我在正确设置系统时遇到了一些麻烦。我的系统包括：
lspci正确检测到我的GPU：
06:00.0 VGA compatible controller: NVIDIA Corporation Device 1b06 (rev a1) (prog-if 00 [VGA controller])
Subsystem: ASUSTeK Computer Inc. Device 85ea
Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx-
Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
Latency: 0, Cache Line Size: 128 bytes
Interrupt: pin A routed to IRQ 18
Region 0: Memory at c4000000 (32-bit, non-prefetchable) [size=16M]
Region 1: Memory at a0000000 (64-bit, prefetchable) [size=256M]
Region 3: Memory at b0000000 (64-bit, prefetchable) [size=32M]
Region 5: I/O ports at 2000 [size=128]
Expansion ROM at c5000000 [disabled] [size=512K]
Capabilities: <access denied>
Kernel driver in use: nvidia
Kernel modules: nvidiafb, nouveau, nvidia_384_drm, nvidia_384
以及下面的命令，我为GPU，CUDA和cuDNN安装了驱动程序：
# Show thunderbolt port / authorize eGPU
$ cat /sys/bus/thunderbolt/devices/0-1/device_name
$ echo 1 | sudo tee -a /sys/bus/thunderbolt/devices/0-1/authorized
# eGPU - on Ubuntu 16.04 - nvidia-384
$ sudo ubuntu-drivers devices
$ sudo ubuntu-drivers autoinstall
$ sudo apt-get install nvidia-modprobe
# CUDA - Download CUDA from Nvidia - http://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1604/x86_64/
$ wget https://developer.nvidia.com/compute/cuda/9.0/Prod/local_installers/cuda_9.0.176_384.81_linux-run
$ chmod +x cuda_9.0.176_384.81_linux-run
$ ./cuda_9.0.176_384.81_linux-run --extract=$HOME
$ sudo ./cuda-linux.9.0.176-22781540.run
$ sudo ./cuda-samples.9.0.176-22781540-linux.run
$ sudo bash -c "echo /usr/local/cuda/lib64/ > /etc/ld.so.conf.d/cuda.conf"
$ sudo ldconfig
# add :/usr/local/cuda/bin (including the ":") at the end of the PATH="/blah:/blah/blah" string (inside the quotes).
$ sudo nano /etc/environments
# CUDA samples - Check installation
$ cd /usr/local/cuda-9.0/samples
$ sudo make
$ /usr/local/cuda/samples/bin/x86_64/linux/release/deviceQuery
# cuDNN
$ wget http://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1604/x86_64/libcudnn7_7.0.5.15-1+cuda9.0_amd64.deb
$ dpkg -i libcudnn7_7.0.5.15-1+cuda9.0_amd64.deb
# Authorization (Security risk!)
$ sudo nano /etc/udev/rules.d/99-local.rules
# Add
ACTION=="add", SUBSYSTEM=="thunderbolt", ATTR{authorized}=="0", ATTR{authorized}="1"
借助sudo prime-select nvidia，我可以在eGPU和Intel集成的GPU之间切换。注销+登录后，eGPU似乎可以工作（nvidia-smi）：
Fri Sep 21 09:25:18 2018
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 384.130 Driver Version: 384.130 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 GeForce GTX 108... Off | 00000000:06:00.0 Off | N/A |
| 0% 47C P0 84W / 275W | 214MiB / 11172MiB | 4% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| 0 6734 G /usr/lib/xorg/Xorg 138MiB |
| 0 7081 G kwin_x11 23MiB |
| 0 7084 G /usr/bin/krunner 2MiB |
| 0 7092 G /usr/bin/plasmashell 47MiB |
+-----------------------------------------------------------------------------+
CUDA示例也可以工作：
$ /usr/local/cuda/samples/bin/x86_64/linux/release/deviceQuery
/usr/local/cuda/samples/bin/x86_64/linux/release/deviceQuery Starting...
