jed351/whisper_small_cantonese_cm_voice

whisper-small-zh-hk

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

• Loss: 0.3003
• Wer: 0.5615

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: 16
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• total_train_batch_size: 32
• total_eval_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1556	2.28	1000	0.2708	0.6069
0.038	4.57	2000	0.2674	0.5701
0.0059	6.85	3000	0.2843	0.5635
0.0017	9.13	4000	0.2952	0.5622
0.0013	11.42	5000	0.3003	0.5615

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.1
• Datasets 2.8.0
• Tokenizers 0.13.2