This repository houses the main driver for executing an AutoMech workflow.
The AutoMech code is in development, so we encourage users to install it in developer mode and send us bug reports. See here for instructions to get set up.
End-users who are unable to contribute (see above) can install the code using Conda, Mamba, or Pixi.
conda install automech # option 1
mamba install automech # option 2
pixi add automech # option 3
Before running the above command, you will need to add auto-mech to your list of
channels:
conda config --append channels auto-mech # option 1 and 2
pixi project channel add auto-mech # option 3
If conda-forge isn't the default channel for your Conda/Mamba installation, you will
also need to set this additional configuration using the command above with --prepend.
To run the main AutoMech workflow, you will need an inp/ directory with input files as in one of the examples here. You can then run the main workflow as follows:
automech run &> out.log &
To see live output from the workflow, you will need to turn off Python buffering before you run the above command.
export PYTHONUNBUFFERED=1
You can then refresh your shell before running the above command and run tail -f out.log to see the live output and monitor progress.
Workflow parallelization is currently not automated in AutoMech. However, if you are on a cluster with direct SSH node access and permissions to run, you can run the following commands to split an AutoMech workflow into subtasks and run them in parallel.
(1.) You can set-up these subtask jobs as follows:
automech subtasks setup
This will parse your inp/ directory and create individual subdirectories for running each individual task for each individual species or reaction/TS. These directories will go in a folder called subtasks/.
(2.) If you are using the amech-dev Pixi environment, you can run the subtasks in parallel on a list of nodes as follows:
automech subtasks run csed-00{08..10} # expands to csed-0008 csed-0009 csed-0010
If you are running in a different environment, you will need to pass in an activation hook as follows:
automech subtasks run csed-00{08..10} -a <activation hook>
Where the activation hook contains the bash commands to activate your environment.
(3.) To check the progress of your subtask run, you can use the following command:
automech subtasks status
This will print a color-coded table showing which tasks have failed for which species/reactions. It will also generate a check.log file with the paths to log files that have have not completed successfully or have a warning.