Simon Mo
Simon Mo
We plan to start working on aggregating logs from Clipper containers using [fluentd](https://www.fluentd.org/). This can be divided in the few PRs. @rkooo567 will leads the effort. ### PR 1: Pipe...
Currently, the frontend-exporter just flatten the metric json and log each sub-metric as guage. We need to use Prometheus label system for our metrics. For example, ```json { "app:testapp1-app:prediction_latency": {...
Since the python2 EOL is near and python's static type checking tool is getting better and better, we should consider type annotate the `clipper_admin` code base.
User is reporting: > I set up clipper on a Kubernetes cluster and tried to see how models perform. I found that the response time of models (even the helloworld...
I’m starting to work on overhauling our integration test. Our currrent integration test is slow and flaky. We have the following limitation and requirement: - We can run only _one_...
This issue thread is used to gather community feedback on next Clipper version. - What major feature do you want to see in Clipper? - What model deployer do you...
Currently we are not pining version for some of our `clipper_admin` dependencies. We should pin them down. Never use '*'
A blog post?
We should install some bare minimum packages such that model that depens on scimitar-learn can be deployed out of the box. I suggest - numpy - scipy - scikit-learn
Right now we run the docker image builder and then the linter. These two should swap their order or at least run in parallel.