Jeremy Reizenstein e0753f0b0d Build wheels for s3
Summary: For Linux, instead of uploading wheels to PyPI which will only work with one particular version of PyTorch and CUDA, from the next release we will store a range of built wheels on S3.

Reviewed By: nikhilaravi

Differential Revision: D26209398

fbshipit-source-id: 945a6907b78807e1eedb25007f87f90bbf59f80e
2021-02-05 05:52:36 -08:00
..
2021-02-05 05:52:36 -08:00
2021-02-05 05:52:36 -08:00
2021-02-05 05:52:36 -08:00
2021-02-05 05:52:36 -08:00
2021-02-05 05:52:36 -08:00

Building Linux pip Packages

  1. Make sure this directory is on a filesystem which docker can use - e.g. not NFS. If you are using a local hard drive there is nothing to do here.

  2. You may want to docker pull pytorch/conda-cuda:latest.

  3. Run bash go.sh in this directory. This takes ages and writes packages to inside/output.

  4. You can upload the packages to s3, along with basic html files which enable them to be used, with bash after.sh.

In particular, if you are in a jupyter/colab notebook you can then install using these wheels with the following series of commands.

import sys
import torch
version_str="".join([
    f"py3{sys.version_info.minor}_cu",
    torch.version.cuda.replace(".",""),
    f"_pyt{torch.__version__[0:5:2]}"
])
!pip install pytorch3d -f https://dl.fbaipublicfiles.com/pytorch3d/packaging/wheels/{version_str}/download.html