For the most part, you can use W&B Enterprise the same way as our Software-as-a-Service offering. There are a few tweaks to keep in mind.
You'll need to fetch your API key from your W&B Server instead of
wandb.ai with one of the following methods. W&B Enterprise Server API keys have a dedicated prefix to avoid confusion between the SaaS system and your private installation.
# Save a wandb settings file in the `wandb` subfolder of your python path.mkdir -p ~/.config/wandbecho -e "[default]\nbase_url = http://your-server-ip-or-host" > ~/.config/wandb/settingswandb loginpython your-training-script.py# Or, configure the CLI target with environment variables.WANDB_BASE_URL=http://your-server-ip-or-host wandb loginWANDB_BASE_URL=http://your-server-ip-or-host python your-training-script.py
In jupyter notebooks, simply set the environment variable
WANDB_BASE_URL to the host or IP address of the server running W&B enterprise. For example:
import osos.environ['WANDB_BASE_URL'] = 'http://your-server-ip-or-host'# At this point you will be asked for an API KEY to your server.import wandbwandb.init()