Abhinav Reddy
Abhinav Reddy
It would appear that the [PyTorch NGC containers do not work on Jetson](https://forums.developer.nvidia.com/t/failed-to-run-pytorch-ngc-docker-on-jetson-nano/216879/3). Using Dusty's docker image [`dustynv/torch_tensorrt:r35.4.1`](https://github.com/dusty-nv/jetson-containers/tree/master/packages/pytorch/torch_tensorrt), results in this error when importing torch_tensorrt. ```root@36743666126a:/# python3 Python 3.8.10 (default,...
EDIT: The iGPU container from NGC is viable and has a working torch tensorrt, but flash attention seems to fail to compile correctly. ------------------------------------------------------------------------------------------------------------------------- When attempting to run a model...