You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I am running a molecular dynamics (MD) simulation using mace_off for a PET and PETase docking result. After starting the simulation, the temperature becomes NaN, and the energy does not change over the course of the MD steps.
To Reproduce
Steps to reproduce the behavior:
Use a PDB file for PET-PETase docking results (e.g., 2pet_petase_wt_dockpose.pdb).
Set up the mace_off calculator with the small model (mace_off(model="small", device="cuda")).
Initialize MD simulation using the Andersen thermostat and a 1.0 fs timestep.
Run 10 steps of MD simulation and observe the output.
Expected behavior
I expected the temperature to remain stable and the energy to vary as the MD simulation progressed. Instead, the temperature quickly became NaN and the energy did not change after the first step.
Code
The code is almost identical to the 'Mace Tutorial - Molecular Dynamics with MACE,' which can be found here.
You might want to geometry-optimise the structure before you start the MD. Also make sure your initial structure is "clean", no atoms on top of each other, (e.g. hydrogens) etc.
You are using a quite large system. I would suggest you to try to run in OpenMM instead of lammps. You can see the documentation here: https://mace-docs.readthedocs.io/en/latest/guide/openmm.html. Try to see if you can reproduce your errors there. It might be a memory handling probem.
Describe the bug
I am running a molecular dynamics (MD) simulation using mace_off for a PET and PETase docking result. After starting the simulation, the temperature becomes NaN, and the energy does not change over the course of the MD steps.
To Reproduce
Steps to reproduce the behavior:
Use a PDB file for PET-PETase docking results (e.g.,
2pet_petase_wt_dockpose.pdb
).Set up the mace_off calculator with the small model (
mace_off(model="small", device="cuda")
).Initialize MD simulation using the Andersen thermostat and a 1.0 fs timestep.
Run 10 steps of MD simulation and observe the output.
Expected behavior
I expected the temperature to remain stable and the energy to vary as the MD simulation progressed. Instead, the temperature quickly became NaN and the energy did not change after the first step.
Code
The code is almost identical to the 'Mace Tutorial - Molecular Dynamics with MACE,' which can be found here.
Output
MD finished in 0.74 minutes!
Desktop (please complete the following information):
OS: Ubuntu 24.04 LTS
MACE Version: 0.3.6
ACE Version: 3.23.0
Torch Version: 2.4.0+cu124
GPU Info: V100
CUDA Version: 12.5
Support Info
2pet_petase_wt_dockpose.pdb.zip
The text was updated successfully, but these errors were encountered: