Environment:
- Hardware: NVIDIA Jetson Orin NX (16GB)
- JetPack Version: 6.0 (L4T r36.3.0)
- Python Version: 3.10
- Location: Taiwan (Huwei)
Issue Description:
I am developing an autonomous pruning robot and I urgently need to install PyTorch 2.3.0 with CUDA support to run real-time object detection for branch identification. However, my local network environment in Taiwan (Huwei) is strictly blocking access to nvidia.osuosl.org.
What I’ve tried:
- wget/curl from terminal: Resulted in “Connection timed out” or “404 Not Found”.
- Changed DNS to 8.8.8.8: Still unable to resolve host.
- Modified /etc/hosts: “No route to host” when connecting to 140.211.169.60.
- Using ‘pip3 install --index-url https://pypi.jetson-ai-lab.dev/jp6/cu122’: It only installs a 100MB CPU-only version.
Questions:
- Is there any mirror link for the 1.1GB ‘.whl’ file (torch-2.3.0-cp310-cp310-linux_aarch64.whl)?
- Since torch.cuda.is_available() returns ‘False’, my YOLO detection is too slow. Is there a way to get the CUDA-enabled version in a restricted network?