Edited by: @gaocegege
- π Repository about hyperparameter tuning
- π New demo: MNIST example using S3 for Training, Serving, and Tensorboard monitoring
Edited by: @gaocegege
The frontend is now using hashRouter instead of browserRouter, meaning the location of the UI is determined only by the portion of the url after the # (similar to k8s dashboard).
Request to the backend are also properly routed now. The backend request are sent to the portion of the url until the first occurence of tfjobs + /api i.e: http://127.0.0.1:8001/api/v1/namespaces/kubeflow/services/ambassador:80/proxy/tfjobs/ui/#/new will translate into: http://127.0.0.1:8001/api/v1/namespaces/kubeflow/services/ambassador:80/proxy/tfjobs/api
This means this would break if we decide to route the dashboard with ambassador on /dashboard instead of /tfjob, so the route should be part of a dashboard config at some point.
- Update images for our 0.1 release
- Upgrade ks version to 0.9.2
- Add parameter to mount path for notebook PVC
- Add Central UI
- Add TfJob dashboard to ksonnet
- v1alpha1: Fix e2e test failures
- v1alpha1: Support testing on minikube
- v1alpha1: Add proxying to front-end development server
- v1alpha1: Add a hack to create example TFJob config
- v1alpha1: Don't leave pods running when a job completes
- v1alpha2: Support condition in TFJobStatus and add corresponding test cases
- v1alpha2: Check running status more gracefully
- Add @mitake as reviewer π