whisper-small-canto

This model is a fine-tuned version of openai/whisper-small on the thisiskeithkwan/canto dataset. It achieves the following results on the evaluation set:

• Loss: 1.5061
• Cer: 0.4485

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: 0.0003
• train_batch_size: 2
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 16
• 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: 500
• training_steps: 5000

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
1.5909	0.76	500	1.6890	0.7769
1.2636	1.52	1000	1.4067	0.7641
0.7889	2.27	1500	1.3118	0.5474
0.6929	3.03	2000	1.2825	0.5516
0.4827	3.79	2500	1.2360	0.5446
0.236	4.55	3000	1.3457	0.5044
0.0982	5.31	3500	1.4736	0.4841
0.064	6.07	4000	1.5103	0.4809
0.035	6.82	4500	1.5110	0.4563
0.0103	7.58	5000	1.5061	0.4485

Framework versions

• Transformers 4.32.0.dev0
• Pytorch 2.0.1+cu118
• Datasets 2.14.3
• Tokenizers 0.13.3