Giter Site home page Giter Site logo

federatedai / fate-board Goto Github PK

View Code? Open in Web Editor NEW
97.0 28.0 61.0 13.9 MB

FATE's Visualization Toolkit

License: Apache License 2.0

Java 39.84% HTML 0.07% JavaScript 0.79% Vue 27.24% Shell 1.00% SCSS 1.33% TypeScript 29.72%
visualization machine-learning fate federated-learning

fate-board's Introduction

DOC|中文

FATEBoard as a suite of visualization tool for federated learning modeling designed to deep explore models and understand models easily and effectively.

To make it easier to understand, track, debug, and explore federated learning modeling, as well as examine, evaluate, and compare various federation learning models. FATEBoard provides a visual way to probe models, from which you can reshape and improve models efficiently.

As a built-in visualization tool for FATE, FATEBoard presents abundant data visualization and model visualization information. In the Running Management, you can monitor the overall progress of all job runs. In the Dashboard, you can track the progress and logs of the algorithm components in real time. And in the Job Management, you can obtain component parameters, and view the results of component execution output,including model output,data output, log output, etc.

Usage

Starting FATEBoard

The FATEBoard source code uses the spring-boot framework and the embedded tomcat container. The default web port provided is 8080. Before starting, it is necessary to check whether port 8080 of the machine is already occupied. If the port is occupied, FATEBoard will fail to start.

FATEBoard access FATE-Flow through HTTP protocol. If FATE-Flow is not started properly, FATEBoard will not display charts.

You can access FATEBoard by visiting http://{fateboard-ip}:8080.

Starting a new job

  1. Preparing a job: you should configure a pipeline, create graph, and define parameters.
  2. Launching FATEBoard: submit the job and get started, returning a job URL. Click the URL to view the job on web. Alternatively, once the job is running, you can navigate your web browser to http://{fateboard-ip}:8080 to view the FATEBoard. Host or arbiter can also access the FATEBoard through http://{fateboard-ip}:8080.
  3. Monitoring the job: dashboard visualizes the whole running progress.
  4. Viewing the job outputs: you may view visual model outputs, data outputs and logs for each component.

FATEBoard can be used in Google Chrome, IE (10.0 and above) and other mainstream browsers. Some browsers might work, but there may be bugs or performance issues.

The Visualizations

Job Dashboard

FATEBoard’s dashboard visualizes basic statistics that vary over time, which include running time of job, real-time Log of job, running status for each component. Once you submit your job, you may have to wait for it to run. You can check the RUNNING page to see the progress of all running jobs and all waiting jobs.

Job Visualization

Job visualization provides overviews of the overall execution of the job, visualizes all the results as much as possible. There are some simple interactions as following:

  • Basic information about the job is shown on the left side of the page
  • Clicking on component and you can see the parameters
  • Clicking on the button under the parameter and you can see the component‘s outputs,Switching tabs to see model output/summary/metric, data output, and log

Visualizing the job workflow

The job workflow of federated learning modeling is easy to understand, which can help you track the running progress intuitively. For each role, you may see your own graph in the federated learning modeling.

Visualizing the model graph

FATEBoard provides rich and diverse visualization methods for federated learning models. You can compare the performance of multiple training models on the same dataset, or inspect a single model’s performance for continued tuning and training, which all probe your models better.

Visualizing the data

Preview the data of each component and you can view 100 lines of output data, from which you can also see the prediction data, including prediction result, prediction score and prediction detail.

Frequently Asked Questions

My FATEBoard isn’t showing any data of components!

FATEBoard sends a request to access FATE-Flow via HTTP to obtain all the data needed by a model. Log of httpclient is defined separately in the logback.xml in the source code, through which you can check communication between FATEBoard and FATE-Flow, and you can easily locate the problem if there is an exception.

My FATEBoard isn’t showing any log!

FATEBoard gets the list of jobs and details by querying fateflow. In a stand-alone environment, FATEBoard reads the local log file and returns it to the user through WebSocekt. If the log file cannot be displayed, you can first check whether the local log file has been generated. In a clustered environment, FATEBoard could access log files on different machines with SSH, and push them to the browser through WebSocket. The default log lookup path is /data/projects/fate/logs. If you cannot view the logs, it may be an error in SSH information in the cluster. you can set the correct SSH information by clicking the button in the upper right corner of the page.

fate-board's People

Contributors

caigaga avatar caohuachun avatar dependabot[bot] avatar dylan-fan avatar easson001 avatar forgivedengkai avatar happycooperxu avatar idwenwen avatar jarviszeng-zjc avatar jat001 avatar mgqa34 avatar pizhicheng avatar wgcitgkaka avatar wmqwxb avatar

Stargazers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

Watchers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

fate-board's Issues

Update readme.md

1.Optimize discriptions for fateboard
2.Add chinese document of readme
3.Add new document of deploy

[Feature Request] Enable Customized Prefix for URLs

Hi, Is it possible to add a customized prefix to the Server-side?
In our scenario, we route requests base on the URL rather than the hostname, assumes we have a gateway and two fateboard services as follows:

          ---->  fateboard1
gateway -|
          ---->  fateboard2

we use the following URLs to access fatebord1 and fateboard2 respectively:

Before forwarding the request, the gateway needs to rewrite the URL from /fateboard1 to /, and then send the request to fateboard1. However, the further requests triggered by the JS will be sent to an address without fatebord* pattern, for example, "http://gateway.domain/ssh". Resulting in the gateway can't forward this request correctly.

