TyperClub / tinyframework

Using pure numpy to construct deep learning computational graph framework

tinyframework

Introduction

Using pure numpy to construct deep learning computational graph framework.

note：To view the code, please switch to the master branch

Dev Tools：

• pycharm

Dependencies

• python: >= 3.7

Features

• Based on computational graph, can be used to build common machine learning models.
• Support automatic gradient.
• Support common optimization methods (such as gd, momentum, adagrad, rmsprop, Adam, etc.)
• Support common evaluation methods (such as accuracy, precision, auc, f1_score, etc.)
• Support model save and load
• Support drawing calculate graph by pyecharts
• Support model serving by grpc
• Support model export
• Support distribute trainning.

logs：

To be added