whisper-finetuned-shortened

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

• Loss: 1.0079
• Wer Ortho: 45.1868
• Wer: 35.0595

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: 16
• 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: constant_with_warmup
• lr_scheduler_warmup_steps: 50
• training_steps: 60
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
No log	1.0	8	1.7547	56.6092	46.2561
No log	2.0	16	1.4652	52.9454	42.1973
No log	3.0	24	1.2884	49.7126	39.2582
1.5685	4.0	32	1.1201	47.7730	37.3688
1.5685	5.0	40	1.0467	46.6954	36.7390
1.5685	6.0	48	1.0030	46.8391	36.0392
0.69	7.0	56	0.9990	43.9655	34.9195
0.69	7.5	60	1.0079	45.1868	35.0595

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0