<>RTX3080ti Pytorch Pit Guide
* Write it in the front, wordy
All say 3080 I won't support it cuda10, But reproduce the code of the previous two years , Corresponding to lower version pytorch, But how ?
After personal experiment , install cuda10, Run various codes cuda Related error reporting .
RuntimeError: CUDA error: CUBLAS_STATUS_EXECUTION_FAILED when calling
`cublasSgemm( handle, opa, opb
（ resolvent ： New environment ,pytorch Official website pip install . In the end ）
Finally found a metaphysical method .（ Personal experience only ）
If in cuda110 Directly use the following command to install pytorch1.4.0, it won 't work .（ come from pytorch Official website ）
（ Da baa ）conda install pytorch==1.4.0 torchvision==0.5.0 cudatoolkit=10.1 -c pytorch
<> serious affairs
3080 Your server is still installed cuda11. stay anaconda When creating an environment , Directly create the corresponding environment python, then conda install
pytorch==1.X.X, Can . The following two are my experiments . Can use , Make a record .
Python3.8-》 Pytorch1.4.0 -》 Cuda101
Python3.7 -》 Pytorch1.3.1 -》 Cuda100
Said the younger martial brother , Official website code pip It's easy to pretend , I didn't try this , Easy to use !!!
（ Not professional , The above is only personal experience , Do not spray ）