![]() 1.2 Copy the codeĪfter installation, be sure to also be on the outermost package of the main code if _name_ = '_main_': Copy the codeĬonsider the example in my article yesterday, because PyTorch's multithreaded library doesn't work properly on Windows. If you can't stand the snail-like network speed of Conda, then I will upload the package to the public cloud for you, you can try seven niuyun or Baidu cloud, after you download, enter the following commands: conda install numpy mkl cffiĬonda install -offline path/to/tar/pytorch- 0.1. To install, if you don't mind the speed of anaconda Cloud, just type the following command: conda install -c peterjc123 pytorch= 0.1. If you don't want to give up Anaconda2, you can also create a virtual environment to use it. ![]() CuDNN V5 + (already shipped with cuDNN V5).Windows 10 64-bit OS (others not available due to INCONSISTENT CUDA binaries).To use it directly, you need to meet the following criteria: Update: Support has been added for all graphics cards with Compute Capability=2.0ĭon't get too excited just yet. Thank 1494598013779 for testing the installation of conda package. Yesterday I posted a post about the process of compiling PyTorch for 64-bit Windows. Python The compiler Deep learning pytorch windows
0 Comments
Leave a Reply. |