Packaging app¶
In this step, the MONAI Deploy application can now be built into a deployable Docker image using the MONAI Application Packager.
Overview on MONAI Application Packager¶
The MONAI Application Packager (Packager) is a utility for building an application developed with the MONAI Deploy App SDK into a structured MONAI Application Package (MAP). The Packager makes use of the package
command of the Holoscan SDK CLI, albeit with monai-deploy package
as the actual command.
The MAP produced by the Packager is a deployable and reusable docker image that can be launched locally or hosted.
It is required that the application configuration yaml file as well as the dependency requirements file are present in the application folder, as described here.
Basic Usage of MONAI Application Packager¶
monai-deploy package <APP_PATH> --config <COMFIG> --tag <TAG> --platform <x64-workstation> [--models <MODEL_PATH>] [--log-level <LEVEL>] [-h]
Required Arguments¶
<APP_PATH>
: A path to MONAI Deploy Application folder or main code.--config, -c <CONFIG>
: Path to the application configuration file.--tag, -t <TAG>
: A MAP name and optionally a tag in the ‘name:tag’ format.--platform <PLATFORM>
: Platform type of the container image, must bex64-workstation
for x86-64 system.
Note
If <APP_PATH>
refers to a python code (such as ./my_app.py
), the whole parent folder of the file would be packaged into the MAP container image, effectively the same as specifying the application folder path which includes __main__.py
file. In both cases, the image’s environment variable, HOLOSCAN_APPLICATION
will be set with the path of the application folder in the image, i.e. HOLOSCAN_APPLICATION=/opt/holoscan/app
. So, it is essential to provide the __main__.py
file in the application folder, which usually would look like below:
# Assuming that the class `App` (that inherits Application) is defined in `app.py`.
from app import App
if __name__ == "__main__":
App().run()
Optional Arguments¶
The following lists a few most likely used optional arguments
--models, -m <MODEL_PATH>
: Path to a model file or a directory containing all models as subdirectories-h, --help
: Show the help message and exit.--log-level, -l <LEVEL>
: Set the logging level ("DEBUG"
,"INFO"
,"WARN"
,"ERROR"
, or"CRITICAL"
).--sdk-version <SDK_VERSION>
: Set the version of the SDK that is used to build and package the Application. If not specified, the packager attempts to detect the installed version.--monai-deploy-sdk-file <MONAI_DEPLOY_SDK_FILE>
: Path to the MONAI Deploy App SDK Python distribution Wheel file.--version <VERSION>
: An optional version number of the application. When specified, it overrides the value specified in the configuration file.
Example of using MONAI Application Packager¶
Given an example MONAI Deploy App SDK application with its code residing in a directory ./my_app
, packaging this application with the Packager to create a Docker image tagged my_app:latest
would look like this:
$ monai-deploy package ./my_app -c --config ./my_app/app.yaml -t my_app:latest --models ./model.ts --platform x64-workstation
Building MONAI Application Package...
Successfully built my_app:latest
The MAP image name will be postfixed with the platform info to become my_app-x64-workstation-dgpu-linux-amd64:latest
, and will be seen in the list of container images on the user’s local machine when the command docker images
is run.
Note
The current implementation of the Packager ONLY supports a set of platform specific base images from
nvcr.io
as base images for the MAP.To package a MAP to run on ARMv8 AArch64 on Linux with discrete GPU, replace the commandline option
--platform x64-workstation
with--platform igx-orin-devkit --platform-config dgpu
. It has been tested on NVIDIA IGX Orin.
Next Step¶
See the next page to learn more on how to locally run a MONAI application package image built by the Packager.