为什么我获得关于正在运行哪个版本的CUDA的不一致信息

时间:2019-07-19 19:33:53

标签: tensorflow cuda

当我检查nvcc --version时,我看到我正在运行CUDA 9.2

$ nvcc --version
  nvcc: NVIDIA (R) Cuda compiler driver
  Copyright (c) 2005-2018 NVIDIA Corporation
  Built on Wed_Apr_11_23:16:29_CDT_2018
  Cuda compilation tools, release 9.2, V9.2.88

但是,当我运行Tensorflow程序时,我(相信)是几个CUDA 10库:

2019-07-19 14:54:04.173534: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudart.so.10.1
2019-07-19 14:54:04.205212: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcublas.so.10
2019-07-19 14:54:04.221681: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcufft.so.10
2019-07-19 14:54:04.227366: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcurand.so.10
2019-07-19 14:54:04.265624: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusolver.so.10
2019-07-19 14:54:04.273467: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcusparse.so.10
2019-07-19 14:54:04.366697: I tensorflow/stream_executor/platform/default/dso_loader.cc:42] Successfully opened dynamic library libcudnn.so.7

我使用的驱动程序是396.57,我认为它足以应付CUDA 9.2(请参阅https://docs.nvidia.com/deploy/cuda-compatibility/index.html#source-compatibility)。但是,当我运行Tensorflow程序时,它会崩溃,并抱怨它无法(以某种方式)使用GPU。经过大量谷歌搜索后,我发现一个建议,可能是我的驱动程序对于CUDA版本而言过旧。由于我看到了CUDA 10种语言的库,因此我升级到418.56,现在一切正常。

那么,我正在运行CUDA 9.2还是CUDA 10?如果仅运行9.2,为什么升级我的驱动程序可以解决我的问题,又如何获得*so.10库,我将其解释为属于CUDA 10?如果我正在运行CUDA 10,为什么nvcc --version告诉我我正在运行9.2。如果我每个都做一点,如何使自己完全处于1个或另一个上?

0 个答案:

没有答案