whisper-base-common-voice-16-pt

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

• Loss: 0.3608
• Wer: 25.7805
• Wer Normalized: 19.8889

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: 2.05e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 300
• training_steps: 3000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Wer Normalized
0.4466	0.37	500	0.4422	30.9405	25.0028
0.3661	0.74	1000	0.4012	29.2259	22.9756
0.1767	1.11	1500	0.3841	27.5178	21.3383
0.1896	1.48	2000	0.3698	26.7728	20.5807
0.1637	1.85	2500	0.3616	25.5398	20.0286
0.1033	2.22	3000	0.3608	25.7805	19.8889

Framework versions

• Transformers 4.36.2
• Pytorch 2.1.1
• Datasets 2.16.1
• Tokenizers 0.15.0