Bhaveen/500medimix

Whisper Small En Medimix

This model is a fine-tuned version of openai/whisper-small on the 500 pes_colab 10.6 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4537
• Wer: 10.6667

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-06
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• 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
• lr_scheduler_warmup_steps: 200
• num_epochs: 8
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0384	1.0	25	0.4705	12.2222
0.0375	2.0	50	0.4689	12.0
0.0332	3.0	75	0.4666	11.0370
0.0293	4.0	100	0.4636	11.1111
0.0242	5.0	125	0.4612	11.0370
0.0197	6.0	150	0.4586	10.9630
0.0152	7.0	175	0.4555	10.7407
0.0116	8.0	200	0.4537	10.6667

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0