whisper-small-ko-E10_Yspeed

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

• Loss: 0.2292
• Cer: 10.0153

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• num_epochs: 2
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.4177	0.13	100	0.2959	7.1898
0.2523	0.26	200	0.2551	7.4483
0.2141	0.39	300	0.2561	7.2133
0.1974	0.52	400	0.2412	7.0724
0.1863	0.64	500	0.2483	7.3484
0.1582	0.77	600	0.2423	6.7375
0.1597	0.9	700	0.2388	8.4998
0.0697	1.03	800	0.2335	6.6142
0.0589	1.16	900	0.2343	9.2634
0.0642	1.29	1000	0.2338	8.9932
0.0609	1.42	1100	0.2353	10.1974
0.0659	1.55	1200	0.2309	10.4382
0.0608	1.68	1300	0.2290	8.9168
0.0605	1.81	1400	0.2293	10.7789
0.0644	1.93	1500	0.2292	10.0153

Framework versions

• Transformers 4.37.0.dev0
• Pytorch 2.1.0+cu121
• Datasets 2.15.0
• Tokenizers 0.15.0