Linux workstation setup with Tensorflow/Keras/PyTorch and NVIDIA GPU.
RHEL 6.8 (with default gcc 4.4.7)
CUDA 8.0
cuDNN 6.0
python 3.6.3 (Anaconda)
gcc 4.8.4 (built from source)
bazel 0.5.4 (built from source)
tensorflow-gpu r1.3 (built from source)
Ubuntu 16.04.3 LTS
CUDA 8.0
cuDNN 6.0
python 3.5.2 (default with Ubuntu)
gcc 5.4.0 (default with Ubuntu)
bazel 0.5.4 (installed from binary)
tensorflow-gpu r1.4 (pip installed from binary)
keras 2.1.1 (pip installed from binary)
pytorch 0.2.0 (pip installed from binary)
torchvision 0.1.9 (pip installed from binary)
jupyter 4.4.0 (pip installed from binary)
Custom TF for Ubuntu Linux README
Ubuntu 16.04.3 LTS
python 3.6
gcc 5.4.0 (default with Ubuntu)
bazel 0.15.0 (installed from binary)
tensorflow-gpu r1.12 (custom build from source)