search

baker-laboratory / RoseTTAFold-All-Atom

Other
572 stars 97 forks source link

The protein-small molecule complex is successfully generate yet no GPU is engaged along the process #108

Open Sinsilcobio opened 1 month ago

Sinsilcobio commented 1 month ago

I successfully installed the package and executed it for the protein-ligand complex using the following command on a system equipped with an NVIDIA GeForce RTX 4070 Ti:

python -m rf2aa.run_inference --config-name protein_sm

The model generated successfully, utilizing only the CPU and GPU memory, but without engaging the CUDA cores.

Could you please confirm if the system has recognized the GPU? If so, what could be the reason for the non-utilization of the CUDA cores?

Thank you!

aw10279 commented 1 month ago

I got the same problem. I imported torch, executed "torch.cuda.is_available()" and got TRUE. But the model still run not recognizing GPU. Have you solved this issue?

Sinsilcobio commented 1 month ago

I am unsure whether the Cuda compute core should be engaged or not. However, I can successfully generate the model without it. Is it the same story on your side or no model is generated in your case?

merttopcu-rhea commented 1 month ago

When I am running protein-small molecule complex, I can see that my P0 GPU (NVIDIA L$ at Google Cloud) and python process is in the GPU memory however the GPU utilization shows as 0%... I assume it is the same problem. Any suggestions?