Skip to content

README.md

Latest commit

 

History

History
130 lines (73 loc) · 2.24 KB

README.md

File metadata and controls

130 lines (73 loc) · 2.24 KB

GitML - Git for Data Science Projects

GitML helps you manage and move around between iterations during model building.

How gitml can help you?

We provide a git like command line tool for you to take a snapshot of your current state with all necessary information and dependent files.

How to use it?

Installation

pip install gitml

Usage

Initializing a gitml project.

It will also initialize a new git repository, if one doesn't exist already.

gitml init

Building your model with gitml

Continue with your model building. Say you are starting with model.py. Add gitml package to your code, like below.

# model.py

import gitml

## your code goes here ##

## Add the below lines.

with gitml.state() as state:
	state.set( model=MODEL_OBJECT, params = { "PARAM_KEY_1" : "PARAM_VALUE_1" }, metrics = { "METRICS_KEY_1" : "METRICS_KEY_2" }, 
	remarks = "REMARKS" )

Note : Now run you can run model.py as usual with python model.py.

Save your current state by passing a command line argument save to your model.py.

python model.py save

This takes a snapshot of all your files and along with other information. It will save the iteration with a unique id (Iteration Id).

# You will be seeing this line when you run `python model.py save`.

[GitML] iteration id : <ITERATION_ID>

List all saved iterations.

gitml ls

Choose and commit a saved iteration.

This will add a permenant commit to your git. For the iteration id, try gitml ls.

gitml commit <ITERATION_ID>

List all commited iterations.

gitml commit ls

Reusing a commited iteration.

gitml reuse <COMMITED_ITERATION_ID>

Note : COMMITED_ITERATION_ID is the iteration id from gitml commit ls.

Stash the changes on your current workspace.

gitml stash

Restore the stash.

gitml restore

Loading a saved/commited iteration on your code for serving.

import gitml

## For saved iterations `gitml ls` and commited iterations `gitml commit ls`.

model = gitml.load("<ITERATION_ID>")

## Your code.

Happy Model Building :)

Development - For contributors

  1. Fork the repo.
  2. Make changes and install using ./tasks/clean_install.sh.