noflm/whisper-small-ja-cv11

Whisper Small Japanese

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

• Loss: 0.4317
• Wer: 13.3262

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: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant_with_warmup
• lr_scheduler_warmup_steps: 500
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.042	3.03	2000	0.3056	12.9174
0.0085	7.01	4000	0.3752	13.1746
0.0047	10.04	6000	0.4103	13.5817
0.0042	14.01	8000	0.4202	13.5323
0.0051	17.05	10000	0.4317	13.3262

Framework versions

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