The ideal solution is to add a variable to the routing policy like ${prefix}/, if the prefix is empty, then the use /.
If the prefix is fateboard1, then use /fateboard1/ as the root. So that the rewrite could be avoided and further request can contain the fatebord* part.

logs bug

Update the condition to pull logs.

Frontend source code has not been updated

It seems in recently updates only the built version of frontend code has been updated in FATE-Board/src/main/resources/static/ and the source code in FATE-Board/static/ hasn't been updated for a while.

FATE-Board V1.1 Web front end outcomes of the second round of proposals

  1. The last round of bug repair;

  2. Secure Boost tree zooming method optimization, display content modification;

  3. Selection feature selection (from chart to table) to separate guest and host data;

  4. Redraw dashboard;

  5. Running, dashboard, detail pages polling for query data;

  6. Workflow is displayed according to the role: guest, host or arbiter are grayed out in the workflow, and component names are not displayed.

Inappropriate getting the fate path in `JobManagerService.java`

Hi, I'd like to report a bug.

In

//get fate path
String webPath = System.getProperty("user.dir");
int i1 = webPath.lastIndexOf("/");
String fatePath = webPath.substring(0, i1)+"/fateflow";

This only works fine in the standalone deployment. (That is, the parent directory of the fate-flow and the fate-board is the same),
In the following situation, it will trigger a File doesn't exist! exception when the user wants to download the dsl/runtime conf in the job details dashboard. Please see the pic follows.

  1. In the kube-fate docker-compose deployment: the fate-flow and fate-board are not in the same container.
  2. In the standalone deployment: If the user runs the spring-boot application in another directory(not the /standalone_fate_install_1.7.0_release/fateboard directory )

image

Since the dsl and the runtime conf have already been stored in the table t_job, I'd like to suggest getting them from the database. If this makes scene, May I submit a PR?

update versiont to 1.5.0

support more display of algorithm components

  • Add model visualization of new algorithm components:Hetero/Homo Data Split, Table reader, Data Statistic, PSI,Hetero Fastsecureboost,Hetero Kmeans

    • Hetero/Homo Data Split:Visualize dataset split results (training/validate/test)
    • Table reader:read in the original modeling data and output data overview
    • Data Statistic:display statistical indicators of each variable
    • PSI:PSI detailed data display for each variable
    • Hetero Fastsecureboost:fast secureboost model visualization in layered-mode and mixed-mode
    • Hetero Kmeans:visualization of sample clustering results
  • Feature selection:supports connection with the output models of Data Statistic, PSI, SecureBoost, and other components for feature selection

  • Feature binning:Support binning visualization without calculating the iv value, and increase the display of the number of bins

  • Evaluation:Supports visualization of cluster model evaluation results

  • Hetero secureboost:Supports visualization of Completely SecureBoost models

  • Evaluation:Supports visualization of cluster model evaluation results

  • Support downloading model and data of algorithm components from FATEBoard(Feature Binning,Feature Selection,Secureboost,LR,Evaluation)

  • The page supports partial data refresh and global data refresh to update the data in the iteration

  • Support retry for failed or canceled job

  • Data input port of the component supports separate access according to train/validate

  • Job list optimization:search, filter, and sort optimization

  • Significantly improve page fluency when the amount of data is large

  • Optimize the log pull method and improve efficiency

  • Enhance safety check

  • Interface optimization

  • Data decoupling:separate the relationship between presentation, interaction, and data of the page

  • Interaction logic decoupling:Disassembly and refinement of complex functional logic

  • Improve the cohesion and reusability of components and tools

Update README.md

Adjust the descriptions about the command of using fateboard.

FATE-Board V1.1 Web front end outcomes of the first round of proposals

  1. Format component parameters and truncate component decimal point;

  2. Secure Boost-tree graph;

  3. Secure Boost-the importance of features;

  4. Search and sort the job table page;

  5. Waiting task cancel;

  6. Workflow is differentiated by roles: guest, host or arbiter components that are not running by their own side are ashed on the workflow, and the name of the component is not displayed;

  7. Component shape, running state (success or failure), dynamic effect of running state change;

  8. Input and output ports of components: separate data from model input and output.

fateboard 1.5 unable to display diagram

I run on a standalone mode, i able to run jobs successfully. Log are showing up in the fateboard, but i can't see the dashboard diagrams. There is no error in browser console. Any idea why?

image

[BUG]Fateboard display predict_score wrong if data are in format of scientific notation.

