Vulkan backend is great

[thewh1teagle]

Just want to say that Vulkan backend makes the project 10x better.
It works fast as with cuda. it works with Amd/Nvidia/Intel GPUs. Finally almost every computer on Linux/Windows get good performance!
Waiting for the next release!

Some tests I made:

System	Time Taken (1 Hour Audio)	Model
RTX 3060 on Linux	2 minutes	Medium
AMD Ryzen 5 4500U on Windows	25 minutes	Medium
Mac Mini with M1 and CoreML	10 minutes	Medium

[stopthinking102]

android supports vulkan . have u tested it on android

[soundwarrior]

I also am curious if anyone has tried Vulkan on Android. With the current CPU implementation in the example, I get about 700 mils on average to decode the average sentence which is pretty good, frankly. Also, I saw that there is a project called Insanely Fast Whisper, but, of course, it is only in Python currently. Thanks for any feedback on this.