Closed adamrhyshill closed 5 months ago
The goal of ichor is to produce the ML csv files which contain all the required information in them, so that any program can then be used to make models or do further processing of the datasets. I do not think adding program-specific config files is necessary because the information in the config file (atom names and ALFs) is already in the csv files and can be parsed easily already. Adding ALFs and atom names to the config file should not be necessary.
After producing the CSVs from ichor, these are then used with a sampling script to create training, validation and test sets, then plugged into FEREBUS for model training. FEREBUS uses a job parameters file which would be useful to generate automatically from ichor without users having to create this manually. Attached is an example of the file format (hardcoded filename is "job-details" without .txt) job-details.txt