Hello, I’m trying to upgrade to pytorch 1.2. I’ve found that official guidance seems to need cudatoolkit version to be 9.2 or higher. Is it possible to use pytorch 1.2 with cudatoolkit 9.0? Thanks for any help!
You should be able to compile PyTorch using CUDA9.0.
However, the binaries dropped the support for CUDA8.0 and CUDA9.0 in the latest release.
Thank you! Sorry for the late reply. I have tried to build pytorch 1.2 from source following the instruction on the official github website. I came across the following problem: “ninja: build stopped: subcommand failed.” I can’t provide the detailed error report now because our server is in the progress of update. When it’s done, I’ll try it again. Thanks for your help!