Modified version of Alphafold to divide CPU part (MSA and template searching) and GPU part. This can accelerate Alphafold when predicting multiple structures
I can run the CPU step using modest amounts of resource - 16 core, 12G per core. But the GPU step requires 400G of VMEM. I am using a Nvidia A100 with 80G of GPU memory - and it is being used according to 'nvidia-smi'. The protein I am folding is very short - 197 amino acids. Is this normal? I would have hoped the GPU step would require much less CPU RAM.
I can run the CPU step using modest amounts of resource - 16 core, 12G per core. But the GPU step requires 400G of VMEM. I am using a Nvidia A100 with 80G of GPU memory - and it is being used according to 'nvidia-smi'. The protein I am folding is very short - 197 amino acids. Is this normal? I would have hoped the GPU step would require much less CPU RAM.