kliuae
kliuae
> I was going to try this out soon, is this in a good spot or is it still being worked? It's in a good state for testing, though occasionally...
@WoosukKwon Merge conflicts are resolved
We have a working version of mamba on ROCm. We've been able to run generation on AMD's MI210, and the unit tests are passing for the port. It uses pytorch's...
Hi @ppanchad-amd, the recent ROCm 6.4.2 release fixed this issue - confirmed working in docker containers. Thanks for your help.
It seems that allocating virtual memory with hipMemCreate causes it to throw an OOM error in sleep mode when trying to allocate a large contiguous memory space in one go,...