arun100/whisper-medium-vi-2

Whisper Medium Vietnamese

This model is a fine-tuned version of openai/whisper-medium on the mozilla-foundation/common_voice_11_0 vi dataset. It achieves the following results on the evaluation set:

• Loss: 0.5686
• Wer: 18.8638

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: 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: 1400
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0063	12.01	200	0.5238	19.2915
0.0046	24.01	400	0.5686	18.8638
0.0067	37.01	600	0.5924	20.6076
0.0004	49.01	800	0.6239	19.8070
0.0005	62.01	1000	0.6354	19.7631
0.0001	74.01	1200	0.6447	19.5547
0.0001	87.01	1400	0.6473	19.5547

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.1+cu117
• Datasets 2.8.1.dev0
• Tokenizers 0.13.2