Sergey O
Sergey O
This fix hasn't been pushed for local installation yet. (Are you seeing same error for local install?) (reopening issue until we fix for local installations)
I believe lastest update should fix this. try again!
I think I finally fixed it! Please try gain. In short: esmfold uses openfold google colab updated to cuda12 only pytorch2 is available for cuda12 openfold was compiled with pytorch1,...
You might need to remove the {commit} part. As that I'm guessing doesn't exist in your branch
The issue should now be fixed, can you try again?
This appears to be a google colab issue... The new default is high-memory and A100 GPU (currently in high demand and unavailable). To fix this issue, you'd need to downgrade...
Try increase to high-memory but keep GPU at V100 or T4
Thanks for report! Looks like the issue is with rfdiffusion weights, the provided links by uw/ipd no longer work. eg: http://files.ipd.uw.edu/pub/RFdiffusion/6f5902ac237024bdd0c176cb93063dc4/Base_ckpt.pt I've contacted the authors to see if we can...
Can you provide more details? Do you have the zip output files from both runs? Can you share these with me? [email protected]
Hi Marian, I'm not seeing any big differences... There are gonna be slight differences. Even changing GPU will cause a slight difference... but the differences should be small.