Integration of PyTorch to Collective Knowledge workflow framework to provide unified CK JSON API for AI (customized builds across diverse libraries and hardware, unified AI API, collaborative experiments, performance optimization and model/data set tuning):
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
.cm
package
program
soft
.ckr.json
.gitignore
CHANGES
CONTRIBUTIONS
COPYRIGHT
LICENSE
README.md

README.md

Collective Knowledge repository for PyTorch

compatibility License

Introduction

This repository provides high-level, portable and customizable workflows for PyTorch as a part of our long-term community initiative to unify and automate AI using Collective Knowledge Framework (CK).

Coordination of development

Minimal CK installation

The minimal installation requires:

  • Python 2.7 or 3.3+ (limitation is mainly due to unitests)
  • Git command line client.

Linux/MacOS

You can install CK in your local user space as follows:

$ git clone http://github.com/ctuning/ck
$ export PATH=$PWD/ck/bin:$PATH
$ export PYTHONPATH=$PWD/ck:$PYTHONPATH

You can also install CK via PIP with sudo to avoid setting up environment variables yourself:

$ sudo pip install ck

Windows

We still need to provide proper support to build PyTorch via CK on Windows

First you need to download and install a few dependencies from the following sites:

You can then install CK as follows:

 $ pip install ck

or

 $ git clone https://github.com/ctuning/ck.git ck-master
 $ set PATH={CURRENT PATH}\ck-master\bin;%PATH%
 $ set PYTHONPATH={CURRENT PATH}\ck-master;%PYTHONPATH%

CK workflow installation for PyTorch

CPU

$ ck pull repo:ck-pytorch
$ ck install package --tags=lib,pytorch,vcpu

GPU

$ ck pull repo:ck-pytorch
$ ck install package --tags=lib,pytorch,vcuda

Checking classification example (and automatically installing available MXNet model(s) via CK)

$ ck install package --tags=lib,pytorch-vision
$ ck run program:pytorch
  • Select 'classify-squeezenet-1.1'
  • Select image to classify
  • Observe result

Next steps

We plan to add PyTorch to our ReQuEST tournament framework: http://cKnowledge.org/request

Feedback

Get in touch with CK-AI community here.