RTX 5070Ti GPU and CUDA error

1

Hello,

I recently tried to run RFantibody on my computer, and got the following errors. RFantibody was designed to run in a Docker container. Does anyone know how to fix this? I tried ChatGPT, but that did not help at all. Really appreciate any input. If you need more information please let me know.

NVIDIA GeForce RTX 5070 Ti with CUDA capability sm_120 is not compatible with the current PyTorch installation.
The current PyTorch install supports CUDA capabilities sm_50 sm_60 sm_70 sm_75 sm_80 sm_86 sm_37 sm_90.
If you want to use the NVIDIA GeForce RTX 5070 Ti GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/

RuntimeError: CUDA error: no kernel image is available for execution on the device
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions

2

You need to install a PyTorch binary with CUDA 12.8 runtime dependencies for Blackwell support.

3

image
image711×949 32.9 KB

I changed my pyproject file and setup file to something like this, and I keep getting this error: “Because rfantibody depends on torch (2.4.0) which doesn’t match any versions, version solving failed.”

I think the problem lies in the versions of dgl and torch, but I do not know much about coding and just want to set up this model for my application. Thanks for all the help.

4

image
image857×494 20.7 KB

5

You need to install PyTorch >= 2.7.0 with CUDA 12.8 (or a nightly binary created after ~Feb. 2025) to support your Blackwell GPU. I don’t know what rfantibody is but you won’t be able to run any application on your 5070Ti with PyTorch 2.4.0.