whisper-small-atc-co-spanish

This model is a fine-tuned version of openai/whisper-small on the atc-co-spanish dataset. It achieves the following results on the evaluation set:

• Loss: 1.4043
• Wer: 55.5556

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• training_steps: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.5633	14.2857	50	1.7417	65.0794
0.2582	28.5714	100	1.3265	60.3175
0.0009	42.8571	150	1.2653	52.3810
0.0003	57.1429	200	1.3243	53.9683
0.0002	71.4286	250	1.3494	53.9683
0.0002	85.7143	300	1.3700	53.9683
0.0001	100.0	350	1.3853	55.5556
0.0001	114.2857	400	1.3966	55.5556
0.0001	128.5714	450	1.4019	55.5556
0.0001	142.8571	500	1.4043	55.5556

Framework versions

• Transformers 4.45.1
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.0