Andrey Velichkevich
Andrey Velichkevich
/kind feature User should be able to sort Trial table by columns and Trial table should have pages. Ref PRs for the old UI: https://github.com/kubeflow/katib/pull/1251, https://github.com/kubeflow/katib/pull/1262. /cc @kimwnasptd
/kind feature We should add unit test for our [File Metrics Collector](https://github.com/kubeflow/katib/tree/master/pkg/metricscollector/v1beta1/file-metricscollector) to verify log parsing. That will help to avoid issues similar to: https://github.com/kubeflow/katib/issues/1754 /help /good-first-issue --- Love this...
/kind feature The new Katib UI should have a page to manage Trial templates. /cc @kimwnasptd
/kind feature The new Katib UI should show the YAML for the Experiment's Suggestion. One solution can be: 1. Rename YAML tab to Experiment YAML 2. Add new tab Suggestion...
/kind feature The new Katib UI should work in the standalone mode when user deploys Katib without Kubeflow central dashboard. Let's continue discussion here how we should do that. Also...
/kind feature Since Katib supports various frameworks with the same algorithms: https://www.kubeflow.org/docs/components/katib/katib-config/#suggestion-settings, we should try to be consistent with Algorithm Settings names. For example, [`n_EI_candidates `](https://github.com/kubeflow/katib/blob/master/pkg/suggestion/v1beta1/hyperopt/service.py#L72) in TPE algorithm in...
/kind feature To follow discussion in this PR: https://github.com/kubeflow/katib/pull/1667, we should add section: `Why Katib ?` in the README. cc @johnugeorge @gaocegege
/kind discussion We discussed on the AutoML and Training summit on 2021-07-16 that **Experiments (AutoML)** and **Experiments (KFP)** in Kubeflow Central Dashboard might be confusing for our users. Let's start...
/kind feature We discussed in our WG meetings and in the summit on 2021-07-16 about Trial training container's source code automation. Many users request automation for container build instead of...
/kind feature In the AutoML and Training summit on 2021-07-16 we discussed that Bayesian Optimisation and Bandit Based approach ([BOHB](https://arxiv.org/pdf/1807.01774.pdf)) produces very good results in case of HP tuning. Let's...