whisper-medium-bem

This model is a fine-tuned version of openai/whisper-medium on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3519
• Wer: 33.5877

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: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• 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
0.6509	0.34	500	0.4872	50.9031
0.5212	0.67	1000	0.3972	40.5156
0.3957	1.01	1500	0.3451	36.4793
0.2956	1.34	2000	0.3421	37.3866
0.2987	1.68	2500	0.3206	34.5374
0.1665	2.02	3000	0.3290	34.1135
0.1557	2.35	3500	0.3334	35.0462
0.1345	2.69	4000	0.3374	33.8506
0.0617	3.02	4500	0.3445	33.6216
0.0661	3.36	5000	0.3519	33.5877

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.12.1+cu113
• Datasets 2.7.1
• Tokenizers 0.13.2