<!-- 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. -->
Whisper Small Ko(FLUERS) - by p4b
This model is a fine-tuned version of openai/whisper-small on the FLUERS Korean dataset. It achieves the following results on the evaluation set:
- Loss: 0.4512
- Wer: 148.1005
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-07
- train_batch_size: 96
- eval_batch_size: 64
- seed: 42
- distributed_type: multi-GPU
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 500
- training_steps: 4000
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 0.6003 | 32.0 | 800 | 0.5913 | 167.2749 |
| 0.459 | 64.0 | 1600 | 0.4978 | 170.9841 |
| 0.4035 | 96.0 | 2400 | 0.4653 | 168.5911 |
| 0.3812 | 128.0 | 3200 | 0.4531 | 149.4765 |
| 0.3766 | 160.0 | 4000 | 0.4512 | 148.1005 |
Framework versions
- Transformers 4.26.0.dev0
- Pytorch 1.14.0.dev20221208+cu116
- Datasets 2.7.1.dev0
- Tokenizers 0.13.2
