Skip to content

Commit

Permalink
feat: Add 2020 roadmap (kubeflow#1121)
Browse files Browse the repository at this point in the history
* feat: Add 2020 roadmap

Signed-off-by: Ce Gao <gaoce@caicloud.io>

* fix: Address comments in kubeflow#1121

Signed-off-by: Ce Gao <gaoce@caicloud.io>

* fix: Address comments

Signed-off-by: Ce Gao <gaoce@caicloud.io>
  • Loading branch information
gaocegege authored and alfred.xx committed Apr 7, 2020
1 parent 5e84a01 commit e2f5666
Showing 1 changed file with 18 additions and 30 deletions.
48 changes: 18 additions & 30 deletions ROADMAP.md
Original file line number Diff line number Diff line change
@@ -1,43 +1,31 @@
# Katib 2020 Roadmap

This document provides a high level view of where Katib will grow in 2020.
## New Features

The original Katib design document can be found [here](https://docs.google.com/document/d/1ZEKhou4z1utFTOgjzhSsnvysJFNEJmygllgDCBnYvm8/edit#heading=h.7fzqir88ovr).
### Hyperparameter Tuning

# Katib 1.0 Readiness
- Support Early Stopping [#692](https://github.com/kubeflow/katib/issues/692)

* Stabilize APIs for Experiments
* Reconsider the design of Trial Template [#906](https://github.com/kubeflow/katib/issues/906)
* Early Stopping [#692](https://github.com/kubeflow/katib/issues/692)
* Resuming Experiment [#1061](https://github.com/kubeflow/katib/issues/1061), [#1062](https://github.com/kubeflow/katib/issues/1062)
* Fully integrate Katib with existing E2E examples:
* Xgboost
* Mnist
* GitHub issue summarization
* Publish API documentation, best practices, tutorials
* [Issues list](https://github.com/kubeflow/katib/issues)
### Neural Architecture Search

# Enhance HP Tuning Experience
- Support Advanced NAS Algorithms like DARTs, ProxylessNAS [#461](https://github.com/kubeflow/katib/issues/461)

The objectives here are organized around the three stages defined in the CUJ:
### Other Features

## 1. Defining Model and Parameters
- Support Auto Model Compression [#460](https://github.com/kubeflow/katib/issues/460)
- Support Auto Feature Engineering [#475](https://github.com/kubeflow/katib/issues/475)

Integration with KF distributed training components
* TFJob
* PyTorch
* Allow Katib to support other operator types generically [#341](https://github.com/kubeflow/katib/issues/341)
## Enhancements

## 2. Configuring a Experiment
* Supporting additional suggestion algorithms [#15](https://github.com/kubeflow/katib/issues/15)
### Hyperparameter Tuning

## 3. Tracking Model Performance
* UI enhancements: allowing data scientists to visualize results easier
* Support for persistent model and metadata storage
* Ideally users should be able to export and reuse trained models from a common storage
- Delete Suggestion deployment after Experiment is finished [#1061](https://github.com/kubeflow/katib/issues/1061)
- Save Suggestion state after deployment is deleted [#1062](https://github.com/kubeflow/katib/issues/1062)
- Reconsider the design of Trial Template [#906](https://github.com/kubeflow/katib/issues/906)
- Add validation for algortihms (a.k.a suggestions) [#1126](https://github.com/kubeflow/katib/issues/1126)

# Test and Release Infrastructure
### Neural Architecture Search

* Improve e2e test coverage
* Improve test harness
* Enhance release process; adding automation (see https://bit.ly/2F7o4gM)
- Refactor structure for NAS algorithms [#1125](https://github.com/kubeflow/katib/issues/1125)
- Refactor the design for NAS model constructor[#1127](https://github.com/kubeflow/katib/issues/1127)
- Katib UI fixes and enhancements

0 comments on commit e2f5666

Please sign in to comment.