tensorflow gpu踩坑记录

声明版本：

python版本--3.7.6

cuda版本--11.4.0

cudnn版本--8.2

tensorflow-gpu版本--2.4.0

①解决Not creating XLA devices, tf_xla_enable_xla_devices not set：

在import tensorflow后加(如果不加会报错，则每次都要加)

import os

os.environ['TF_XLA_FLAGS'] = '--tf_xla_enable_xla_devices'

②解决Could not load dynamic library 'cusolver64_10.dll'; dlerror: cusolver64_10.dll not found

将C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.4\bin中的cusolver64_11.dll更名为cusolver64_10.dll

③导入tensorflow测试是否成功

import tensorflow as tf

tf.test.is_gpu_available()

④解决Could not load dynamic library 'cudart64_100.dll'; dlerror: cudart64_100.dll not found

缺少cudar64_100.dll,去官网下载，并放入C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.4\bin中​​​​​​https://download.zip.dll-files.com/f0ab3cac7e90959a38e97b262ebdf3f2/cudart64_100.zip?token=TgabjjguiZ8vuxLodfS1MQ&expires=1637069619https://download.zip.dll-files.com/f0ab3cac7e90959a38e97b262ebdf3f2/cudart64_100.zip?token=TgabjjguiZ8vuxLodfS1MQ&expires=1637069619

 ⑤解决pip install --upgrade tensorflow报错

ERROR: Could not find a version that satisfies the requirement werkzeug>=0.11.15 (from tensorboard~=2.6->tensorflow) (from versions: none)
ERROR: No matching distribution found for werkzeug>=0.11.15 (from tensorboard~=2.6->tensorflow)

 将pip install --upgrade tensorflow改为pip --trusted-host pypi.python.org install --upgrade tensorflow