whisper-small-nose

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: 0.0078
• Wer: 0.6383

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
1.348	2.1277	100	0.3435	66.1702
0.9069	4.2553	200	0.6359	127.2340
0.5917	6.3830	300	0.5703	93.1915
0.5389	8.5106	400	0.5150	98.9362
0.5074	10.6383	500	0.4959	101.4894
0.4695	12.7660	600	0.4411	100.2128
0.4338	14.8936	700	0.3799	83.1915
0.3685	17.0213	800	0.3144	76.3830
0.3018	19.1489	900	0.2604	65.1064
0.238	21.2766	1000	0.1752	46.1702
0.1896	23.4043	1100	0.1110	25.7447
0.1132	25.5319	1200	0.0687	17.0213
0.0649	27.6596	1300	0.0401	7.8723
0.0217	29.7872	1400	0.0078	0.6383

Framework versions

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