| description |
|---|
Run Weights and Biases on your own machines using Docker |
To run the W&B server locally you'll need to have Docker installed. Then simply run:
$ wandb local
Behind the scenes the wandb client library is running the wandb/local docker image, forwarding port 8080 to the host, and configuring your machine to send metrics to your local instance instead of our hosted cloud. If you want to run our local container manually, you can run the following docker command:
$ docker run --rm -d -v wandb:/vol -p 8080:8080 --name wandb-local wandb/local
Running wandb local configures your local machine to push metrics to http://localhost:8080. If you want to host local on a different port you can pass the --port argument to wandb local. If you want to configure DNS with your local instance you can run wandb login --host=http://wandb.myhost.com on any machines that you want to report metrics from. You can also set the WANDB_BASE_URL environment variable to a host or IP on any machines you wish to report to your local instance. To restore a machine to reporting metrics to our cloud hosted solution, run wandb login --host=https://api.wandb.ai.
The base install of wandb/local starts with a default user [email protected]. The default password is perceptron. The frontend will attempt to login with this user automatically and prompt you to reset your password. An unlicensed version of wandb will allow you to create up to 4 users. You can configure users in the System Settings page of wandb/local found at http://localhost:8080/admin
All metadata and files sent to W&B are stored in the /vol directory. If you do not mount a persistent volume at this location all data will be lost when the docker process dies. If you purchase a license for wandb/local, you can store metadata in an external MySQL database and files in an external storage bucket removing the need for a stateful container.
We are pushing new versions of wandb/local to dockerhub regularly. To upgrade you can run:
$ wandb local --upgrade
To upgrade your instance manually you can run the following
$ docker pull wandb/local
$ docker restart wandb-local
If you're interested in configuring teams, using external storage, or deploying wandb/local to a Kubernetes cluster send us an email at [email protected]