system

[haizhu12]

Is it possible to run the sample code under windows10 system?

[PeterWang512]

I haven't test the code on windows, but it should be doable I think. You'll need nvcc and cuda installed.

[haizhu12]

@PeterWang512 Excuse me, When I run generate sample script, but processing stop at line 11 of [fused_act.py]。 RuntimeError: Ninja is required to load C++ extensions

[0sure]

@PeterWang512 Excuse me, When I run generate sample script, but processing stop at line 11 of [fused_act.py]。 RuntimeError: Ninja is required to load C++ extensions

Hello, I think you can use pip install ninja to build a C++ extension.

[haizhu12]

@PeterWang512 @0sure Excuse me, the GPU I use is NVIDIA RTX3060, can RTX3060 run the sample code? When I run this sample code after configuring the relevant environment, there is an error, and the program error is as follows:

UserWarning: NVIDIA GeForce RTX 3060 with CUDA capability sm_86 is not compatible with the current PyTorch installation. The current PyTorch install supports CUDA capabilities sm_37 sm_50 sm_60 sm_70 sm_75. If you want to use the NVIDIA GeForce RTX 3060 GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/

relevant environment：ubuntu 20.04 GPU:NVIDIA RTX3060 12G tensorflow-gpu-2.4.0 cuda:11.0 cuDNN:8.0 pytorch:1.7.1

[0sure]

I think the GPU you are currently using should be able to run the code in this repo. I use Quadro P6000, pytorch = 1.8.1, tensorflow-gpu=2.3.0, cuda=10.2.89, and cudnn=8.2.1.32. I guess the problem may be related to CUDA. I think you should also check the local configuration according to the link of pytorch given to you.