run_pyrosetta_ver.sh works but run_e2e_ver.sh fails with RuntimeError: CUDA out of memory.

[Joseph-Vineland]

run_pyrosetta_ver.sh works just fine and runs to completion but run_e2e_ver.sh fails with RuntimeError: CUDA out of memory during step 4. end-to-end prediction.

It is surprising to me that the full atomic model can work but the backbone-only model fails.

It appears many other users are experiencing the same problem. I hope the developers can offer a solution beyond just "get a better GPU" or "don't use the GPU".