jempf/whisper-small-medical-de

Whisper Small Medical Swiss-German

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

• Loss: 0.3155
• Wer: 16.8239

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• 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: 20
• num_epochs: 4
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0.4619	50	0.3501	20.2859
0.3346	0.9238	100	0.3196	16.6067
0.3346	1.3788	150	0.3138	16.7493
0.1855	1.8406	200	0.3054	17.4463
0.1855	2.2956	250	0.3073	15.3230
0.1107	2.7575	300	0.3073	18.4576
0.1107	3.2125	350	0.3130	17.4754
0.0687	3.6744	400	0.3155	16.8239

Framework versions

• Transformers 4.49.0
• Pytorch 2.5.1+cu124
• Datasets 3.3.1
• Tokenizers 0.21.0