whisper-small-cs

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

• Loss: 0.0010
• Wer: 44.7674

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: 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: 10
• training_steps: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	10.0	10	2.9041	96.5116
No log	20.0	20	0.2452	109.3023
2.545	30.0	30	0.0064	53.4884
2.545	40.0	40	0.0015	44.7674
0.0053	50.0	50	0.0010	44.7674

Framework versions

• Transformers 4.49.0
• Pytorch 2.5.1+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0