Installation issue and setup failed

[wenchangzhou-qtx]

Hey @fjclark,

Thanks for such a wonderful automation package! I did a quick run and encountered two issues so far:
The first one is after installation, I only can import the module inside the installation directory but cannot load the module outside, not sure if this is because of my installation.

The second one is after I copied over the input files from a3fe/a3fe/data/example_run_dir/input, and run up to calc.setup(), I then got error saying RuntimeError: Could not find slurm output file name in /home/softwares/a3fe/input/solvate_bound.sh, do I missing anything there?

[fjclark]

Hey @wenchangzhou-qtx,

No problem - hope it's useful!

Installation Issue: The steps to install (once you have GROMACS and SLURM) are:

git clone https://github.com/michellab/a3fe.git
cd a3fe
mamba env create -f environment.yaml
python -m pip install --no-deps .

It sounds like you've missed the last step - python -m pip install --no-deps ., which installs the package into your environment. To check this, run conda list | grep a3fe to make sure that it's installed. For example, I see:

a3fe                      0.1.1+2.g80436ae.dirty          pypi_0    pypi

If you don't see anything, run python -m pip install --no-deps . in the a3fe base directory.

RuntimeError: This sounds like an issue with run_somd.sh. The options in this are used as a template to create all of your other SLURM scripts, such as solvate_bound.sh. Could not find slurm output file name shows that you need to specify the SLURM output file name with #SBATCH -o in run_somd.sh. To fix this, add e.g. #SBATCH -o somd-array-gpu-%A.%a.out to run_somd.sh, and make sure that all of the other SLURM options at the top of the script are correct for your cluster.

Hope that helps!

[wenchangzhou-qtx]

Hey @fjclark,

Thanks for the guide and suggestions!
Regarding the installation issue, I made sure I ran the last step and also got the exact output when I run conda list | grep a3fe, but still I cannot import a3fe outside the installation base directory. What else should I check?

When I import a3fe inside the installation directory, I can run the submission steps up to a3.setup() but still got the same error, this is my run_somd.sh for you to take a look

#!/bin/bash

#SBATCH --account=qt
#SBATCH --nodes=1
#SBATCH --time=24:00:00
#SBATCH --gres=gpu:1
#SBATCH --ntasks-per-node=1
#SBATCH --cpus-per-task=2
#SBATCH --exclude=node018
#SBATCH --mem=50G
#SBATCH --output=somd-array-gpu-%A.%a.out

lam=$1
echo "lambda is: " $lam

srun somd-freenrg -C somd.cfg -l $lam -p CUDA

[fjclark]

Hey,

Thanks for the additional details. The issue with the SLURM file was a bug - apologies for this. I forgot to handle the case when "=", rather than spaces are used (e.g. #SBATCH -o somd-array-gpu-%A.%a.out works but #SBATCH --output=somd-array-gpu-%A.%a.out doesn't. I've now fixed this and added corresponding tests (see cb0d532). If you pull the latest changes then your input should work - alternatively you can reformat your run_somd.sh to use spaces. Thanks for raising this and sorry for the sloppy error!

For the installation, can you check:

• That you are using the version of Python installed in your environment e.g. which python shows .../anaconda3/envs/mamba/envs/a3fe_test/bin/python for me.
• That A3FE is correctly linked in the environment. Type pip show a3fe. For me, I see Location: .../anaconda3/envs/mamba/envs/a3fe_test/lib/python3.12/site-packages
• That the install location is in the Python path. Start an ipython/python session and type

import sys
sys.path

I see .../anaconda3/envs/mamba/envs/a3fe_test/lib/python3.12/site-packages in the resultant list, matching the location given by pip show a3fe.

Hopefully this gives us a clue as to what's wrong! If it's still not working, please post the output of the above commands.

Thanks!

[wenchangzhou-qtx]

Thanks @fjclark !

Regarding the installation, I found the cause, guess it's a silly one from my side. I did not switch on the a3fe environment before running the last step, now it's working.

Regarding the test run, I figured the slurm output file option has to be in the first line in run_somd.sh, everything else needs to put after that. I will continue from here :)

[fjclark]

No problem!

Ah, glad you found the issue. My fault - I should update the installation notes to be explicit with this.

The slurm output file option shouldn't have to be the first line in run_somd.sh, but before I fixed the bug it didn't work with and "=" in e.g. #SBATCH --output=somd-array-gpu-%A.%a.out. The updated version should now work with either "=" or a space (and I now have a test which uses the run_somd.sh file you provided to check that the output file name can be read.

Great, good luck and let me know if there are any more issues!

[fjclark]

Hey @wenchangzhou-qtx,

Did it work for you? If so, feel free to close the issue. Thanks!

[wenchangzhou-qtx]

Yes, it's working for me now, thanks!