pytorch 0.3发布了
性能提升、新的网络层，支持ONNX，正式支持CUDA9、CuDNNv7，大量bug修复。
Performance improvements, new layers, ship models to other frameworks (via ONNX), CUDA9, CuDNNv7, lots of bug fixes
具体更新信息请移步官方的release：https://github.com/pytorch/pytorch/releases
新的编译好的更新包请移步这里：https://zhuanlan.zhihu.com/p/31747695
快捷门更新包（whl）：https://pan.baidu.com/s/1dF6ayLr#list/path=%2F&parentPath=%2F
更新包是上面那个连接的童鞋编译好的，他只编译好了whl版的conda版的暂时没编译好。
不想编译的童鞋可以直接到官网release界面下载源码版自行进行编译
安装步骤
下好上面的whl文件。在win10下进行pip安装。
C:\Users\dell\PycharmProjects\Pytorch-Learn>C:\Users\dell\Anaconda3\envs\my-pytorch\Scripts\pip install "torch-0.3.0b0+591e73e-cp36-cp36m-win_amd64.whl"
Processing c:\users\dell\pycharmprojects\pytorch-learn\torch-0.3.0b0+591e73e-cp36-cp36m-win_amd64.whl
Requirement already satisfied: pyyaml in c:\users\dell\anaconda3\envs\my-pytorch\lib\site-packages (from torch==0.3.0b0+591e73e)
Requirement already satisfied: numpy in c:\users\dell\anaconda3\envs\my-pytorch\lib\site-packages (from torch==0.3.0b0+591e73e)
Installing collected packages: torch
Found existing installation: torch 0.2.1+a4fc05a
Uninstalling torch-0.2.1+a4fc05a:
Successfully uninstalled torch-0.2.1+a4fc05a
Successfully installed torch-0.3.0b0+591e73e
