whisper-a-nomimose

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: 0.0723
• Wer: 15.2655

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: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.3968	0.9217	100	0.7677	512.9794
0.3388	1.8387	200	0.3331	93.3628
0.2711	2.7558	300	0.2512	87.0944
0.2383	3.6728	400	0.2198	87.9056
0.2096	4.5899	500	0.1971	80.3835
0.2131	5.5069	600	0.1680	75.5900
0.1498	6.4240	700	0.1433	56.1209
0.1152	7.3410	800	0.1094	41.0767
0.0833	8.2581	900	0.1193	65.9292
0.0653	9.1751	1000	0.0728	25.1475
0.0444	10.0922	1100	0.0781	24.4100
0.0383	11.0092	1200	0.0537	17.6991
0.0269	11.9309	1300	0.0658	18.0678
0.0182	12.8479	1400	0.0641	19.3215
0.0128	13.7650	1500	0.0679	15.8555
0.0068	14.6820	1600	0.0723	15.2655

Framework versions

• Transformers 4.47.0.dev0
• Pytorch 2.4.0
• Datasets 3.0.1
• Tokenizers 0.20.0