Usage
Description
Run code in a Docker container with W&B configured. Start a Docker container, inject the WANDB_DOCKERs environment variable, and mount the current working directory at/app by
default. Inject WANDB_API_KEY if logged in. Override the
container entrypoint to ensure wandb is installed.
Pass additional arguments to insert them into docker run before
the image name. Use a default image if none is specified.
Use --jupyter to install and start JupyterLab on port 8888.
Enable the NVIDIA runtime automatically if NVIDIA Docker is
available on the host.
Requires Docker to be installed and running on the host machine.
To inject W&B environment variables into an existing docker run
command without modifying the entrypoint, use wandb docker-run.
For example, to run the default image and mount a dataset into the container:
Arguments
| Name | Default | Type |
|---|---|---|
docker_run_args | STR | False |
docker_image | STR | False |
Options
| Flag | Type | Description |
|---|---|---|
--nvidia | BOOL Flag | Use the nvidia runtime, defaults to nvidia if nvidia-docker is present. Default: False |
--digest | BOOL Flag | Output the image digest and exit. Default: False |
--jupyter | BOOL Flag | Run jupyter lab in the container. Default: False |
--dir | STR | Which directory to mount the code in the container. Default: /app |
--no-dir | BOOL Flag | Don’t mount the current directory. Default: False |
--shell | STR | The shell to start the container with. Default: /bin/bash |
--port | STR | The host port to bind jupyter on. Default: 8888 |
--cmd | STR | The command to run in the container. Default: None |
--no-tty | BOOL Flag | Run the command without a tty. Default: False |