CPSIGN CLI DEMO

Here is a short demo of how to run CPSign using the command line interface (CLI) in an efficient manner. The demo is written for Mac and linux-based platforms and test-run in bash, minor adjustments might be required to run on a different platform. To run the examples, it is assumed that the commands are runned from the cli-api directory, as all paths are relative to this directory. It is also required that you have a recent version of cpsign fatjar which can be downloaded from CPSign releases page.

Setting up an environment

When downloading cpsign from GitHub the execution file privileges are stripped, so here we recommend/assume that you create an alias for running cpsign:

alias cpsign='java -jar <path-to-cpsign>'

This will make the following commands identical:

> java -jar <cpsign-jar> train @args
> cpsign train @args

Run some basic commands:

> cpsign --version
> cpsign

Precompute

From running cpsign without any extra parameters, we saw the top-most usage text which lists the available "programs" that can be run. The first step is typically to start with precompute - in order to compute descriptors. First check the available parameters for precompute:

> cpsign precompute

That shows a lot of parameters which might be daunting at first. Note that only the parameters following a star (*) are required parameters and that there are default parameters for the remaining ones. Some arguments might be more complex due to syntax or require further information, these you can typically get more information about using the explain program, e.g. to see more information about the available descriptors and sub-arguments run:

> cpsign explain descriptors

There is a file precompute.args that is already filled in with the parameters necessary for a rudimentary default precomputed model. Run this using:

> cpsign @precompute.args

Evaluate the default parameters

Check what parameters that we can use

> cpsign cv

Notice that we have the --data-set parameter, which takes a precomputed dataset - no need to add labels, property, transformers or descriptors again. Check out the file cv.args and run it using:

> cpsign @cv.args

Training a final model

The train program has similar parameters as the crossvalidate, but here we need to define where the model should be saved and give it a name. Again we can print the usage help text to see the available parameters:

> cpsign train 

Run train using the train.args file.

> cpsign @train.args

Predict using the model

The predict program requires a trained model to be given and only have parameters for what should be predicted and the type of output that should be generated. There are quite a few tweakable parameters, again to display the usage manual run:

> cpsign predict

Note that there are many parameters prefixed with -si:/--si: or -gi:/--gi: which are for image generation of either significant signature or atom gradients. Look in the predict.args file to see how some of these are used, to predict a single SMILES you can run the file using:

> cpsign @predict.args

Concluding remarks

This concludes this demo, for more information we refer to the CPSign readthedocs page and the CLI usage help texts that should explain the most important parameters - often more information is available using the explain program - run cpsign explain in order to list the available extra help texts.