Taeyeun72/whisper-small-denoising_2

whisper-kor3_de_2

This model is a fine-tuned version of openai/whisper-small on the whisper-kor3_de_2 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3685
• Wer: 23.2718
• Cer: 10.5941

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
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 200
• training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.3178	0.21	50	0.3333	23.0343	11.1119
0.2684	0.42	100	0.3293	22.9024	10.6660
0.279	0.64	150	0.3264	23.1662	10.8458
0.2813	0.85	200	0.3314	28.3905	14.9597
0.2363	1.06	250	0.3325	31.3720	16.7002
0.1909	1.27	300	0.3333	26.4908	13.0538
0.171	1.48	350	0.3384	24.4063	12.1044
0.1699	1.69	400	0.3330	22.6649	10.3711
0.1824	1.91	450	0.3336	23.3509	10.8242
0.0784	2.12	500	0.3425	22.7441	10.4574
0.0694	2.33	550	0.3492	23.2982	10.7379
0.0946	2.54	600	0.3442	24.2744	11.4212
0.0785	2.75	650	0.3486	22.6913	10.4646
0.0838	2.97	700	0.3466	22.7441	10.5941
0.0423	3.18	750	0.3600	24.2480	11.3780
0.0448	3.39	800	0.3615	23.0871	10.5509
0.0492	3.6	850	0.3640	23.3509	10.6228
0.04	3.81	900	0.3649	23.4565	10.5941
0.0385	4.03	950	0.3635	24.1689	11.2558
0.0276	4.24	1000	0.3685	23.2718	10.5941

Framework versions

• Transformers 4.33.2
• Pytorch 2.0.1+cu117
• Datasets 2.14.5
• Tokenizers 0.13.3