kaos
kaos copied to clipboard
KI-labs
open source platform for simplifying machine learning deployment
#### What is the current behaviour? Build serve failures cause "failed" serve pipelines. #### What is the expected behaviour? Removed serve pipelines if build serve fails.
#### What is the feature/concept/idea? We only output model quality scores in `train info`, but it is highly likely that user would like to compare models they trained with some...
#### What is the current behaviour? Error from GCP when supplying a namespace within the kubeconfig from `kubectl`. #### What is the expected behaviour? No error. #### How to reproduce?...
#### What is the feature/concept/idea? The current version of the CI/CD is not testing the infrastructure. It should be properly implemented for a branch with a specific cloud provider name...
#### What is the feature/concept/idea? Add the ability to kill an ingestion job. An incorrectly deployed manifest can hang, which should be killed from the client side. #### How will...
#### What is the current behaviour? JOB_SUCCESS for a running training job when it is still running. #### What is the expected behaviour? Only show JOB_SUCCESS when a job is...
#### What is the desired change? As a SuperHero user going through the *Quick Start Guide*, I didn't know whether I can/can't build more than one cluster so I did...
#### What is the feature/concept/idea? Sharing a cluster with kaos will likely result in poor management of resources. For this reason, a user should be allowed the following when creating...
#### What is the desired change? kaos 2.0 should be cloud-agnostic and deployable from a single command from the CLI. Thus, it should be available to run in any cloud....
#### What is the current behaviour? Any image stored on ACR registry cannot be pulled by a pachyderm sidecar image (supporting a pipeline image). #### What is the expected behaviour?...