Michael Green MichaelScript

Tensorflow 1.7 with CUDA on macOS High Sierra 10.13.4 for eGPU

Largely based on the Tensorflow 1.6 gist, and Tensorflow 1.7 gist for xcode, this should hopefully simplify things a bit.

Requirements

NVIDIA Web-Drivers 387.10.10.10.30.103 for 10.13.4
CUDA-Drivers 387.178
CUDA 9.1 Toolkit

Tensorflow

This is fully functional version of Tensorflow with GPU on macOS 10.13 it also works on 10.13.1 I hope it helps.

System information

OS - High Sierra 10.13
Tensorflow - 1.4
Xcode command line tools - 8.2 (Download from here: Xcode - Support - Apple Developer & Switch to different clang version: sudo xcode-select --switch/Library/Developer/CommandLineTools & check version: clang -v)
Cmake - 3.7

Tensorflow

System information

OS - High Sierra 10.13
Tensorflow - 1.4
Xcode command line tools - 8.2 (Download from here: Xcode - Support - Apple Developer & Switch to different clang version: sudo xcode-select --switch/Library/Developer/CommandLineTools & check version: clang -v)
Cmake - 3.7
Bazel - 0.7.0

This configuration worked for me, hope it helps

It is based on: https://becominghuman.ai/deep-learning-gaming-build-with-nvidia-titan-xp-and-macbook-pro-with-thunderbolt2-5ceee7167f8b

and on: https://stackoverflow.com/questions/44744737/tensorflow-mac-os-gpu-support

Hardware

Debian 8 Hardening

Update root's mail recipient. Open /etc/aliases replacing [email protected] with an administrator's email address. This is where logs will be emailed.
```
root:     [email protected]
```
Update the the default umask to 027. Edit the file /etc/init.d/rc and change the following setting:

umask 027

Latency numbers every programmer should know

L1 cache reference ......................... 0.5 ns
Branch mispredict ............................ 5 ns
L2 cache reference ........................... 7 ns
Mutex lock/unlock ........................... 25 ns
Main memory reference ...................... 100 ns             
Compress 1K bytes with Zippy ............. 3,000 ns  =   3 µs
Send 2K bytes over 1 Gbps network ....... 20,000 ns  =  20 µs
SSD random read ........................ 150,000 ns  = 150 µs

Read 1 MB sequentially from memory ..... 250,000 ns = 250 µs

	import site
	import os
	import pathlib
	# Big thanks to Match on StackOverflow for helping me with this
	# see https://stackoverflow.com/a/48713998/5614280

	# This is some cool hackery to overwrite the default functionality of
	# the builtin print function within you're entire python environment
	# to display the file name and the line number as well as always flush
	# the output. It works by creating a custom user script and placing it

	# see https://www.topbug.net/blog/2013/04/14/install-and-use-gnu-command-line-tools-in-mac-os-x/

	# core
	brew install coreutils

	# key commands
	brew install binutils
	brew install diffutils
	brew install ed --default-names
	brew install findutils --with-default-names