Fix runtime error in detectron2 sample due to wrong variable usage by luxi78 · Pull Request #4665 · NVIDIA/TensorRT

@luxi78

@luxi78

…ons was passed to the execution context instead of self.device_memories.

This likely resulted from an incomplete refactor of the cuda memory allocation code. The sample now runs correctly.

Signed-off-by: luxi78 <luxi78@gmail.com>