Issue on integrating lammps with MACE potential
aishwaryo opened this issue · 10 comments
Dear developers, I was attempting to install the CPU implementation of lammps+MACE from https://mace-docs.readthedocs.io/en/latest/guide/lammps.html. This clones and proceeds with the "mace" branch of the repo.
On the step with cmake I am getting the following error :
CMake Error at Modules/Packages/ML-MACE.cmake:3 (find_package):
By not providing "FindTorch.cmake" in CMAKE_MODULE_PATH this project has
asked CMake to find a package configuration file provided by "Torch", but
CMake did not find one.Could not find a package configuration file provided by "Torch" with any of
the following names:TorchConfig.cmake
torch-config.cmakeAdd the installation prefix of "Torch" to CMAKE_PREFIX_PATH or set
"Torch_DIR" to a directory containing one of the above files. If "Torch"
provides a separate development package or SDK, be sure it has been
installed.
Call Stack (most recent call first):
CMakeLists.txt:526 (include)-- Configuring incomplete, errors occurred! '
Could you please assist in resolving this?
Do you install torch in your machine? i met this error, and i install suitable version torch then the error is disappeared.
Adding to @stargolike's comment, have you followed this part of the instructions (using the appropriate libtorch version)?
wget https://download.pytorch.org/libtorch/cpu/libtorch-shared-with-deps-1.13.0%2Bcpu.zip
unzip libtorch-shared-with-deps-1.13.0+cpu.zip
rm libtorch-shared-with-deps-1.13.0+cpu.zip
Hi! I fixed this issue with correct installation of libtorch. However now when I try to run a simulation I get the error :
"pair_mace does not support vflag_atom."
It possibly comes from a compute of stress/atom. I am attaching the LAMMPS script and .pt files as well. This script has previously been tested with other potentials and it at-least runs. Any help is appreciated!
send.zip
We don't currently offer a way to extract the local stress. Is that essential for your application?
It would be nice to put that on our todo list
@aishwaryo You can slove the vflg_atom issue using by my repository and this code to pair with lammps. @wcwitt I want to pull this code but I don't know which branch is correct one. Can you make a new one?
@pobo95 i have tried using your repo and code to pair with lammps. However, while training I am getting the following over all the epochs.
INFO: Epoch 12: loss=0.0000, RMSE_E_per_atom=243.9 meV, RMSE_F=0.0 meV / A
This does not happen while using the original MACE repo. Could you suggest a way through? I have also attached the log and the xyz file so that you can check.
combined.zip
Maybe you didn't use test sets, don't you?
Can you share the result with my data
train_validate.zip
set using this line?
mace_run_train --name="mace" --train_file="train.xyz" --valid_fraction=0.05 --test_file="validate.xyz" --E0s={14:-0.08755009} --model="MACE" --num_interactions=2 --hidden_irreps='64x0e + 64x1o' --max_L=1 --correlation=2 --r_max=5.0 --forces_weight=1000 --energy_weight=10 --batch_size=5 --valid_batch_size=2 --max_num_epochs=5000 --compute_stress=”true” --start_swa=450 --scheduler_patience=5 --patience=15 --eval_interval=3 --ema --swa --swa_forces_weight=10 --default_dtype="float64" --device=cuda --seed=123 --restart_latest --save_cpu --error_table=PerAtomRMSEstressvirials When I test with your data sets withi this line
mace_run_train --name="mace" --train_file="combined.xyz" --valid_fraction=0.05 --E0s=average --model="MACE" --num_interactions=2 --hidden_irreps='64x0e + 64x1o' --max_L=1 --correlation=2 --r_max=5.0 --forces_weight=1000 --energy_weight=10 --batch_size=5 --valid_batch_size=2 --max_num_epochs=5000 --compute_stress=”true” --start_swa=450 --scheduler_patience=5 --patience=15 --eval_interval=3 --ema --swa --swa_forces_weight=10 --default_dtype="float64" --device=cuda --seed=123 --restart_latest --save_cpu --error_table=PerAtomRMSEstressvirials It seems fine.
When I chckded your log file, looks like you didn't take max_L.
It can be the resaon too.