目的は、DeepLearning (cifar10_train.py 動きました)
環境
- MacBook Pro (15-inch, 2016)
- macOS High Sierra 10.13.6（17G65）
- GeForce GTX 1080
私が購入したものは、下記です。
GIGABYTE ビデオカード GTX1080搭載 外付VGA BOX GV-N1080IXEB-8GD
注意
検証は、まだ十分に行なっておりません。
eGPU が認識でき、外部モニターに画面出力できている状態です。
macOS High Sierra 10.13.6（17G3015）で、認識しなくなりました。
（一旦、TimeMachineで戻しました。今の所、17G65 でしか動きません。）
手順
基本的には、下記のスクリプトを使用させていただきました。
https://github.com/learex/macOS-eGPU/tree/master#macos-egpush
- まずは、バックアップ。(TimeMachineを使用して、現状保存が良いとおもいます。)
- SIPの無効化。再起動時に command + R で起動させ、
ユーティリティメニューのターミナルを選択し、下記を実行する。
csrutil disable; reboot
- スクリプト実行
bash <(curl -s https://raw.githubusercontent.com/learex/macOS-eGPU/master/macOS-eGPU.sh)
- スクリプト実行時のログは、以下です。
- ２度目の実行なので、ライセンスアグリーメントは、省略。及び、スクリプトの実行のみ。
- 途中、eGPU の抜き差しの指示があるので、行なってください。
$ macos-egpu
macOS-eGPU.sh (v1.5)
Accept license terms... [done]
Killing all other running programs... [OK]
Internet connection established... [YES]
Fetching system information...
macOS info [done]
system integrity protection [done]
thunderbolt version [done]
GPU information [done]
installed eGPU software [done]
installed patches [done]
installed programs [done]
Setting internal switches...
Automatic eGPU information fetching...
locking script execution [done]
elevating privileges
Password:
checking for elevated privileges [OK]
preparing secure eGPU connection [done]
waiting 20 seconds for user to connect eGPU
20..19..18..17..16..15..14..13..12..11..10..9..8..7..6..5..4..3..2..1..0
fetching eGPU information [done]
preparing secure eGPU disconnection [done]
waiting 20 seconds for user to disconnect eGPU
20..19..18..17..16..15..14..13..12..11..10..9..8..7..6..5..4..3..2..1..0
stetting switches [done]
opening script execution lock [done]
Fetching CUDA needs...
fetching CUDA requiring apps list [done]
preparing matching [done]
matching [done]
Checking for incompatibilies and up to date software...
NVIDIA drivers [install scheduled]
NVIDIA eGPU enabler [skip, incompatible]
AMD legacy drivers [skip]
T82 unblocker [skip]
NVIDIA dGPU deactivator [skip]
macOS 10.13.4/.5 NVIDIA patch [skip, incompatible]
macOS 10.13.4+ thunderbolt 1/2 unlock [skip, incompatible]
CUDA software
CUDA drivers [skip]
CUDA developer driver [skip]
CUDA toolkit [skip]
CUDA samples [skip]
thunderbolt daemon [skip]
IO PCIE Tunnelled patch [install scheduled]
Checking if SIP is sufficently disabled... [OK]
Download external content...
--- NVIDIA drivers ---
######################################################################## 100.0%
Checking for elevated privileges...
Uninstalling...
Installing...
NVIDIA driver [done]
IO PCIE Tunnelled patch [done]
Patching...
deactivating auto-updates...
Finish...
cleaning system [done]
Rebuilding caches
kext cache [done]
system cache [done]
A reboot will soon be performed...
5..4..3..2..1..0
reboot: / is busy updating, waiting for lock (this might take approx 15-30s)...
reboot: / is busy updating; waiting for lock
結果
おまけ
tensorflow r1.10 を最新CUDAの仕様でコンパイルに成功しました。
python 3.6.6
CUDA Toolkit 10.0
cuDNN v7.3.1 (Sept 28, 2018), for CUDA 10.0
TF_NCCL_VERSION 2.3.5
作成したパッケージです。
tensorflow-1.10.1-cp36-cp36m-macosx_10_13_x86_64.whl
CPU 版に比べて、６倍くらいの処理スピードです。（こんなもんなのですかね？）
python cifar10_train.py
Filling queue with 20000 CIFAR images before starting to train. This will take a few minutes.
2018-10-05 23:04:13.815498: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:858] OS X does not support NUMA - returning NUMA node zero
2018-10-05 23:04:13.815687: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1405] Found device 0 with properties:
name: GeForce GTX 1080 major: 6 minor: 1 memoryClockRate(GHz): 1.7335
pciBusID: 0000:46:00.0
totalMemory: 8.00GiB freeMemory: 5.71GiB
2018-10-05 23:04:13.815706: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1484] Adding visible gpu devices: 0
2018-10-05 23:04:14.187784: I tensorflow/core/common_runtime/gpu/gpu_device.cc:965] Device interconnect StreamExecutor with strength 1 edge matrix:
2018-10-05 23:04:14.187826: I tensorflow/core/common_runtime/gpu/gpu_device.cc:971] 0
2018-10-05 23:04:14.187831: I tensorflow/core/common_runtime/gpu/gpu_device.cc:984] 0: N
2018-10-05 23:04:14.188000: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1097] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 5481 MB memory) -> physical GPU (device: 0, name: GeForce GTX 1080, pci bus id: 0000:46:00.0, compute capability: 6.1)
2018-10-05 23:04:17.923348: step 0, loss = 4.68 (139.6 examples/sec; 0.917 sec/batch)
2018-10-05 23:04:18.492055: step 10, loss = 4.60 (2250.7 examples/sec; 0.057 sec/batch)
2018-10-05 23:04:18.914545: step 20, loss = 4.62 (3029.7 examples/sec; 0.042 sec/batch)
2018-10-05 23:04:19.332597: step 30, loss = 4.60 (3061.8 examples/sec; 0.042 sec/batch)
2018-10-05 23:04:19.756255: step 40, loss = 4.38 (3021.3 examples/sec; 0.042 sec/batch)
2018-10-05 23:04:20.198415: step 50, loss = 4.42 (2894.9 examples/sec; 0.044 sec/batch)
2018-10-05 23:04:20.622042: step 60, loss = 4.40 (3021.5 examples/sec; 0.042 sec/batch)
2018-10-05 23:04:21.040028: step 70, loss = 4.21 (3062.3 examples/sec; 0.042 sec/batch)
2018-10-05 23:04:21.449060: step 80, loss = 4.19 (3129.3 examples/sec; 0.041 sec/batch)
2018-10-05 23:04:21.868983: step 90, loss = 4.12 (3048.2 examples/sec; 0.042 sec/batch)
2018-10-05 23:04:22.424941: step 100, loss = 4.13 (2302.3 examples/sec; 0.056 sec/batch)