CUDA Device Query (Runtime API) version (CUDART static linking)
Detected 1 CUDA Capable device(s)
Device 0: "GeForce GTX 1080 Ti"
CUDA Driver Version / Runtime Version 9.0 / 9.0
CUDA Capability Major/Minor version number: 6.1
Total amount of global memory: 11172 MBytes (11715084288 bytes)
(28) Multiprocessors, (128) CUDA Cores/MP: 3584 CUDA Cores
GPU Max Clock rate: 1683 MHz (1.68 GHz)
Memory Clock rate: 5505 Mhz
Memory Bus Width: 352-bit
L2 Cache Size: 2883584 bytes
Maximum Texture Dimension Size (x,y,z) 1D=(131072), 2D=(131072, 65536), 3D=(16384, 16384, 16384)
Maximum Layered 1D Texture Size, (num) layers 1D=(32768), 2048 layers
Maximum Layered 2D Texture Size, (num) layers 2D=(32768, 32768), 2048 layers
Total amount of constant memory: 65536 bytes
Total amount of shared memory per block: 49152 bytes
Total number of registers available per block: 65536
Warp size: 32
Maximum number of threads per multiprocessor: 2048
Maximum number of threads per block: 1024
Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
Max dimension size of a grid size (x,y,z): (2147483647, 65535, 65535)
Maximum memory pitch: 2147483647 bytes
Texture alignment: 512 bytes
Concurrent copy and kernel execution: Yes with 2 copy engine(s)
Run time limit on kernels: Yes
Integrated GPU sharing Host Memory: No
Support host page-locked memory mapping: Yes
Alignment requirement for Surfaces: Yes
Device has ECC support: Disabled
Device supports Unified Addressing (UVA): Yes
Supports Cooperative Kernel Launch: Yes
Supports MultiDevice Co-op Kernel Launch: Yes
Device PCI Domain ID / Bus ID / location ID: 0 / 6 / 0
Compute Mode:
< Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >
deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 9.0, CUDA Runtime Version = 9.0, NumDevs = 1
Result = PASS
我们来解决这个问题：仅当我以root权限执行python时，以下代码段才有效。
$ python
Python 3.6.6 |Anaconda, Inc.| (default, Jun 28 2018, 17:14:51)
[GCC 7.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> torch.cuda.is_available()
True
>>> torch.Tensor([0]).cuda()
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
RuntimeError: CUDA error: unknown error
vs。
$ sudo ~/miniconda3/envs/PyTorch/bin/python
Python 3.6.6 |Anaconda, Inc.| (default, Jun 28 2018, 17:14:51)
[GCC 7.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> torch.cuda.is_available()
True
>>> torch.Tensor([0]).cuda()
tensor([0.], device='cuda:0')
TLDR： 我的CUDA安装似乎可以正常工作，并且我的eGPU已被系统正确识别。但是在PyTorch（以及Tensorflow）中，当我尝试将相同数据移至eGPU时，尽管框架已识别出该错误，但会抛出“未知错误”。当我以root权限执行相同的代码时，所有内容都像一个超级按钮。有人知道在哪里我必须采用一些权限，以便可以作为标准用户执行代码？
答案 0 :(得分：0)
试用了不同版本的nvidia驱动程序后，我得出的结论是384版存在问题。
我通过sudo apt-get purge nvidia.删除了所有与nvidia驱动程序相关的内容，并安装了390版：
$ sudo add-apt-repository ppa:graphics-drivers/ppa
$ sudo ubuntu-drivers devices
$ sudo apt install nvidia-390
结果如下：
(PyTorch) max@MaxKubuntuNUC:~$ python
Python 3.6.6 |Anaconda, Inc.| (default, Jun 28 2018, 17:14:51)
[GCC 7.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> torch.cuda.is_available()
True
>>> torch.tensor([1]).cuda()
tensor([1], device='cuda:0')
现在，所有内容都像魅力一样。