TensorFlow GPU 无法识别 GPU dll

Question

我按照以下网站安装了必要的 GPU 对 Tensorflow 2.4.1 的支持： https://www.tensorflow.org/install/gpu然后是实际的 Tensorflow 安装： 8810563352708588：https://38.864 . 然后我运行一段代码示例，它应该检查 GPU 并尝试运行张量。 它说它看不到某些 CUDA 文件，其中我粘贴了下面的所有文件（也与路径匹配）。 我把我能想到的所有相关信息都粘贴在下面。

我检查以确保所有 dll 都已正确创建：

PS C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\bin> ls


    Directory: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\bin


Mode                 LastWriteTime         Length Name
----                 -------------         ------ ----
d-----         2/13/2021  10:04 PM                crt
-a----         7/23/2020   6:58 AM         203776 bin2c.exe
-ar---         7/27/2020   4:11 PM             58 compute-sanitizer.bat
-a----          8/4/2020   1:57 AM       89057280 cublas64_11.dll
-a----          8/4/2020   1:57 AM      167523328 cublasLt64_11.dll
-a----         7/23/2020   6:58 AM         373248 cuda-memcheck.exe
-a----         7/23/2020   6:58 AM        4604928 cudafe++.exe
-a----         7/23/2020   6:59 AM         335872 cudart32_110.dll
-a----         7/23/2020   6:59 AM         401408 cudart64_110.dll
-a----         2/13/2021   8:49 PM         222720 cudnn64_8.dll
-a----         2/13/2021   8:49 PM       99908096 cudnn_adv_infer64_8.dll
-a----         2/13/2021   8:49 PM       65183232 cudnn_adv_train64_8.dll
-a----         2/13/2021   8:49 PM      508834304 cudnn_cnn_infer64_8.dll
-a----         2/13/2021   8:49 PM       51145728 cudnn_cnn_train64_8.dll
-a----         2/13/2021   8:49 PM      240333312 cudnn_ops_infer64_8.dll
-a----         2/13/2021   8:49 PM       25651712 cudnn_ops_train64_8.dll
-a----         7/23/2020   6:59 AM      162878976 cufft64_10.dll
-a----         7/23/2020   6:59 AM         272896 cufftw64_10.dll
-a----         7/23/2020   6:59 AM        1305600 cuinj64_110.dll
-a----         7/23/2020   6:58 AM        2828800 cuobjdump.exe
-a----         7/23/2020   6:59 AM       52446720 curand64_10.dll
-a----         7/23/2020   6:59 AM      525173248 cusolver64_10.dll
-a----         7/23/2020   6:58 AM      314601984 cusolverMg64_10.dll
-a----         7/23/2020   6:59 AM      159789568 cusparse64_11.dll
-a----         7/23/2020   6:58 AM         302080 fatbinary.exe
-a----         7/23/2020   6:59 AM         232448 nppc64_11.dll
-a----         7/23/2020   6:59 AM       10907648 nppial64_11.dll
-a----         7/23/2020   6:59 AM        4569600 nppicc64_11.dll
-a----         7/23/2020   6:59 AM        7840256 nppidei64_11.dll
-a----         7/23/2020   6:59 AM       51866112 nppif64_11.dll
-a----         7/23/2020   6:59 AM       26065920 nppig64_11.dll
-a----         7/23/2020   6:59 AM        6411264 nppim64_11.dll
-a----         7/23/2020   6:59 AM       20231168 nppist64_11.dll
-a----         7/23/2020   6:59 AM         203264 nppisu64_11.dll
-a----         7/23/2020   6:59 AM        2856960 nppitc64_11.dll
-a----         7/23/2020   6:59 AM        8873472 npps64_11.dll
-a----          8/4/2020   1:57 AM         299520 nvblas64_11.dll
-a----         7/23/2020   6:58 AM         408576 nvcc.exe
-a----         7/23/2020   6:58 AM            334 nvcc.profile
-a----         7/23/2020   6:58 AM       28080128 nvdisasm.exe
-a----         7/23/2020   6:58 AM        3133952 nvjpeg64_11.dll
-a----         7/23/2020   6:58 AM        7523328 nvlink.exe
-a----         7/23/2020   6:58 AM        2173440 nvprof.exe
-a----         7/23/2020   6:58 AM         225792 nvprune.exe
-a----         7/23/2020   6:58 AM        5118976 nvrtc-builtins64_110.dll
-a----         7/23/2020   6:58 AM       16161792 nvrtc64_110_0.dll
-a----         7/23/2020   6:58 AM             53 nvvp.bat
-a----         7/23/2020   6:58 AM        7407104 ptxas.exe

