tf-gpu 2.6.5

TensorFlow is an open source machine learning framework for everyone.

TensorFlow is an open source software library for high performance numerical computation. Its flexible architecture allows easy deployment of computation across a variety of platforms (CPUs, GPUs, TPUs), and from desktops to clusters of servers to mobile and edge devices.

Originally developed by researchers and engineers from the Google Brain team within Google's AI organization, it comes with strong support for machine learning and deep learning and the flexible numerical computation core is used across many other scientific domains.

build environment is: 
cuda 11.3.1 
cudnn 8.2.1 
nccl 2.9.9 
tensorrt: 8.0.1 GA optional
nvidia Compute Capability 6.0 6.1 7.0 7.5 8.0 8.6

build https://github.com/nvidia/tensorflow by https://github.com/ssbuild with mkl support. and test gpu pass as follow cuda 116 and cuda 113 , any other you can try also.

docker pull ssdog/cuda:11.6.2-runtime-ubuntu18.04
docker pull ssdog/cuda:11.6.2-runtime-ubuntu20.04
docker pull ssdog/cuda:11.3.1-runtime-ubuntu18.04
docker pull ssdog/cuda:11.3.1-runtime-ubuntu20.04