Modelpublic

syp1229/whisper-small-ko-Ynormal

sourceHugging Faceapache-2.0updated 3y agoView on Hugging Face

0likes7downloads

Model Card

Whisper Small ko-Ynormal - syp1229

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

—Loss: 0.3226
—Cer: 7.2687

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: 2e-05
—trainbatchsize: 16
—evalbatchsize: 16
—seed: 42
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—lrschedulerwarmup_steps: 50
—num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.1988	0.3	100	0.3257	8.7640
0.1965	0.59	200	0.3302	8.6157
0.1374	0.89	300	0.3257	9.6481
0.0528	1.18	400	0.3255	8.6097
0.0515	1.48	500	0.3126	8.0104
0.0461	1.78	600	0.3063	7.9096
0.0204	2.07	700	0.3209	7.8206
0.0174	2.37	800	0.3216	7.6070
0.0225	2.66	900	0.3095	7.4586
0.0204	2.96	1000	0.3186	8.0342
0.0048	3.25	1100	0.3188	7.6485
0.0062	3.55	1200	0.3149	7.4586
0.0044	3.85	1300	0.3156	7.4527
0.0076	4.14	1400	0.3182	7.1975
0.004	4.44	1500	0.3211	7.1738
0.0027	4.73	1600	0.3226	7.2687

Framework versions

—Transformers 4.34.0.dev0
—Pytorch 2.0.1+cu118
—Datasets 2.14.4
—Tokenizers 0.13.3