-
Notifications
You must be signed in to change notification settings - Fork 40
Installing MaTEx TensorFlow CPU
Vinay Amatya edited this page Jul 13, 2018
·
16 revisions
We have provided a set of scripts for easy installation on CPU clusters. We expect that minimal changes would be required to the script for installation.
For using CPU enabled TensorFlow, you would need to set up the MPI_HOME environment variable to point where MPI and should be found. The installation script assumes that the host system has Anaconda flavor of python available. For the same, ANACONDA_HOME environment variable needs to be set, pointing to where ANACONDA has been installed.
As an example, for bash shells:
$ export ANACONDA_HOME=/where/ANACONDA/installation/resides
$ export MPI_HOME=/where/mpi_library/resides
Installation for bash shells
$ cd matex/src/deeplearning/tensorflow/cpu/py3.x
$ source ./install_tf_matex_0.7.0.sh
Installation for C-shells
$ cd matex/src/deeplearning/tensorflow/cpu/py3.x
$ source ./install_tf_matex_0.7.0.sh
Deactivating the virtual environment:
$ source deactivate_mtx.sh
Activating the virtual environment (if MaTEx has already been installed):
$ source activate_mtx.sh
Getting Started on MaTEx-TensorFlow
- Required Software
- Installing MaTEx-TensorFlow on CPU Clusters
- Installing MaTEx-TensorFlow on GPU Clusters
- MaTEx-TensorFlow on Older glibc(v<2.19)
- DataSet Reader
- Testing Scripts
- Performance
- Running on PNNL Systems
- Running on NERSC Systems
- Restarting the MaTEx TensorFlow environment