mmarron14/whisper-medium-cv17-es-5-steps_proc2

Whisper Medium CV17 Es 5 steps with processing of text using method 2; with same configuration as 5-steps-sin_proc: no filtering 30sec, and with customised optimizer- María Marrón

This model is a fine-tuned version of openai/whisper-medium on the Common Voice 17.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3279
• Wer Ortho: 12.4854
• Wer: 7.0710

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: 2
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 32
• total_train_batch_size: 64
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• training_steps: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
No log	0.2	1	0.6498	12.5746	7.1028
No log	0.4	2	0.5543	12.3537	7.0297
No log	0.6	3	0.4379	12.3734	7.0234
No log	0.8	4	0.3593	12.4320	7.0380
No log	1.0	5	0.3279	12.4854	7.0710

Framework versions

• Transformers 4.53.2
• Pytorch 2.9.1+cu128
• Datasets 2.14.4
• Tokenizers 0.21.4