dav-ell

I think I discovered the issue. It seems that using nn.Upsample causes a segmentation fault for unknown reasons. It also looks like F.upsample has the same issue. @narendasan any ideas?...

Potentially related to https://github.com/pytorch/pytorch/issues/27376 and https://github.com/onnx/onnx-tensorrt/issues/361#issuecomment-576004279.

I was able to get the debug logs you're looking for, and the error seems to be an off-by-one issue (see attached for full log). ```python DEBUG: [Torch-TensorRT] - ITensor...

Yeah, trying to find a minimal reproducible script now. Is there a way to make the conversion fall back to PyTorch for these kind of issues, or is fixing this...

I've been able to convert the most problematic part of my network (so far), but when performing inference it gives me a segmentation fault. @narendasan can you help with this...

Also interested in an answer to this.

@justinmeister, your other game, [The Stolen Crown](https://github.com/justinmeister/The-Stolen-Crown-RPG), is licensed under an MIT license. Any chance of this repository being licensed similarly?

Also looking for an update on this. These features are pretty critical to using this repo.

This issue is still present in ROS2 Humble (using `ros:humble` docker image) and a script like the following: ```python import numpy as np import rclpy from rclpy.node import Node from...