通过NVIDIA tensorflow使用cuda11和cudnn8

官方的tensorflow1.1x只支持cuda10.0和cudnn7 如何在更高的版本cuda和cudnn8使用tensorflow1.1x呢 最简单的方法是使用nvidia修改维护的tensorflow
GitHub - NVIDIA/tensorflow: An Open Source Machine Learning framework for Everyone

TensorFlow User Guide :: NVIDIA Deep Learning frameworks documentation

RTX3080 Ubuntu18.04 cuda11.1 cudnn8.0.4 TensorFlow1.15.4 PyTorch1.7.0环境配置_wu496963386的博客-CSDN博客

版本关系

TensorFlow Release Notes :: NVIDIA Deep Learning frameworks documentation

直接使用镜像

docker pull nvcr.io/nvidia/tensorflow:21.06-tf1-py3 # example of use 21.06-tf1-py3

rebuild tensorflow

TensorFlow User Guide :: NVIDIA Deep Learning frameworks documentation

FROM nvcr.io/nvidia/tensorflow:21.08
# Bring in changes from outside container to /tmp
# (assumes my-tensorflow-modifications.patch is in same directory as Dockerfile)
COPY my-tensorflow-modifications.patch /tmp
# Change working directory to TensorFlow source path
WORKDIR /opt/tensorflow
# Apply modifications
RUN cd tensorflow-source 
 patch -p1 /tmp/my-tensorflow-modifications.patch
# Rebuild TensorFlow
RUN ./nvbuild.sh
# Reset working directory
WORKDIR /workspace

也就是在镜像里面包含了重新编译tf的代码和流程。