Revision 8cc162e593c41df13c6e8824cce95c612c6a90fc authored by YujiOshima on 19 December 2018, 01:34:46 UTC, committed by YujiOshima on 19 December 2018, 15:14:48 UTC
Signed-off-by: YujiOshima <yuji.oshima0x3fd@gmail.com>
1 parent 28c5b1c
README.md
# Katib
[![Go Report Card](https://goreportcard.com/badge/github.com/kubeflow/katib)](https://goreportcard.com/report/github.com/kubeflow/katib)
<img src="./img/Katib_Logo.png" width="320px">
Hyperparameter Tuning on Kubernetes.
This project is inspired by [Google vizier](https://static.googleusercontent.com/media/research.google.com/ja//pubs/archive/bcb15507f4b52991a0783013df4222240e942381.pdf). Katib is a scalable and flexible hyperparameter tuning framework and is tightly integrated with kubernetes. Also it does not depend on a specific Deep Learning framework (e.g. TensorFlow, MXNet, and PyTorch).
## Name
Katib stands for `secretary` in Arabic. As `Vizier` stands for a high official or a prime minister in Arabic, this project Katib is named in the honor of Vizier.
## Concepts in Google Vizier
As in Google Vizier, Katib also has the concepts of Study, Trial and Suggestion.
### Study
Represents a single optimization run over a feasible space. Each Study contains a configuration describing the feasible space, as well as a set of Trials. It is assumed that objective function f(x) does not change in the course of a Study.
### Trial
A Trial is a list of parameter values, x, that will lead to a single evaluation of f(x). A Trial can be “Completed”, which means that it has been evaluated and the objective value f(x) has been assigned to it, otherwise it is “Pending”.
One trial corresponds to one job, and the job kind can be [k8s Job](https://kubernetes.io/docs/concepts/workloads/controllers/jobs-run-to-completion/), [TFJob](https://www.kubeflow.org/docs/guides/components/tftraining/) or [PyTorchJob](https://www.kubeflow.org/docs/guides/components/pytorch/), which depends on the Study's worker kind.
### Suggestion
A Suggestion is an algorithm to construct a parameter set. Currently Katib supports the following exploration algorithms:
* random
* grid
* [hyperband](https://arxiv.org/pdf/1603.06560.pdf)
* [bayesian optimization](https://arxiv.org/pdf/1012.2599.pdf)
## Components in Katib
Katib consists of several components as shown below. Each component is running on k8s as a deployment.
Each component communicates with others via GRPC and the API is defined at `pkg/api/api.proto`.
- vizier: main components.
- vizier-core : API server of vizier.
- vizier-db
- suggestion : implementation of each exploration algorithm.
- vizier-suggestion-random
- vizier-suggestion-grid
- vizier-suggestion-hyperband
- vizier-suggestion-bayesianoptimization
- modeldb : WebUI
- modeldb-frontend
- modeldb-backend
- modeldb-db
## Getting Started
Please see [MinikubeDemo.md](./examples/MinikubeDemo.md) for more details.
## Web UI
Katib provides a Web UI.
You can visualize general trend of Hyper parameter space and each training history.
![katibui](https://user-images.githubusercontent.com/10014831/48778081-a4388b80-ed17-11e8-938b-fc59a5d2e574.gif)
## API Documentation
Please refer to [api.md](./pkg/api/gen-doc/api.md).
## CONTRIBUTING
Please feel free to test the system! [developer-guide.md](./docs/developer-guide.md) is a good starting point for developers.
## TODOs
* Integrate KubeFlow (TensorFlow, Caffe2 and PyTorch operators)
* Support Early Stopping
* Enrich the GUI
![swh spinner](/static/img/swh-spinner.gif)
Computing file changes ...