Building conda binaries

• Change the BUILD_VERSION and BUILD_NUMBER variables in build_pytorch.sh as appropriate
• If pytorch-$BUILD_VERSION folder doesn't exist, copy over the last version and change the meta.yaml if necessary (if tests change etc.)
  • cp -r pytorch-0.1.3 pytorch-$BUILD_VERSION
  • git add pytorch-$BUILD_VERSION
• Run ./build_pytorch.sh on an OSX machine and a Linux machine

build base docker image

cd ..
nvidia-docker build -t soumith/conda-cuda -f conda/Dockerfile .
docker push soumith/conda-cuda

building pytorch / torchvision etc.

nvidia-docker run -it --ipc=host --rm -v $(pwd):/remote soumith/conda-cuda bash
cd remote/conda

# versioned
export PYTORCH_FINAL_PACKAGE_DIR="/remote"
export TORCH_CONDA_BUILD_FOLDER=pytorch-1.1.0
export PYTORCH_REPO=pytorch
export PYTORCH_BRANCH=v1.1.0
./build_pytorch.sh 100 1.1.0 1 # cuda 10.0 pytorch 1.0.1 build_number 1

# nightly
export PYTORCH_FINAL_PACKAGE_DIR="/remote"
export TORCH_CONDA_BUILD_FOLDER=pytorch-nightly
export PYTORCH_REPO=pytorch
export PYTORCH_BRANCH=master
./build_pytorch.sh 100 nightly 1 # cuda 10.0 pytorch 1.0.1 build_number 1

building magma-cuda91

nvidia-docker run -it --ipc=host --rm -v $(pwd):/remote soumith/conda-cuda bash
git clone https://github.com/pytorch/builder
cd builder/conda
conda install -y conda-build
. ./switch_cuda_version.sh 10.0
conda build magma-cuda100-2.5.1

Test magma builds on K80 machine, to check against the bug pytorch/pytorch#29096