Apex安装失败（笔记记录分享）

Apex安装失败（笔记记录）

1. 错误合集

No.1: error detected in the compilation of "csrc/multi_tensor_scale_kernel.cu"

No.2: module 'torch.nn' has no attribute 'backends'

No.3: AttributeError: module 'torch.distributed' has no attribute '_all_gather_base'

No.4: ImportError was: ModuleNotFoundError("No module named 'amp_C'")

No.56789...: 乱七八糟 

2. 解决

1. 直接delete原本的Apex

2. 命令

   git clone https://github.com/NVIDIA/apex.git
   cd apex
   git checkout f3a960f80244cf9e80558ab30f7f7e8cbf03c0a0
   python setup.py install --cuda_ext --cpp_ext
   pip3 install -v --no-cache-dir ./
   

3. 勿杠，仅做笔记分享。

4. 参考文章（反正我也不明白为啥就能运行跑通，多方参考，大力出奇迹）：
   https://blog.csdn.net/qq_17783559/article/details/127813381、https://zhuanlan.zhihu.com/p/360412833