whisper-small

This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1540
• Wer: 13.8083

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-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.3889	0.0	1	3.1044	49.3314
0.174	0.29	1000	0.2346	20.9796
0.1521	0.58	2000	0.1945	17.9616
0.1301	0.88	3000	0.1747	16.2713
0.0951	1.17	4000	0.1684	15.3962
0.0955	1.46	5000	0.1606	14.7689
0.096	1.75	6000	0.1561	14.3492
0.0668	2.04	7000	0.1554	14.0853
0.062	2.34	8000	0.1555	14.0599
0.0664	2.63	9000	0.1548	13.9191
0.0678	2.92	10000	0.1540	13.8083

Framework versions

• Transformers 4.27.4
• Pytorch 2.0.0
• Datasets 2.11.0
• Tokenizers 0.13.3