Simon Beck
Simon Beck
@mjustynaPhD i was actually wondering the same thing. I was trying to submit Kaggle's 3D RNA competiton with GraphaRNA, but turns out P100 or T2 GPU couldn't process Flash-Attention. So...
@doheon114 if u give me ur email, i can send u 80epoch weights. Since I trained GraphaRNA with Coarse-grained mode, which doesn't include C1' coordinates, I have to use models...
@mjustynaPhD @RJPenic I successfully infered with **RiNALMo NON-flash attn rotary embeding** and GraphaRNA's model_800epoch.h5 weight even though it(model_800epoch.h5) was pretrained with **RiNALMo's flash-attn rotary embedding.** I did some editing in...