ROCm VS Deepspeed-Windows

ROCm is not spelled out anywhere in their documentation and the best answers in search come from Github and not AMD official documents

"Radeon Open Compute Platform"

https://github.com/ROCm/ROCm/issues/1628

And they wonder why they are losing. Branding absolutely matters.

https://github.com/ROCm/ROCm/issues/1353

Bought in 2020. Stopped working in 2020. Not the latest, but in-production, advertised ROCm-capable, and what I could find during the Great GPU Shortage of 2020.

Maybe so. But it isn't confidence inspiring when I go to see which cards are supported and I see this issue:

https://github.com/ROCm/ROCm/issues/1714

With Nvidia cards, I know that if I buy any Nvidia card made in the last 10 years, CUDA code will run on it. Period. (Yes, different language levels require newer hardware, but Nvidia docs are quite clear about which CUDA versions require which silicon.)

The will-they-won't-they and the rapidly dropped support is hurting the otherwise excellent ROCm and HIP projects. There is a huge API surface to implement and it looks like they're making rapid gains.

Like, there is: https://github.com/ROCm/ROCm.github.io/blob/master/hardware.md but I'm pretty sure that's very very outdated, maybe from 4.x?

Fair enough I'll give you that. Although it is listed as officially supported here, other documentation says it works but is not officially supported.

Do you have a source? The 580 and several older cards are listed as officially supported here, and even some 2xx/3xx cards are listed as unofficially supported.

Some of the GPUs listed in your link are for consumers. For a more extensive list, see https://github.com/ROCm/ROCm.github.io/blob/master/hardware.md

Very conveniently, your linked page (the therein linked pages) do not talk about which GPUs actually do support ROCm. This is probably because AMDs newest cards do not support ROCm in any way, and would guess they don't want the sales pact this lack of feature could cause. Please do evaluate yourself, here: https://github.com/ROCm/ROCm.github.io/blob/master/hardware.md

I just went through this this weekend - If you're running in Windows and want to use deepspeed, you have to still use Cuda 12.1 because deepspeed 13.1 is the latest that works with 12.1. There's no deepspeed for windows that works with 12.3.

I tried to get it working this weekend but it was a huge PITA so I switched to putting everything into WSL2 then in arch on there pytorch etc in containers so I could flip versions easily.

I'm still working on that part, halfway into it my WSL2 completely broke and I had to reinstall windows. The p9 networking stopped working.

https://github.com/S95Sedan/Deepspeed-Windows