longriverlong/whisper-tiny-chinese

Whisper Tiny Chinese - Tommy

This model is a fine-tuned version of openai/whisper-tiny on the google fleurs dataset. It achieves the following results on the evaluation set:

• Loss: 2.0604
• Cer: 686.3631

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: linear
• lr_scheduler_warmup_steps: 100
• training_steps: 3500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
2.7861	2.18	500	2.9273	541.9488
2.2443	4.37	1000	2.5364	440.6712
1.9019	6.55	1500	2.3037	705.8989
1.6275	8.73	2000	2.1769	746.9195
1.4541	10.92	2500	2.1022	756.6663
1.3193	13.1	3000	2.0731	699.0618
1.2989	15.28	3500	2.0604	686.3631

Framework versions

• Transformers 4.37.0.dev0
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0