Make it possible to run the same grid search for multiple base configurations within a single experiment. E.g. by stating them in the experiment config:
# Meta Arguments: What job? What train .py file? Base config? Where to store?
meta_job_args:
project_name: "drl-lth"
experiment_type: "hyperparameter-search"
base_train_fname: "cluster_exec_mle.sh"
base_train_config:
- "configs/train/bullet/config_1.json"
- "configs/train/bullet/config_2.json"
experiment_dir: "runs/bullet/calibrate/v3_2/"
report_generation: false
The hyper_log.pkl should in that case have a separate variable that saves the configuration file name. It would also be nice to automatically calculate the number of jobs/batches given a fixed amount of jobs per batch/max to run.
Make it possible to run the same grid search for multiple base configurations within a single experiment. E.g. by stating them in the experiment config:
The
hyper_log.pklshould in that case have a separate variable that saves the configuration file name. It would also be nice to automatically calculate the number of jobs/batches given a fixed amount of jobs per batch/max to run.