TensorFlow 2.4.0 Keras GPU 问题 解决

运行环境

Acer Swift X RTX3050
Win10
VS2019+Python 3.7

安装过程

Pip install tensorflow_gpu
自动安装Tensorflow 2.6.0
经过优快云查询得知
Tensorflow 2.6.0对应
cuda=11.2
cudnn=8.1.0 (for cuda 11.2)
详见Dr.Debug
文章
下载地址
CUDA
Cudnn

问题

安装CUDA和Cudnn并设置环境变量后，运行bandwidthTest.exe和deviceQuery.exe均显示Result = PASS!

运行Python测试
s = tf.version
显示正确
开始训练模型
在

model.fit(train_images,train_labels, batch_size=128, epochs=7)

VS2019出现
Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above.
[[node sequential/conv2d/Conv2D (defined at…
错误
程序输出
2021-08-19 12:16:27.527652: E tensorflow/stream_executor/cuda/cuda_dnn.cc:336] Could not create cudnn handle: CUDNN_STATUS_NOT_IN