thiagobarbosa/whisper-base-common-voice-16-pt-v6

Whisper Base using Common Voice 16 (pt)

This model is a fine-tuned version of openai/whisper-base on the Mozilla Common Voices - 16.0 - Portuguese dataset. It achieves the following results on the evaluation set:

• Loss: 0.3552
• Wer: 25.4363
• Wer Normalized: 19.4668

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: 2e-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: 400
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Wer Normalized
0.6085	0.19	500	0.4465	32.1833	25.3383
0.4624	0.37	1000	0.4131	28.9867	22.8488
0.4375	0.56	1500	0.3936	27.8135	21.3817
0.4372	0.74	2000	0.3784	27.5695	21.7171
0.4704	0.93	2500	0.3630	26.1167	20.5133
0.2013	1.11	3000	0.3600	25.5462	19.7750
0.2261	1.3	3500	0.3570	25.5010	19.5181
0.2118	1.48	4000	0.3552	25.4363	19.4668

Framework versions

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