whisper-a-nomi-15

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: 0.0280
• Wer: 196.1573

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
No log	1.0	88	0.2007	80.2502
0.9112	2.0	176	2.4319	100.0
2.8333	3.0	264	0.0985	34.4951
0.1561	4.0	352	0.0809	36.6399
0.0444	5.0	440	0.0607	30.4736
0.0244	6.0	528	0.0363	28.5076
0.0121	7.0	616	0.0430	24.8436
0.009	8.0	704	0.0398	31.0992
0.009	9.0	792	0.0360	198.9276
0.0084	10.0	880	0.0401	201.4298
0.0028	11.0	968	0.0278	196.2466
0.0	12.0	1056	0.0279	196.1573
0.0	13.0	1144	0.0280	196.1573
0.0	14.0	1232	0.0280	196.1573
0.0	15.0	1320	0.0280	196.1573

Framework versions

• Transformers 4.47.0.dev0
• Pytorch 2.4.0
• Datasets 3.1.0
• Tokenizers 0.20.3