candle-distributed

Introduction

This is a distributed deep learning framework based on candle

TODO

• multi worker detail: prefetch
• test of multi worker dataloader. use this dataloder to train a model
• distributed dataset, dataloader and sampler like torch
• [] things related to GPU
• [] parameter server
• [] ring all-reduce

ROADMAP

This project is still on very early stage. Since I was major in Computer Vision, my plan is to implement a multi gpu training example of Object Detection model like YOLO series.