Mezosky/whisper-base-clinical-assistance

Whisper Chilean Spanish Small

This model is a fine-tuned version of openai/whisper-base on the Mezosky/es_clinical_assistance dataset. It achieves the following results on the evaluation set:

• Loss: 4.4659
• Wer: 204.9755

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: 1e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.6451	6.25	100	4.5135	105.7912
3.2485	12.5	200	3.3821	126.5905
2.3839	18.75	300	2.9779	215.0897
1.6538	25.0	400	3.0304	212.1533
0.887	31.25	500	3.4092	221.3703
0.3317	37.5	600	3.7754	191.3540
0.1065	43.75	700	4.0480	235.1550
0.0374	50.0	800	4.2473	185.4812
0.0173	56.25	900	4.4145	187.5204
0.014	62.5	1000	4.4659	204.9755

Framework versions

• Transformers 4.39.3
• Pytorch 2.2.2+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2