vumichien/whisper-large-v2-jp

openai/whisper-large-v2

This model is a fine-tuned version of openai/whisper-large-v2 on the mozilla-foundation/common_voice_11_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2352
• Wer: 8.1166
• Cer: 5.0032

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.0897	0.1	1000	0.1884	11.0068	6.6992
0.0396	0.2	2000	0.1749	9.7399	5.9350
0.036	1.1	3000	0.1698	9.1419	5.6781
0.012	1.2	4000	0.1849	9.3041	5.7661
0.0151	2.09	5000	0.1879	9.1959	5.6761
0.0047	2.19	6000	0.2097	8.6706	5.4422
0.0046	3.09	7000	0.2040	8.8277	5.4717
0.0015	3.19	8000	0.2260	8.4949	5.3101
0.0013	4.09	9000	0.2339	8.3716	5.1471
0.0005	4.19	10000	0.2352	8.1166	5.0032

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.0+cu117
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2