yum-archive/TaSTT-Whisper

High-performance GPGPU inference of OpenAI's Whisper automatic speech recognition (ASR) model

git clone https://git.yummers.dev/yum-archive/TaSTT-Whisper

KonstantinUpdated performance summaryff1bb57

master
1.4 KiB17 linesraw
1Audio Clip	Model	GPU	Total, sec	Relative speed	Encode, sec	Decode, sec	RAM, MB	VRAM, MB
2jfk.wav	medium	GeForce 1080Ti	1.13909	9.6568	0.599964	0.451832	2.84049	2185.90208
3jfk.wav	medium	GeForce 1650	3.16174	3.4791	1.95373	0.987441	2.84049	2185.90208
4jfk.wav	medium	Ryzen 5 5600U	8.79853	1.2502	7.34021	1.30925	2.84049	2233.35424
5jfk.wav	medium	Ryzen 7 5700G	4.95485	2.2200	3.82894	1.03424	2.84049	2233.35424
6jfk.wav	large	GeForce 1080Ti	2.19628	5.0085	1.29615	0.7739170000000001	2.85543	4052.89984
7jfk.wav	large	GeForce 1650	8.33686	1.3194	3.98133	4.07331	2.85543	4052.89984
8jfk.wav	large	Ryzen 5 5600U	14.2729	0.7707	11.9404	2.12941	2.85543	4112.35328
9jfk.wav	large	Ryzen 7 5700G	9.46787	1.1618	7.35144	2.01739	2.85543	4112.35328
10columbia.wma	medium	GeForce 1080Ti	14.9475	13.2973	6.01034	8.78676	91.929	2247.34208
11columbia.wma	medium	GeForce 1650	48.7479	4.0773	19.2258	29.233	91.929	2247.34208
12columbia.wma	medium	Ryzen 5 5600U	81.256	2.4461	51.1656	29.6502	91.929	2295.52128
13columbia.wma	medium	Ryzen 7 5700G	62.1145	3.1999	37.8702	23.9262	91.929	2295.52128
14columbia.wma	large	GeForce 1080Ti	27.5329	7.2191	11.3967	15.9412	93.1329	4118.28224
15columbia.wma	large	GeForce 1650	109.423	1.8165	36.3141	72.8459	93.1329	4118.28224
16columbia.wma	large	Ryzen 5 5600U	140.747	1.4122	88.7441	51.4306	93.1329	4178.78016
17columbia.wma	large	Ryzen 7 5700G	110.474	1.7992	65.7998	44.3232	93.1329	4178.78016