Skip to content

Latest commit

 

History

History
153 lines (102 loc) · 4.63 KB

structure.md

File metadata and controls

153 lines (102 loc) · 4.63 KB

Building ML applications with Gramex

Objective

We'll build a ML Classification application with Gramex.

Outline

The workshop content is prepared for 1.5 hrs.

Section Content Duration
Introduction Gramex intro, controlling data with URL params 20 mins
Build an ML app Use snippets and build the application step-by-step 50 mins
Q&A, Support Q&A session, help the participants 10 mins
Possibilities Other ML applications, Email alerts, Charts, SMS, Screenshots, Admin module 10 mins

Assumptions

  • you've installed Gramex correctly. If you've noticed any errors please share now.
  • you're familiar with ML and would like to build web apps easily.

What we will do today

  • construct a simple app with building blocks (HTML, JS, Python, Gramex, little Tornado).
  • create an app that trains, tests datasets and review the output.
  • we won't be reviewing ML algorithms or scikit-learn.

This is how the application would look at the end of it:

final page

What it needs

  1. Make sure you install the necessary software before you arrive at the workshop.

  2. Clone this repository, stick with master branch.

  3. Use Visual Studio Code, add the cloned directory as a workspace.

  4. Optionally, you can add our VSCode extension Gramex Snippets. Include the below lines in settings.json in VSCode for autosuggestions to work.

{
  "editor.tabCompletion": "on",
  "editor.quickSuggestions": {
    "other": true,
    "comments": false,
    "strings": true
  }
}

Introduction

Gramex is a FOSS web and data server. One can do the below with minimal setup

  • fetch, edit, delete data records
  • execute arbitrary python
  • auth support: database-based, active directory-based, google, social (twitter, facebook)
  • capture screenshots
  • configure smart email alerts and a lot more...

most of it via YAML configuration.

URLs

Consider a dataset that you'd like to explore for exploratory data analysis, we'll use the model iris dataset.

Before that you'd like to see the composition of the file. You can explore it using FormHandler:

url:
  data-endpoint:
    pattern: /$YAMLURL/iris
    handler: FormHandler
    kwargs:
      url: $YAMLPATH/datasets/iris.csv

Most handlers support kwargs section to control data from server-side.

Notes

  • We've limited the number of rows to 5 irrespective of how many rows end user requests.
  • $YAMLPATH and $YAMLURL are Gramex predefined variables. Read more in the documentation.
formats: JSON, CSV, HTML

URL sections

Multiple URL sections can be defined for different endpoints.

Arbitrary python

Arbitrary python code can be run using FunctionHandler.

url:
  custom-python-code:
    pattern: /$YAMLURL/code
    handler: FunctionHandler
    kwargs:
      function: app.my_func

  1. Access URL parameters via attributes

  2. Return values

Other Handlers, services

Gramex is built on top of tornado which uses RequestHandler for everything and we've sub-classed to different utilities for ease of use.

Snippets

Head to snippets directory to begin the application building.

Possibilities

Overview of file structure

Files for building the application:

  • gramex.yaml - route configuration for endpoints
  • index.html - page that serves the application content
  • template-navbar.html - navbar content
  • style.css - CSS rules
  • report.html - classification output renderer

Files for the workshop:

  • structure.md - workshop introduction
  • snippets - code snippets to build application
  • final.html - compare your final application against this

Contact