Closed priyanka9991 closed 4 weeks ago
@priyanka9991 did you have a chance to test it? It looks good, but I assume it wouldn't work as is. I think os
should be imported inside each python_app
.
@adpartin Thanks for catching that. I added the imports inside the apps. Also tested on GraphDRP.
@wilke pointed out that if the subprocess call fails inside the parsl app for some reason (eg: missing model scripts), the corresponding output dirs (model_dir, infer_dir or ml_data_dir) will not be created. So, we need to check if the output directories exist and create them if needed, to save the log files.