whisper-medium-bemgen-baseline-62

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

• Loss: 0.5268
• Wer: 0.4506

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: 62
• 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
0.9781	0.5125	200	0.7509	0.6631
0.5787	1.0231	400	0.5970	0.5311
0.4938	1.5356	600	0.5545	0.4928
0.243	2.0461	800	0.5268	0.4506
0.242	2.5586	1000	0.5386	0.4832
0.0958	3.0692	1200	0.5889	0.4549
0.1067	3.5817	1400	0.5833	0.4583
0.0525	4.0922	1600	0.5935	0.4436

Framework versions

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