README.md

Running the examples

Edit hparams.yaml to configure your conda environment and slurm job details. hparams.yaml right now will save experiment details to ./test_shopty/, optimize for the minimum result, and run the simple training script in ./train.py.

Change run_command in hparams.yaml if you want to see the neural network example in action. Be sure to modify the slurm directives so you run your experiments on a GPU (if you're on a cluster, which you probably should be.)

hyperband on a cpu:

shopty hyperband --config_file hparams.yaml --supervisor cpu

hyperband on a slurm cluster, submitting default number of experiments

shopty hyperband --config_file hparams.yaml --supervisor slurm

hyperband on a slurm cluster, only submitting 20 experiments max

shopty hyperband --config_file hparams.yaml --supervisor slurm -n 20

run 20 random hyperparameter configs each for 100 iterations on cpu

shopty random --config_file hparams.yaml --supervisor cpu --max_iter 100 --n_experiments 20

same thing, on slurm:

shopty random --config_file hparams.yaml --supervisor slurm --max_iter 100 --n_experiments 20