Support for NVIDIA RTX 50-series GPUs

[retryLatr]

Hi,
I am currently trying to set up the environment and run the provided Octo and RT-1 inference on a PC with an NVIDIA RTX 50series GPU. However, it seems that the project does not yet support the latest NVIDIA GPU architecture.
I would like to ask whether there are plans to support RTX 50-series GPUs in the future？
The main error message is as follows:

TensorFlow was not built with CUDA kernel binaries compatible with compute capability 12.0.
CUDA kernels will be jit-compiled from PTX
...
Falling back to the CUDA driver for PTX compilation; ptxas does not support CC 12.0
Used ptxas at /home/.../miniconda3/envs/simpler_env/lib/python3.10/site-packages/nvidia/cuda_nvcc/bin/ptxas

Any guidance or suggestions would be greatly appreciated. Thanks for your work!

[xuanlinli17]

Hi,

Tensorflow is optional (only used to infer RT-1 and Octo, not used in simpler env itself), so you can try to update to newer tensorflow versions and see if the checkpoints can run on RTX 50-series.