Saeed
Saeed
I was searching for a similar technique and found out this article but I don't have full access since it is paid. It reduces llama3.1 8B fp16 from 16GB to...
Hello team, I wanted to also check if there is specific timeline for AMD ROCm implementation? I have 2xAMD MI60 (GCN5 architecture). I would be great to see a general...
oh man, I feel your pain. I have two AMD MI60s. Even though I successfully built FA2 from https://github.com/ROCm/flash-attention.git by changing file setup.py line 126 by adding gfx906 (allowed_archs =...
Thanks, @nktice , Unfortunately, AMD MI60 is not CDNA or RDNA. It is GCN architecture. Yeah, I don't think sales team cares about older AMD cards (even though these cards...
Thank you for your great advice, @nktice. Since you mentioned Eric, let me ask him a question related to flash attention here.
Hi @ehartford, I see you mentioned that you had a pile of MI50 cards. Were you able to progress in implementing/compiling FA2 package for them? I have two MI60 cards....
oh I see. I think MI100 cards do not suffer from FA2 since they have support for FA2. But have you had MI50/60/Vega 20 cards before? Did you have any...
Hello team, Any updates on AMD GPU support?
Hello @lamikr , Thank you for your amazing work! I am really glad I found this repo. I have two AMD MI60 cards (gfx906). I will also compile this repo...
Quick update. I did a fresh installation of Ubuntu 24.04.1 today which takes around 6.5GB SSD storage. It installs Nvidia GPU drivers by default. I assumed this repo would install...