Run a command with a memo prefix and that particular file along with its outputs gets saved to a folder. You can track progress easily within your browser by starting server.py and opening http://localhost:5000/ in your browser.
Current limitation: only the file you execute gets saved. Thus, if your file has external dependencies, they must be specified with absolute paths and, unfortunately, they won't get saved.
No dependencies outside the standard Python 3 libraries.
The database (which is just a plain folder) is expected to be on braintree.mit.edu but can be modified freely in the config file.
You need to create a new environment variable MEMO that points to a folder where you will store your runs.
Create a config file in your home folder (~/.memo) with the following structure:
[db]
user = ...
host = braintree.mit.edu
[braintree]
user = ...
[om]
user = ...
qos = dicarlo
memo <your command as usual>
e.g.:
memo python run.py value1 --arg2 value2
Usually, you want to run it a couple of reserved memo arguments, such as -d (description).
In theory, you should be able to run your command on your local machine but specify that it should actually be executed on a remote server (e.g., --cluster braintree --node gpu3). I haven't tested it too much yet though.