thad0ctor
thad0ctor
> Gave a test with my 3 rtx 5090, slow as hell generation on 30k tokens prompt with command-r 2024 (3 tokens per second) > > With pytorch nightly +...
Any word on a fix, this messes up DDP with some workflows
As others mentioned you can get multi gpu working with accelerate. I posted how I got it working with 5090s here: https://github.com/thad0ctor/unsloth-5090-multiple
I don’t see why it wouldn’t, you may have to configure it differently
I just looked at GitHub discussions and asked chat gpt mostly, sorry
> I wonder what the relationship between https://github.com/anhvth/opensloth and this PR is, if any. Good question, I haven't tried it yet. Looking at the github it has more files patched...
I can’t say for sure because I haven’t tested a recent unsloth build however I was successful at getting unsloth to load and recognize multiple GPUs before my patch, it...
> Oh sorry I misunderstood your question. So training other models doesn't have similar issues with liger-kernel enabled and exact same configs? Only on Qwen3-VL(`Qwen3ForConditionalGeneration`)? > > Currently qwen3-vl liger...