whisper-small-yue-mdcc-1

This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2556
• Cer: 13.0495

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: 64
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 128
• 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	Cer
0.5014	1.05	100	0.3315	27.2477
0.1895	2.11	200	0.2321	16.3903
0.1272	3.16	300	0.2210	15.7561
0.0759	4.21	400	0.2191	14.2006
0.0363	5.26	500	0.2249	16.1079
0.02	6.32	600	0.2320	13.4516
0.0112	7.37	700	0.2398	13.0711
0.0062	8.42	800	0.2497	13.0902
0.0048	9.47	900	0.2556	13.0495

Framework versions

• Transformers 4.38.0.dev0
• Pytorch 2.1.0+cu121
• Datasets 2.17.0
• Tokenizers 0.15.2