Modelpublic

jethrowang/android_topline-whisper-tiny_tat

sourceHugging Faceapache-2.0updated 2y agoView on Hugging Face

0likes8downloads

Model Card

Whisper Tiny Taiwanese 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: 0.6536
—Cer: 10.3016

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
0.371	0.9985	681	0.4334	14.4492
0.2637	1.9971	1362	0.3950	13.0672
0.1725	2.9956	2043	0.3962	12.1858
0.1102	3.9941	2724	0.4102	11.8710
0.0715	4.9927	3405	0.4442	11.9113
0.0467	5.9912	4086	0.4830	12.2436
0.0322	6.9897	4767	0.5100	11.6466
0.0234	7.9883	5448	0.5315	11.5878
0.0182	8.9868	6129	0.5542	11.8786
0.012	9.9853	6810	0.5834	11.5762
0.0083	10.9839	7491	0.5833	11.4945
0.0061	11.9824	8172	0.6000	11.1774
0.0045	12.9809	8853	0.6136	11.0700
0.0027	13.9795	9534	0.6144	10.8808
0.0008	14.9780	10215	0.6320	10.6295
0.0006	15.9765	10896	0.6380	10.6150
0.0003	16.9751	11577	0.6385	10.4755
0.0003	17.9736	12258	0.6498	10.4047
0.0001	18.9721	12939	0.6537	10.3546
0.0001	19.9707	13620	0.6536	10.3016

Framework versions

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