SrivastavaKshitij

#!/usr/bin/env python
# coding: utf-8

# # [TVMConf 2020] BYOC Tutorial Demo
# 
# Cody Yu (hyuz@amazon.com), Zhi Chen (chzhi@amazon.com) from AWS AI.
# 
# 
# 
# This demo has two parts. In the first part, we use a simple Relay graph to walkthrough the BYOC workflow. In the second part, we build a SSD model with TensorRT BYOC integration to showcase a real world application.

mingfeima

PyTorch Channels Last Memory Format Performance Optimization on CPU Path

("mkldnn" has been renamed to "oneDNN", but exsiting PyTorch APIs still use "mkldnn", future work will align PyTorch user level APIs to "oneDNN")

Table of Contents

• PyTorch Channels Last memory format introduction
• oneDNN API for NHWC layout
• Generic Channels Last memory format optimization with ATen native
• oneDNN NHWC integration

NB: Memory format refers to data representation that describes how multidimensional arrays (nD) are stored in linear (1D) memory address space. Memory format has the same semantic with layout in oneDNN. Layout in PyTorch has other semantic ofdescribing dense or sparse with the attributes: 'torch.strided', 'torch.sparse_coo'.

asroy

I've tested it on Fedora 23 and Ubuntu 16.04. I'm using gcc-5.3.1, python-3.4, VS Code-1.14.0
You can debug mixed Python/C++ in the same GUI. It also works for MPI applications. You can switch between the debuggers and corresponding call stacks.

1. Packages needed
  1) Visual Studio Code

  2) Extensions for VS Code:
      "Python" from Don Jayamanne (I'm using 0.6.7)
        This allows VS Code act as the front end to debug python.
        This gives VS Code ability to attach to a python script that uses module "ptvsd".