This repository was archived by the owner on Jul 17, 2023. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 5
Standalone
Rohan Rao edited this page May 13, 2021
·
4 revisions
This guide shows how to use Wave ML as a standalone package. This might be useful for a developer who wants to try it in a notebook before doing a regular Wave app.
Although Wave ML is meant to be used along with Wave, it's independent enough to be run in standalone.
There is no special setup if you want to use H2O-3 backend. Just import package and use provided functions.
This setup will use a private cloud used for Wave ML development. The steps necessary to enable Steam + DAI + MLOps integration:
- Install the Wave ML package from the GitHub repo. Check releases section. (The package on
pipdoesn't have all dependencies.) - Set up your OpenID account. Ask on the slack channel.
- Create a DAI instance in Steam here (login with OpenID). We will be able to use DAI multinode cluster in the future, making this step obsolete.
- Use these variables in your environment:
export H2O_WAVE_OIDC_CLIENT_ID=
export H2O_WAVE_OIDC_CLIENT_SECRET=
export H2O_WAVE_OIDC_PROVIDER_URL=http://keycloak.44.238.254.140.nip.io/auth/realms/wave
export H2O_WAVE_OIDC_REDIRECT_URL=http://localhost:10101/_auth/callback
export H2O_WAVE_OIDC_END_SESSION_URL=http://keycloak.44.238.254.140.nip.io/auth/realms/wave/protocol/openid-connect/logout
export H2O_WAVE_ML_STEAM_ADDRESS=https://steam.44.238.254.140.nip.io/
export H2O_WAVE_ML_STEAM_INSTANCE_NAME=
export H2O_WAVE_ML_MLOPS_GATEWAY=http://api.44.238.254.140.nip.io/
export H2O_WAVE_ML_STEAM_VERIFY_SSL=False- Provide the missing credentials and your DAI instance name. Ask on the slack channel.
- Go to this address http://h2oaicloud.44.238.254.140.nip.io/auth/get-token , authenticate with OpenID and use the token provided in the API calls where
refresh_tokenis required (build_model(),get_model()).
Project Management
Guides
Documentation