thiagobarbosa/whisper-base-common-voice-16-pt-v8

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: 0.4574
• Wer: 26.1926
• Wer Normalized: 20.0029

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: 32
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 7000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Wer Normalized
0.4883	0.74	1000	0.3803	28.0317	21.8327
0.2659	1.48	2000	0.3677	26.3688	20.1666
0.1251	2.22	3000	0.3730	26.3752	20.4620
0.1071	2.96	4000	0.3867	25.5026	19.5470
0.0523	3.7	5000	0.4148	25.7094	19.6851
0.02	4.44	6000	0.4491	25.6803	19.5759
0.0134	5.18	7000	0.4574	26.1926	20.0029

Framework versions

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