mlvajra 0.1.4.3

A Last weapon to save Data scientist

mlVajra

A framework or best practices to develop end to end machine learning pipeline (also has some tips for ML-management people ) Aim : To built robust depoyment pipeline strategies using Open source stack planning to add as many strategies in this repo pertaining to ML-deployment

Installation :

pip install mlvajra - only installs mlvajra binaries

To install complete dependencies: (for time being)

git clone https://github.com/rajagurunath/mlvajra.git

create virtualenv

virtualenv -p python3 vajra_env

source vajra_env\bin\activate

cd mlvajra
Install all required dependencies from repo

pip install -r requirements.txt

TODO list

Deploy

• Mlflow
• Tensorflow serving

model-Training /distribuited

• mlflow -generic classification metrics (done)
• nnictl-automl -tensorflow /pytorch

Feature Engineering

• pandas
• pyspark-Flint

preprocessing

• cyclic features (done)
• lag features
• window features