whisper-medium-bigcgen-baseline-42

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

• Loss: 0.6970
• Wer: 0.5261

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: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• 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: 500
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.0745	0.6102	200	0.9366	0.6486
0.6532	1.2197	400	0.7690	0.5467
0.6347	1.8299	600	0.7060	0.5129
0.4066	2.4394	800	0.6970	0.5261
0.2542	3.0488	1000	0.7140	0.5034
0.252	3.6590	1200	0.7221	0.4833
0.137	4.2685	1400	0.7573	0.4878
0.114	4.8787	1600	0.7934	0.4812

Framework versions

• Transformers 4.53.0.dev0
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.0