cu11_config.sh 860 B

1234567891011
  1. export OMPI4_CONTAINER=ompi4-config.sif
  2. export CUDA_BASE_IMAGE="nvidia/cuda:11.4.2-cudnn8-devel-ubuntu18.04"
  3. export SINGULARITYENV_VERBS=without-verbs
  4. export SINGULARITYENV_TENSORFLOW_VERSION=2.7.0
  5. export SINGULARITYENV_PYTORCH_VERSION=1.10.0+cu113
  6. export SINGULARITYENV_PIP_CMD_1="pip install torch==${SINGULARITYENV_PYTORCH_VERSION} torchvision==0.11.1+cu113 torchaudio==0.10.0+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html"
  7. export SINGULARITYENV_PIP_CMD_2="pip install mxnet-cu112 tensorflow-gpu==${SINGULARITYENV_TENSORFLOW_VERSION} keras h5py filelock matplotlib scikit-learn"
  8. export SINGULARITYENV_LDCONFIG_PATH="/usr/local/cuda-11.4/targets/x86_64-linux/lib/stubs"
  9. export SINGULARITYENV_PIP_CMD_3="HOROVOD_GPU_ALLREDUCE=NCCL HOROVOD_WITH_MXNET=1 HOROVOD_WITH_TENSORFLOW=1 HOROVOD_WITH_PYTORCH=1 pip install --no-cache-dir horovod "