mle-infrastructure / mle-toolbox

Lightweight Tool to Manage Distributed ML Experiments 🛠
https://mle-infrastructure.github.io/mle_toolbox/toolbox/
MIT License
2 stars 0 forks source link

Allow external add. CLI inputs for jobs #57

Closed RobertTLange closed 3 years ago

RobertTLange commented 3 years ago

The Experiment class takes as an input a dict extra_cmd_line_input, which adds command line inputs to the execution of the target job script. E.g.:

extra_cmd_line_input = {"imp_iters": 10}

will add another argument -imp_iters 10 to the python job_fname.py -seed_id 0 -config_fname example.json execution. So far this feature is only used in the post_processing experiment type and requires a separate set of args in the .yaml file. I want to add this directly to single_job_args and to make the flexbility available to all experiment types.