python - 未找到：ID 为 0 的 TF GPU 设备未注册，之后出现段错误

Question

你能帮我调试一下这个问题吗？我尝试使用多个版本进行构建，但无法解决。

我的配置：

硬件： MacBook Pro 13,3 eGPU NVIDIA 1080

软件：
macOS 10.13.6
NVIDIA Web 驱动程序 387.10.10.10.40.105
CUDA 驱动程序 396.148
CUDA 9.1 工具包
cuDNN 7.0.5
Python 2.7
NCCL 2.1.15
Xcode 9.2

>>> import tensorflow as tf
>>> tf.test.is_gpu_available()
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:859] OS X does not support NUMA - returning NUMA node zero
tensorflow/core/common_runtime/gpu/gpu_device.cc:1356] 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: 2.32GiB
tensorflow/core/common_runtime/gpu/gpu_device.cc:1435] Adding visible gpu devices: 0
tensorflow/core/common_runtime/gpu/gpu_device.cc:923] Device interconnect StreamExecutor with strength 1 edge matrix:
tensorflow/core/common_runtime/gpu/gpu_device.cc:929]      0 
tensorflow/core/common_runtime/gpu/gpu_device.cc:942] 0:   N 
tensorflow/core/common_runtime/gpu/gpu_device.cc:1053] Created TensorFlow device (/device:GPU:0 with 2025 MB memory) -> physical GPU (device: 0, name: GeForce GTX 1080, pci bus id: 0000:46:00.0, compute capability: 6.1)
True

当我尝试运行某些东西时，最后会收到带有段错误的错误消息：

Filling queue with 20000 CIFAR images before starting to train. This will take a few minutes.
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:859] OS X does not support NUMA - returning NUMA node zero
tensorflow/core/common_runtime/gpu/gpu_device.cc:1356] 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: 3.39GiB
tensorflow/core/common_runtime/gpu/gpu_device.cc:1435] Adding visible gpu devices: 0
tensorflow/core/common_runtime/gpu/gpu_device.cc:923] Device interconnect StreamExecutor with strength 1 edge matrix:
tensorflow/core/common_runtime/gpu/gpu_device.cc:929]      0 
tensorflow/core/common_runtime/gpu/gpu_device.cc:942] 0:   N 
tensorflow/core/common_runtime/gpu/gpu_device.cc:1053] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 3118 MB memory) -> physical GPU (device: 0, name: GeForce GTX 1080, pci bus id: 0000:46:00.0, compute capability: 6.1)
E tensorflow/core/grappler/clusters/utils.cc:127] Not found: TF GPU device with id 0 was not registered
Segmentation fault: 11

在其他程序中，我尝试减少 per_process_gpu_memory_fraction 和批量大小，因此它在第一批后崩溃并出现相同的错误代码。

标签： pythontensorflow