whisper-small-commonvoice-en

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: 0.3573
• Wer: 14.7768

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: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2697	0.4545	50	0.2484	13.1027
0.3113	0.9091	100	0.2571	13.3482
0.0671	1.3636	150	0.2864	13.7946
0.0833	1.8182	200	0.2984	14.3973
0.0297	2.2727	250	0.3126	14.2411
0.0253	2.7273	300	0.3181	14.4866
0.016	3.1818	350	0.3273	14.8661
0.0138	3.6364	400	0.3177	13.9955
0.012	4.0909	450	0.3335	14.4866
0.0054	4.5455	500	0.3401	14.6429
0.004	5.0	550	0.3459	14.5982
0.0011	5.4545	600	0.3535	15.1339
0.0026	5.9091	650	0.3422	15.1116
0.001	6.3636	700	0.3452	14.5313
0.0007	6.8182	750	0.3510	14.6875
0.0005	7.2727	800	0.3540	14.6429
0.0005	7.7273	850	0.3554	14.7545
0.0005	8.1818	900	0.3564	14.7321
0.0005	8.6364	950	0.3570	14.7545
0.0004	9.0909	1000	0.3573	14.7768

Framework versions

• Transformers 4.42.3
• Pytorch 2.1.2
• Datasets 2.20.0
• Tokenizers 0.19.1