Does Ubuntu 22.04, RTX 5090 support the use of isaac_ros_nvblox?

Robotics & Edge Computing Robotics - Isaac Isaac ROS
1

my machine:

ubuntu22.04, rtx5090, Driver Version: 580.95.05 CUDA Version: 13.0, use isacc_ros_nvblox, release:3.2,

Will it run successfully on my pc? Does it currently not support the 5090 graphics card?

I’m using nvblox_torch-0.0.8 from GitHub - nvidia-isaac/nvblox: A GPU-accelerated TSDF and ESDF library for robots equipped with RGB-D cameras. , and I’m getting errors on my machine:

NVIDIA GeForce RTX 5090 Laptop GPU 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_90. If you want to use the NVIDIA GeForce RTX 5090 Laptop GPU GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/ warnings.warn( The embedding dimension of the RADIO model is larger than the maximum feature array size that nvblox supports: 128. The feature extractor will randomly project the output of the model to a lower dimension. Traceback (most recent call last): File “”, line 198, in _run_module_as_main File “”, line 88, in _run_code File “/workspaces/nvblox/nvblox_torch/nvblox_torch/examples/reconstruction/sun3d.py”, line 176, in sys.exit(main()) ^^^^^^ File “/workspaces/nvblox/nvblox_torch/nvblox_torch/examples/reconstruction/sun3d.py”, line 148, in main visualizer = Visualizer(deep_feature_embedding_dim=RadioFeatureExtractor().embedding_dim()) ^^^^^^^^^^^^^^^^^^^^^^^ File “/workspaces/nvblox/nvblox_torch/nvblox_torch/examples/utils/feature_extraction.py”, line 81, in init self.projection_matrix = get_random_projection_matrix( ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File “/workspaces/nvblox/nvblox_torch/nvblox_torch/examples/utils/feature_extraction.py”, line 42, in get_random_projection_matrix projection_matrix = torch.rand(input_channel_num, output_channel_num, device=device) * 2.0 - 1.0 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 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.

Will I encounter the same limitations if I use GitHub - NVIDIA-ISAAC-ROS/isaac_ros_nvblox: NVIDIA-accelerated 3D scene reconstruction and Nav2 local costmap provider using nvblox?

2

Hi @fhqjwyk ,

Yes, Isaac ROS 3.2 targets CUDA 12.6, which does not support Blackwell series GPUs like the 5090.

We have recently released Isaac ROS 4.0 which supports CUDA 13.0 and Blackwell. However, the target Ubuntu distro for Isaac ROS 4.0 is 24.04.

Please refer to the system requirements pages for each release for more information: