Ref: https://github.com/NVIDIA/tensorflow/tree/r1.15.5+nv21.11

灵感来源……
– 最近搞复现发现很多以前的论文会用到tensorflow 1.x框架
– 跑代码时好不容易装完requirements，训练模型报错blas GEMM launch failed
– 疑似显存不足，然后看到网上反馈tensorflow 1.x不支持30系显卡，遂更新到tensorflow 2.x，代码果然在开头import的时候崩了
– 得知Nvidia针对老版本tensorflow开发了适用于新显卡的框架？还有这种好事，Nvidia Fuck Me

记录下安装教程罢别哪天自己忘了（
这是仓库中NVIDIA的介绍：

NVIDIA has created this project to support newer hardware and improved libraries to NVIDIA GPU users who are using TensorFlow 1.x. With release of TensorFlow 2.0, Google announced that new major releases will not be provided on the TF 1.x branch after the release of TF 1.15 on October 14 2019. NVIDIA is working with Google and the community to improve TensorFlow 2.x by adding support for new hardware and libraries. However, a significant number of NVIDIA GPU users are still using TensorFlow 1.x in their software ecosystem. This release will maintain API compatibility with upstream TensorFlow 1.15 release. This project will be henceforth referred to as nvidia-tensorflow.

简单来说，许多用户手上有新显卡，但是原生的tensorflow 1.x并不支持新的显卡架构（例如30系以后的显卡）；针对这种情况，NVIDIA开发了新的TF框架，使我们能够兼得TF 1.x与新架构显卡。此框架基于TF 1.15开发。

接下来我就直接搬仓库的readme了

要求：
– 只能在Linux环境下使用（甚至原生TF现在也不支持在windows下用GPU了Pytorch大法好），Ubuntu 20.04或以上
– CUDA支持的GPU
– GPU驱动版本须在r445以上
– 若whl安装，需要python 3.8以上与pip 19.0以上

由于此TF没有在PyPI上架，所以首先要pip install --user nvidia-pyindex；

然后安装NVIDIA Tensorflow：pip install --user nvidia-tensorflow[horovod]

此TF已支持GPU运算，无需额外安装CUDA和cuDNN（沟槽的在windows上试了好几版CUDA装了卸卸了装）

然后就可以开吃啦