Toolkit for deploying experiments on a cluster.
e.g.:
python3 toolkit/run_exp.py --gin ginfile.gin --user username --branch branch_name --gpu-count 4
usage: run_exp.py [-h] [--gin GIN] --user USER --branch BRANCH [--gpu-count GPU_COUNT] [--gpu-type {1080ti,titanx,titanv,rtx2080ti}]
[--node {asusgpu1,asusgpu2,asusgpu3,asusgpu4,asusgpu5,asusgpu6,arnold,steven,sylvester}] [--script SCRIPT] [--install] [--reinstall] [--ckpt CKPT]
optional arguments:
-h, --help show this help message and exit
--gin GIN gin file path
--user USER username on the cluster
--branch BRANCH branch name
--gpu-count GPU_COUNT
number of gpu
--gpu-type {1080ti,titanx,titanv,rtx2080ti}
type of gpu
--node {asusgpu1,asusgpu2,asusgpu3,asusgpu4,asusgpu5,asusgpu6,arnold,steven,sylvester}
type of gpu
--script SCRIPT custom script
--install Install full global venv along with downloading dataset
--reinstall Reinstall full global venv - without readownloading dataset
--ckpt CKPT Folder name in vatican workspace where checkpoint is stored