Scrya/whisper-large-v2-cantonese

Whisper Large V2 - Cantonese - Augmented

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: 0.1828
• Cer: 6.2133

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

Training:

• mozilla-foundation/common_voice_11_0 (train+validation)

Evaluation:

• mozilla-foundation/common_voice_11_0 (test)

Training procedure

Datasets were augmented on-the-fly using audiomentations via PitchShift and TimeStretch transformations at p=0.3.

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 8
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• training_steps: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.1126	1.21	200	0.1666	7.3103
0.0467	2.42	400	0.1610	6.9419
0.0217	3.63	600	0.1621	6.3874
0.008	4.85	800	0.1699	6.3064
0.0023	6.06	1000	0.1828	6.2133

Framework versions

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