Errors while installing OpenMM 8.0

[saratk1]

Hi openmm-ml developers,
I was trying to install the beta release of OpenMM 8.0 and I created a conda environment with compatible versions of all the packages as described here (conda env create mmh/openmm-8-beta-linux). When trying to run the openmm-ml tests (TestMLPotential.py) I got several error messages, eg.

E RuntimeError: Tried to instantiate class 'cuaev.CuaevComputer', but it does not exist! Ensure that it is registered via torch::class_

I also attach the full error log here: error_log.txt
Is there something I can do to fix this?

[JohannesKarwou]

I get the same error when trying to install the new release...

[peastman]

This looks to me like an issue with incompatible libraries in different packages. The error log contains the exception

E       ImportError: /data/shared/software/conda_environment/envs/endstate_test_2/lib/python3.10/site-packages/torchani/cuaev.cpython-310-x86_64-linux-gnu.so: undefined symbol: _ZN2at4_ops4full4callEN3c108ArrayRefIlEERKNS2_6ScalarENS2_8optionalINS2_10ScalarTypeEEENS8_INS2_6LayoutEEENS8_INS2_6DeviceEEENS8_IbEE

Can you post the output of conda list?

[saratk1]

Thanks for the response! The installation worked with mamba.