Мой tenorflow-gpu не может создать сеанс с gpu.Я установил nvidia-драйвер, CUDA, CUDNN в соответствии с версией GPU.Тензор Flow-GPU был установлен с Anaconda3.6.6.И версия tenorflow-gpu 1.5.0, потому что кажется, что машина не поддерживает AVX2.
%%%% Environment %%%%
CentOS7 KVM
GPU: Tesla P100
Driver: Nvidia 384.145
CUDA-9.0
CUDNN-7.2.1
tensorflow-gpu: 1.5.0
python: Anaconda 3.6.6
%%%% nvidia-smi %%%%
скриншот nvidia-smi здесь
%%%% driver information %%%%
(tf-gpu) [root@gpu gpus]# cat /proc/driver/nvidia/gpus/0000:00:04.0/information
Model: Tesla P100-PCIE-16GB
IRQ: 27
GPU UUID: GPU-????????-????-????-????-????????????
Video BIOS: ??.??.??.??.??
Bus Type: PCIe
DMA Size: 47 bits
DMA Mask: 0x7fffffffffff
Bus Location: 0000:00:04.0
Device Minor: 0
%%%% tf.test.is_gpu_available %%%%
(tf-gpu) [root@gpu ~]# python3
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 tensorflow as tf
>>> tf.test.is_gpu_available()
2018-09-28 00:29:26.255612: E tensorflow/stream_executor/cuda/cuda_driver.cc:406] failed call to cuInit: CU
DA_ERROR_NO_DEVICE
2018-09-28 00:29:26.255679: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:158] retrieving CUDA diag
nostic information for host: gpu2018-09-28 00:29:26.255695: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:165] hostname: gpu
2018-09-28 00:29:26.255761: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:189] libcuda reported ver
sion is: 384.145.0
2018-09-28 00:29:26.255796: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:369] driver version file
contents: """NVRM version: NVIDIA UNIX x86_64 Kernel Module 384.145 Thu May 17 21:47:37 PDT 2018GCC version: gcc version 4.8.5 20150623 (Red Hat 4.8.5-28) (GCC)
"""
2018-09-28 00:29:26.255833: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:193] kernel reported vers
ion is: 384.145.0
2018-09-28 00:29:26.255844: I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:300] kernel version seems
to match DSO: 384.145.0
False
%%%% tensorflow.Session() %%%%
выход для создания сеанса здесь