Starter kit for H2O.ai Wildfire Challenge.
-
Install Python 3.6+, and pip3
-
Install H2O Wave SDK - follow instructions for your platform at https://wave.h2o.ai/docs/installation
-
Install H2O AI Cloud CLI (v0.9.1-rc1) to debug, bundle and execute your H2O Wave app: https://h2oai-cloud-release.s3.amazonaws.com/releases/ai/h2o/h2o-cloud/v0.9.1-rc1/index.html
-
Install
tar
(or an alternative, to create a compressed archive file for submission) -
Visit to mapbox and create a token. Add this token to wave-app/src/plot_v2.py
Go to your H2O Wave SDK directory and run the Wave server:
cd $HOME/wave && ./waved
INFO: On Windows, run
waved.exe
to start the server.
git clone https://github.com/h2oai/challenge-wildfires.git
- Use the provided sample for Australia
- Download the global raw data
- Search for other useful external datasets
cd wave-app
make setup
or simply
cd wave-app
python3 -m venv venv
./venv/bin/python -m pip install --upgrade pip
./venv/bin/python -m pip install -r requirements.txt
This step is using installed h2o-wave package to run the application.
cd wave-app
make run
or
cd wave-app
./venv/bin/wave run src.app
Point your web browser to http://localhost:10101/ to access the app.
This step prepares the Wave app for submission.
cd wave-app
make bundle
or
cd wave-app
h2o bundle
H2O.ai Wildfire & Bushfire Challenge enables participants to deploy, debug, and upload their H2O Wave apps on a managed H2O AI Cloud instance. H2O AI Cloud's Appstore operationalizes AI/ML applications built with H2O Wave. https://challenge.h2o.ai/ is a H2O AI cloud instance managed by H2O.ai and is available for use for Callenge Wildfire.
Developer Guide is available here: https://h2oai.github.io/h2o-ai-cloud/docs/userguide/developer-guide
Wildfire Challenge allows two usage modes for the participants on the cloud:
-
publish-cloud-private: immediately run your current app source in the platform. This command will automatically package your current directory into a .wave bundle, import it into the platform, and run it privately (only visible to you). In the output you will be able to find a URL where you can reach the instance, or visit the "My Instances" in the UI.
-
publish-cloud-public: publish an app to the platform. This command will automatically package your current directory into a .wave bundle and import it into the platform. The app will be visible and available to run for all participants. Participants will be run an instance on H2OAIC Appstore.
To get started, please follow the steps below:
Note: For ease of use, config setup steps have been automated for you. When you get to the token portion, you will need to visit https://challenge.h2o.ai/auth/get-token in order to obtain your token. After entering the token here, you are all set.
WARNING: please ensure that the newly generated config file,
h2o_wildfire_cli_config.toml
, is confidential.
cd wave-app
make generate-cloud-config
cd wave-app
make publish-cloud-private
WARNING: this mode will allow all participants to view and launch an instance of your H2O Wave app on the Appstore.
cd wave-app
make publish-cloud-public
This operation is going to create a new archive file in the root directory of the repo called submission.tar
. The archive follows challenge rules and contains the wave app, Python notebook, and this README.
cd wave-app
make submission
starter_kit_intro.mp4
There are several communities to discuss topics related to AI/ML or application development:
- H2O.ai Community Slack: http://h2oai-community.slack.com/ provides a space to discuss AI/ML related topics or questions related to H2O.ai open source tools.
- H2O.ai Challenge Forum: https://discuss.challenge.h2o.ai/ is a space to discuss challenges, tools, dataset, and ideas.
- H2O.ai Wave Discussions: https://github.com/h2oai/wave/discussions includes technical topics about Wave and Wave applications development.