I notice a small problem in the fateboard of FATE 1.5.2.version
When you click into the data output of homo nn component in the board,
you can see some predict-scores are more than 1.0,
For example:
If a predict scores is '8.940696716308..e-8', you will get '8.940696' only .
Here is the picture:
QQ图片20210826141219
Look like that it just cuts off the string according to the length.
I am not sure this problem still exists in the 1.6.x,
can someone have a check?

Upgrades for 1.2.0 version

  1. New components:
  • Heterogeneous feature correlation component: visualizing the correlation matrix  diagram,filtering by role or variable, and scaling of diagram
  • Upload component: uploading data, and supporting for viewing data output
  • Download component: downloading data, and supporting for viewing data output
  • LocalBaseline component: A LR component based on Sklearn run to compare the performance  of Federated LR
  1. Component optimization:
  • Secureboost: Visual enhancement of secureboost tree diagrams, and support for scaling  and dragging trees
  • Federated Sample: Filtering by label to display the table content
  • Model output of some components supports global fuzzy search
  • Model output of some components supports for table column sorting
  • Model output of some components supports multi-host visualization and filtering by role,  and adds variable mapping
  • Optimize cross-validation curves and loss curves  support visualization and interaction  of  large data volumes, and support for manual refresh during operation
  1. Update the resources for the front-end.

  2. Others:

  • Job detail page redesign, and support for workflow scaling and dragging, and optimize  parameter  display
  • Add job note and support for description and classification
  • Log load optimization

通过项目接入fate-board,ajax请求fate-board接口跨域

我们在自己的系统内调用fate-board接口,fate-board版本为1.5.0 浏览器报跨域错误。
跨域报错信息如下:
/job/query/page/new' from origin 'null' has been blocked by CORS policy: Response to preflight request doesn't pass access control check: No 'Access-Control-Allow-Origin' header is present on the requested resource.
请问这个如何解决呢,我看了下fate-board代码,也没找到相关的跨域配置呢。
谢谢

Adjust version

Change board version 1.2.1 to 1.3.0 to match the version for FATE.

[bug]: Part of the page cannot be opened when use HTTPS

Deploy FATE-Board on K8s, use ingress-nginx for TLS certificate termination.

When I open fateboard's URL, I got some error information.

ERROR: 'WebSocket': An insecure WebSocket connection may not be initiated from a page loaded over HTTPS.

image

add copy plugins to pom for a bit better deploy experience

fateboard could be deploy with following directory structure:

fateboard/
  fateboard.jar
  conf/
    application.properties
    ssh.properties

It maybe a good idea to generate this structure in maven package phase by simply add this configuration to pom.xml:

<plugins>
...
    <plugin>
        <groupId>org.apache.maven.plugins</groupId>
        <artifactId>maven-dependency-plugin</artifactId>
        <version>2.10</version>
        <executions>
            <execution>
                <id>copy-dependencies</id>
                <phase>package</phase>
                <goals>
                    <goal>copy-dependencies</goal>
                </goals>
                <configuration>
                    <outputDirectory>${project.build.directory}/lib</outputDirectory>
                </configuration>
            </execution>
            <!-- code added start -->
            <execution>
                <id>copy-packaged</id>
                <phase>package</phase>
                <goals>
                    <goal>copy</goal>
                </goals>
                <configuration>
                    <artifactItems>
                        <artifactItem>
                            <groupId>${project.groupId}</groupId>
                            <artifactId>${project.artifactId}</artifactId>
                            <version>${project.version}</version>
                            <type>${project.packaging}</type>
                            <destFileName>fateboard.jar</destFileName>
                        </artifactItem>
                    </artifactItems>
                    <outputDirectory>${project.build.directory}/fateboard</outputDirectory>
                </configuration>
            </execution>
        </executions>
    </plugin>
    <plugin>
        <artifactId>maven-resources-plugin</artifactId>
        <version>3.1.0</version>
        <executions>
            <execution>
                <id>copy-resources</id>
                <!-- here the phase you need -->
                <phase>package</phase>
                <goals>
                    <goal>copy-resources</goal>
                </goals>
                <configuration>
                    <outputDirectory>${basedir}/target/fateboard/conf</outputDirectory>
                    <resources>
                        <resource>
                            <directory>src/main/resources</directory>
                            <includes>
                                <include>application.properties</include>
                                <include>ssh.properties</include>
                            </includes>
                        </resource>
                    </resources>
                </configuration>
            </execution>
        </executions>
    </plugin>
    <!-- code added end -->
</plugins>

We get feteboard dir after package phase:

image

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.