-
Notifications
You must be signed in to change notification settings - Fork 403
Add RTX 50 series GPU compatibility via PyTorch 2.7 and CUDA 12.8 #369
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
base: main
Are you sure you want to change the base?
Conversation
|
There are currently no Docker images that can run on RTX 50 series GPU. The one built for RTX 40 series card (see #27) does not work on RTX 50 series due to the issues addressed in this PR. |
|
Thank you! It works on RTX 5090 in WSL with docker. Except that I still need to import |
|
Hi, I have followed your instruction. However, there is an error says "No module named 'torch'" when i tried to bash build_all_conda.sh to install mycpp. I am sure theres pytorch 2.7.1 with cuda 12.8 on my 5090D PC. And the detail of bugs is listed below. Could you please let me know why and what should i do? This warning is for project developers. Use -Wno-dev to suppress it. -- Found Boost: /home/dell/miniconda3/envs/foundationpose/lib/cmake/Boost-1.85.0/BoostConfig.cmake (found version "1.85.0") found components: system program_options -- Found PythonInterp: /home/dell/miniconda3/envs/foundationpose/bin/python (found suitable version "3.9.23", minimum required is "3.8") × Getting requirements to build editable did not run successfully. note: This error originates from a subprocess, and is likely not a problem with pip. |
|
@GXYsteven Hi, have you sovled this problem? I encountered the same issue when using pytorch 2.9 and cuda 13.0. |
You need to install torch in your environment and then run without build isolation. |
This PR addresses multiple compatibility issues that prevent FoundationPose from building and running on RTX 50 series (sm_120 architecture) GPUs with modern PyTorch versions.
bundlesdf/mycuda/common.cuto fix compilation errors with PyTorch 2.7+bundlesdf/mycuda/setup.pyto meet modern PyTorch requirementsmycpp/src/app/pybind_api.cppby changing%dto%zuforsize_tvaluesrequirements.txtto use PyTorch 2.7.1 with CUDA 12.8 for RTX 50 series compatibilitymycppimport path inUtils.pyfrommycpp.build.mycpptomycppreadme.md