whisper-tiny-ml-2

This model is a fine-tuned version of openai/whisper-tiny on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2696
• Wer: 63.2536

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: 3.75e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• 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: 500
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.25	1.0316	500	0.5131	99.8569
0.29	2.0632	1000	0.1842	75.1277
0.1325	4.0264	1500	0.1808	73.6562
0.0741	5.058	2000	0.1953	71.7147
0.0424	7.0212	2500	0.2104	66.1353
0.0241	8.0528	3000	0.2327	66.1762
0.0122	10.016	3500	0.2463	66.4623
0.0054	11.0476	4000	0.2554	64.4186
0.002	13.0108	4500	0.2660	64.0507
0.0007	14.0424	5000	0.2696	63.2536

Framework versions

• Transformers 4.50.3
• Pytorch 2.6.0+cu124
• Datasets 3.5.0
• Tokenizers 0.21.1