whisper-a-nomimo-16

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: 0.0862
• Wer: 25.1543

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: 0.0004
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use 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: 132
• num_epochs: 16
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.9727	1.0	104	0.1997	47.0679
0.231	2.0	208	0.0566	178.1636
0.2066	3.0	312	0.2833	91.6667
0.2809	4.0	416	0.2589	91.9753
0.2872	5.0	520	0.2672	88.8889
0.2384	6.0	624	0.2239	110.1080
0.202	7.0	728	0.1959	79.7840
0.1828	8.0	832	0.1883	78.3951
0.1775	9.0	936	0.1908	79.1667
0.1496	10.0	1040	0.2103	87.8858
0.1162	11.0	1144	0.1416	54.3981
0.0674	12.0	1248	0.0975	61.5741
0.0449	13.0	1352	0.0775	36.4969
0.026	14.0	1456	0.0706	23.6883
0.0197	15.0	1560	0.0873	26.6204
0.0119	15.8502	1648	0.0862	25.1543

Framework versions

• Transformers 4.48.0.dev0
• Pytorch 2.4.0
• Datasets 3.1.0
• Tokenizers 0.21.0