2.18.1.dev0
- Tokenizers 0.15.2