whisper-medium-bigcgen-balanced-52

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: 0.7864
• Wer: 0.5609

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: 52
• 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
1.0799	0.6079	200	1.0164	0.6944
0.6236	1.2158	400	0.8548	0.5904
0.6619	1.8237	600	0.7864	0.5609
0.4126	2.4316	800	0.7878	0.5579
0.2899	3.0395	1000	0.8031	0.5295
0.2227	3.6474	1200	0.8267	0.5078
0.1103	4.2553	1400	0.8692	0.5611

Framework versions

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