whishper

This model is a fine-tuned version of openai/whisper-small on the common_voice_11_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.5474
• Wer: 72.2488
• Cer: 29.9605

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: 4
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 2
• num_epochs: 0.5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.2442	0.0333	5	0.8071	140.3509	157.0811
0.2386	0.0667	10	0.7964	146.2520	136.7877
0.3848	0.1	15	0.7687	146.8900	111.5479
0.3015	0.1333	20	0.7213	157.0973	126.8761
0.2178	0.1667	25	0.6916	159.1707	144.8561
0.2314	0.2	30	0.6551	149.6013	125.3526
0.2112	0.2333	35	0.6239	99.3620	64.2844
0.1571	0.2667	40	0.5794	76.5550	35.1514
0.1934	0.3	45	0.5547	73.0463	33.7596
0.3231	0.3333	50	0.5474	72.2488	29.9605
0.1035	0.3667	55	0.5434	72.5678	32.3491
0.1991	0.4	60	0.5454	74.0032	31.4275
0.196	0.4333	65	0.5495	73.5247	36.0166
0.4541	0.4667	70	0.5448	73.3652	38.5556
0.2166	0.5	75	0.5418	73.3652	39.3455

Framework versions

• Transformers 4.50.0.dev0
• Pytorch 2.5.1+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0