workstation_whisper_base_finetune_teacher__babble_noise_mozilla_100_epochs_batch_4

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

• Loss: 1.3964
• Wer: 36.5051

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: 5e-05
• train_batch_size: 4
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 256
• total_train_batch_size: 1024
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine_with_restarts
• lr_scheduler_warmup_ratio: 0.2
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.0214	7.35	500	0.8448	36.1291
0.3301	14.7	1000	0.9065	35.5511
0.0745	22.06	1500	1.1071	36.1535
0.0089	29.41	2000	1.2245	36.1082
0.0026	36.76	2500	1.3039	36.3171
0.0015	44.12	3000	1.3551	36.4216
0.001	51.47	3500	1.3964	36.5051

Framework versions

• Transformers 4.24.0
• Pytorch 1.12.1
• Datasets 2.7.1
• Tokenizers 0.11.0