code-intelligence
code-intelligence copied to clipboard
ML-Powered Developer Tools, using Kubeflow
We should update our notebooks e.g. Get-GitHub-Issues.ipynb to use HDF5 instead of pickled. HDF5 files are better because they don't have pickling issues and we should be able to store...
Our synchronous training pipeline is currently spawning multiple instances of training rather than the expected 1 model per hour. The problem appears to be the code to decide whether to...
Right now needs sync only compares the latest model to the model on master. As a result, we end up firing off a Tekton pipeline if there is a PR...
We'd like to periodically (weekly) retrain the model and deploy the latest model to production. This would allow us to benefit from * Additional data to train from * Incorporate...
We currently post-process the CRD generated by kubebuilder here https://github.com/kubeflow/code-intelligence/blob/7f514162b674b3bcd28bc4600c7b2f22a7703c39/Label_Microservice/go/Makefile#L55 This post-processing removes validation for PipelineRunRef and just treats it as an arbitrary json. This is undesirable because we'd like...
The ModelSync Controller isn't really a model sync controller. Its more of a generic controller aimed at building declarative APIs. You specify a lambda that determines whether the world needs...
We'd like to be able to 1. Use GitOps to deploy the production version of the controller 1. Deploy a development version of the controller in a different namespace 1....
PR #77 contains skaffold.yaml files to continuously build and deploy the microservices. Skaffold is building the containers on the cluster using Kaniko. The builds look like they are taking O(minutes)...
To continue to reduce toil, I'd like to experiment with build a virtual assistant using Dialogflow to answer common queries. To get some of the scaffolding in place, I'd like...
Opening this issue to log organizational structure changes to our bot projects. Ultimately we want to manage them better using GitOps and ACM.