Modelpublic

kimas1269/whisper-meduim_zhtw

sourceHugging Faceapache-2.0updated 3y agoView on Hugging Face

0likes7downloads

Model Card

Whisper Medium Zh - Kimas

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

—Loss: 0.0635
—Wer: 100.0

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
—trainbatchsize: 8
—evalbatchsize: 8
—seed: 42
—gradientaccumulationsteps: 2
—totaltrainbatch_size: 16
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—lrschedulerwarmup_steps: 500
—training_steps: 10000
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1461	0.28	1000	0.1406	100.0
0.0803	0.57	2000	0.1181	100.0
0.0715	0.85	3000	0.1039	100.0
0.0255	1.14	4000	0.0925	100.0207
0.0199	1.42	5000	0.0810	100.0
0.027	1.7	6000	0.0767	100.0207
0.0328	1.99	7000	0.0706	100.0
0.0026	2.27	8000	0.0700	100.0
0.0082	2.56	9000	0.0646	100.0
0.0099	2.84	10000	0.0635	100.0

Framework versions

—Transformers 4.36.0.dev0
—Pytorch 1.12.1
—Datasets 2.14.6
—Tokenizers 0.14.1