这是我的路径：

C:\Users\jnels>echo %PATH% C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\bin;C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\libnvvp;C :\Windows\system32;C:\Windows;C:\Windows\System32\Wbem;C:\Windows\System32\WindowsPowerShell\v1.0;C:\Windows\System32\OpenSSH;881038565文件\NVIDIA Corporation\PhysX\Common;C:\Program Files\NVIDIA Corporation\Nsight Compute 2020.1.2;C:\Program Files\Docker\Docker\resources\bin;C:\ProgramData\DockerDesktop\version-bin;8586378:8586378 \Users\jnels\AppData\Local\Microsoft\WindowsApps;

然后我按照说明运行示例代码，该代码将测试 GPU 和 Tensorflow 以查看它是否有效。 我应该期望看到张量返回：

PS C:\Users\jnels> python -c "import tensorflow as tf;print(tf.reduce_sum(tf.random.normal([1000, 1000])))"
2021-02-13 23:24:28.372069: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'cudart64_110.dll'; dlerror: cudart64_110.dll not found
2021-02-13 23:24:28.372201: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
2021-02-13 23:24:29.637123: I tensorflow/compiler/jit/xla_cpu_device.cc:41] Not creating XLA devices, tf_xla_enable_xla_devices not set
2021-02-13 23:24:29.638344: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library nvcuda.dll
2021-02-13 23:24:29.692627: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1720] Found device 0 with properties:
pciBusID: 0000:0b:00.0 name: GeForce RTX 3080 computeCapability: 8.6
coreClock: 1.71GHz coreCount: 68 deviceMemorySize: 10.00GiB deviceMemoryBandwidth: 707.88GiB/s
2021-02-13 23:24:29.692892: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'cudart64_110.dll'; dlerror: cudart64_110.dll not found
2021-02-13 23:24:29.694758: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'cublas64_11.dll'; dlerror: cublas64_11.dll not found
2021-02-13 23:24:29.695158: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'cublasLt64_11.dll'; dlerror: cublasLt64_11.dll not found
2021-02-13 23:24:29.695609: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'cufft64_10.dll'; dlerror: cufft64_10.dll not found
2021-02-13 23:24:29.696010: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'curand64_10.dll'; dlerror: curand64_10.dll not found
2021-02-13 23:24:29.696368: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'cusolver64_10.dll'; dlerror: cusolver64_10.dll not found
2021-02-13 23:24:29.696722: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'cusparse64_11.dll'; dlerror: cusparse64_11.dll not found
2021-02-13 23:24:29.697079: W tensorflow/stream_executor/platform/default/dso_loader.cc:60] Could not load dynamic library 'cudnn64_8.dll'; dlerror: cudnn64_8.dll not found
2021-02-13 23:24:29.697327: W tensorflow/core/common_runtime/gpu/gpu_device.cc:1757] Cannot dlopen some GPU libraries. Please make sure the missing libraries mentioned above are installed properly if you would like to use GPU. Follow the guide at https://www.tensorflow.org/install/gpu for how to download and setup the required libraries for your platform.
Skipping registering GPU devices...
2021-02-13 23:24:29.698128: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2021-02-13 23:24:29.699113: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1261] Device interconnect StreamExecutor with strength 1 edge matrix:
2021-02-13 23:24:29.699192: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1267]
2021-02-13 23:24:29.699571: I tensorflow/compiler/jit/xla_gpu_device.cc:99] Not creating XLA devices, tf_xla_enable_xla_devices not set
tf.Tensor(1250.3312, shape=(), dtype=float32)

Answer 1

您尚未将所有必需的安装目录添加到路径环境变量中以支持tensorflow_gpu 。

如下所示将CUDA® 、 CUPTI和cuDNN安装目录添加到%PATH%环境变量，并更新您的%PATH%以匹配：

SET PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\bin;%PATH%
SET PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\extras\CUPTI\lib64;%PATH%
SET PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.0\include;%PATH%
SET PATH=C:\tools\cuda\bin;%PATH%

您可以参考此链接以获取更多详细信息。