Sign inSign up

maxscha/cuda_pytorch_jupyter

By maxscha

•Updated about 5 years ago

Image
3

2.7K

maxscha/cuda_pytorch_jupyter repository overview

⁠ppc64le-pytorch-docker

This repository collects various docker images to cross-build working docker images of multiple version combinations for Cuda, Cudnn, Python, & Pytorch for IBM Power9 computers with Tesla V100-SXM2-32GB GPUs.

⁠Requirements
  • Docker
  • Docker Buildkit⁠
  • For building pytorch with buildkit more than 16GB of RAM is required
⁠How to build all version combinations
  1. cd build
  2. Adapt the required versions in build_all_versions.sh to your need
  3. Run ./build_all_versions.sh
  4. Depending on the number of images & combinations build this process
⁠Working & Tested Combinations

In Progress

⁠Github

Dockerfiles & build-scripts can be found on Github

Tag summary

Content type

Image

Digest

Size

9.4 GB

Last updated

about 5 years ago

docker pull maxscha/cuda_pytorch_jupyter:10.2-cudnn8-devel-centos7-py3.8.1-small-dependencies-gcc8.4.0-pytorchv1.8.1