Modelpublic

jethrowang/android_loss_CH_1.0-whisper-tiny

sourceHugging Faceapache-2.0updated 2y agoView on Hugging Face

0likes2downloads

Model Card

Whisper Tiny Taiwanese Simulated Android

This model is a fine-tuned version of openai/whisper-tiny on the TAT ASR Aligned dataset. It achieves the following results on the evaluation set:

—Loss: 0.7438
—Cer: 11.6466

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: 0.0001
—trainbatchsize: 64
—evalbatchsize: 32
—seed: 42
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—lrschedulerwarmup_steps: 1362
—training_steps: 13620
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.3611	0.9985	681	0.4700	20.9285
0.2547	1.9971	1362	0.4463	15.1381
0.1658	2.9956	2043	0.4418	13.8355
0.1045	3.9941	2724	0.4723	13.4539
0.0687	4.9927	3405	0.4987	13.4172
0.0456	5.9912	4086	0.5397	13.2578
0.0326	6.9897	4767	0.5761	12.9786
0.0219	7.9883	5448	0.6007	13.0098
0.0167	8.9868	6129	0.6061	12.7120
0.0122	9.9853	6810	0.6446	12.8573
0.0087	10.9839	7491	0.6544	12.7846
0.0053	11.9824	8172	0.6783	12.3071
0.0041	12.9809	8853	0.6960	12.3634
0.002	13.9795	9534	0.7046	12.2334
0.0012	14.9780	10215	0.7138	12.0635
0.0004	15.9765	10896	0.7239	12.0304
0.0002	16.9751	11577	0.7270	11.7646
0.0001	17.9736	12258	0.7367	11.6746
0.0001	18.9721	12939	0.7418	11.6619
0.0001	19.9707	13620	0.7438	11.6466

Framework versions

—Transformers 4.42.3
—Pytorch 2.3.0+cu121
—Datasets 2.20.0
—Tokenizers 0.19.1