whisper-small-nomimo

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: 0.0000
• Wer: 1.9149

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 132
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.3211	1.3889	100	0.2361	12.7660
0.7584	2.7778	200	0.1198	12.5532
0.2285	4.1667	300	0.1448	158.5106
0.1734	5.5556	400	0.1582	27.0213
0.1351	6.9444	500	0.0599	11.2766
0.1252	8.3333	600	0.0704	52.9787
0.086	9.7222	700	0.1452	17.2340
0.0865	11.1111	800	0.0404	54.0426
0.0635	12.5	900	0.0221	49.7872
0.0672	13.8889	1000	0.0403	52.9787
0.034	15.2778	1100	0.0459	7.0213
0.0396	16.6667	1200	0.0205	3.8298
0.0262	18.0556	1300	0.0147	5.1064
0.0227	19.4444	1400	0.0101	48.2979
0.0108	20.8333	1500	0.0019	47.0213
0.0044	22.2222	1600	0.0013	1.7021
0.0021	23.6111	1700	0.0001	1.9149
0.0002	25.0	1800	0.0001	1.9149
0.0002	26.3889	1900	0.0000	1.9149
0.0002	27.7778	2000	0.0000	1.9149
0.0001	29.1667	2100	0.0000	1.9149

Framework versions

• Transformers 4.45.0.dev0
• Pytorch 2.4.0
• Datasets 2.21.0
• Tokenizers 0.19.1