labmlml will keep track of experiments if you declare an Experiment. It will keep track of logs, code diffs, git commits, etc.
from labml import experiment
name defaults to the calling python filename. However when
invoking from a Jupyter Notebook it must be provided because the library
cannot find the calling file name.
comment can be changed later from
Starting an experiments creates folders, stores the experiment configurations, git commits, and source diffs.
mnist_pytorch: 2a81805c946011eabd18acde48001122 Test [dirty]: "citation"
You can also start from a previously saved checkpoint.