TensorFlow:调用 cuInit 失败:CUDA_ERROR_NO_DEVICE [英] TensorFlow : failed call to cuInit: CUDA_ERROR_NO_DEVICE

问题描述

我的测试:

import tensorflow as tf
hello = tf.constant('Hello, TensorFlow!')
sess = tf.Session()`

错误:

c:\l\work\tensorflow-1.1.0\tensorflow\stream_executor\cuda\cuda_driver.cc:405]

c:\l\work\tensorflow-1.1.0\tensorflow\stream_executor\cuda\cuda_driver.cc:405]

调用 cuInit 失败:CUDA_ERROR_NO_DEVICE

failed call to cuInit: CUDA_ERROR_NO_DEVICE

-> 但 "/cpu:0" 工作正常

-> but "/cpu:0" works fine

配置:

nvidia-smi :

nvidia-smi :

• CUDA 9.1 版
• tensorflow-1.1.0
• Windows 10
• cudnn64_7.dll(安装在 C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.1\bin)
• 我的 Conda 环境中仅安装了 tensorflow-gpu

为什么 Tensorflow 检测不到我的 GPU?

Why can't Tensorflow detect my GPU ?

推荐答案

问题已解决 on GitHub.如果您为 CUDA_VISIBLE_DEVICES 环境变量设置了无效值，则会显示此错误消息，例如当您只有一个 GPU(ID 0)并设置 CUDA_VISIBLE_DEVICES=1 或 CUDA_VISIBLE_DEVICES=2 时.

The issue was solved on GitHub. This error message will be shown if you set an invalid value for the CUDA_VISIBLE_DEVICES environment variable, e.g. when you only have a single GPU (which has ID 0) and set CUDA_VISIBLE_DEVICES=1 or CUDA_VISIBLE_DEVICES=2.

这篇关于TensorFlow:调用 cuInit 失败:CUDA_ERROR_NO_DEVICE的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！