whisper-a-nomi-17

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: 0.0335
• Wer: 9.5621

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: 17
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	1.0	88	0.3229	37.8910
0.9213	2.0	176	0.0773	21.7158
0.135	3.0	264	0.0585	26.5416
0.0547	4.0	352	0.0728	177.7480
0.0453	5.0	440	0.0574	18.0518
0.0237	6.0	528	0.0531	15.0134
0.0147	7.0	616	0.0271	6.6130
0.0106	8.0	704	0.0362	7.2386
0.0106	9.0	792	0.0491	8.9366
0.0063	10.0	880	0.0327	8.1323
0.0036	11.0	968	0.0352	7.3280
0.0009	12.0	1056	0.0695	11.4388
0.0023	13.0	1144	0.0338	10.2770
0.0001	14.0	1232	0.0374	10.2770
0.0	15.0	1320	0.0335	9.5621
0.0	16.0	1408	0.0335	9.5621
0.0	17.0	1496	0.0335	9.5621

Framework versions

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