pip install wandb you should have a new command available, wandb.
The following sub-commands are available:
Open documentation in a browser
Configure a directory with W&B
Login to W&B
Disable W&B in this directory, useful for testing
Ensure W&B is enabled in this directory
Run a docker image, mount cwd, and ensure wandb is installed
Add W&B environment variables to a docker run command
Pull files for a run from W&B
Restore code and config state for a run
Launch a non-python program, for python use wandb.init()
List runs in a project
Sync a local directory containing tfevents or previous runs files
List current directory status
Create a new sweep given a YAML definition
Start an agent to run programs in the sweep
restore to return to the state of your code when you ran a given run.
# creates a branch and restores the code to the state it was in when run $RUN_ID was executedwandb restore $RUN_ID
How do we capture the state of the code?
wandb.init is called from your script, a link is saved to the last git commit if the code is in a git repository. A diff patch is also created in case there are uncommitted changes or changes that are out of sync with your remote.