whisper-medium-ro_private_dataset

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

• Loss: 3.9763
• Wer Ortho: 104.4025
• Wer: 102.3457

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: 32
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Use 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: 50
• training_steps: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.0288	20.0	100	3.3078	118.7421	116.9136
0.0015	40.0	200	3.6434	98.9937	97.6543
0.0017	60.0	300	3.6502	100.2516	99.1358
0.0002	80.0	400	3.8591	105.6604	103.5802
0.0001	100.0	500	3.9031	113.8365	111.9753
0.0001	120.0	600	3.9312	114.0881	112.0988
0.0001	140.0	700	3.9508	101.6352	99.7531
0.0001	160.0	800	3.9650	103.5220	101.7284
0.0001	180.0	900	3.9731	104.0252	101.9753
0.0001	200.0	1000	3.9763	104.4025	102.3457

Framework versions

• Transformers 4.46.3
• Pytorch 2.5.1+cu118
• Datasets 3.1.0
• Tokenizers 0.20.3