igarciahuidobro/whisper-tiny-300-audios

Whisper Pre Tuned 300 Audios - Nacho v2.0

This model is a fine-tuned version of rasel35/whisper-base-es-medical-terms on the 300 audios 1.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4543
• Wer: 19.9630

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: 5e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• 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: 50
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.8385	1.0	18	1.1087	37.7079
0.6509	2.0	36	0.5255	25.1386
0.2616	3.0	54	0.4827	21.2569
0.1177	4.0	72	0.4747	21.0721
0.0719	5.0	90	0.4630	20.8872
0.0391	6.0	108	0.4802	21.8115
0.0313	7.0	126	0.4613	20.7024
0.023	8.0	144	0.4557	17.7449
0.0104	9.0	162	0.4513	20.1479
0.0042	9.4507	170	0.4543	19.9630

Framework versions

• Transformers 4.48.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.0
• Tokenizers 0.21.0