TensorRT version Error
I got the error during run the demo.py. Does the demo only work on the A100?
container : use nvcr.io/nvidia/pytorch:23.04-py3.
GPU : T4
[05/17/2023-03:30:34] [I] [TRT] Loaded engine size: 14024 MiB
[05/17/2023-03:30:34] [E] [TRT] 6: The engine plan file is not compatible with this version of TensorRT, expecting library version 8.6.1.2 got 8.6.0.12, please rebuild.
[05/17/2023-03:30:34] [E] [TRT] 2: [engine.cpp::deserializeEngine::951] Error Code 2: Internal Error (Assertion engine->deserialize(start, size, allocator, runtime) failed. )
Segmentation fault (core dumped)
@rayleee
Yes, you need Ampere architecture( like A100 or newer) because a lot of kernels we used are compiled under this architecture.
We may release other models for other architectures in the future (like for V100). But for your T4, you may encounter memory problem even you have right model.
it seem you tensorRT not match.
you can do this to solve it.
```bash
# download old tensorRT
wget https://developer.nvidia.com/downloads/compute/machine-learning/tensorrt/secure/8.6.0/local_repos/nv-tensorrt-local-repo-ubuntu2004-8.6.0-cuda-12.0_1.0-1_amd64.deb
# install (maybe it extract)
dpkg -i nv-tensorrt-local-repo-ubuntu2004-8.6.0-cuda-12.0_1.0-1_amd64.deb
# go to this dir
cd /var/nv-tensorrt-local-repo-ubuntu2004-8.6.0-cuda-12.0/
# install all deb file
dpkg -i *.deb
@rayleee @Tlntin We have updated to a new accelerated version and removed the previous TensorRT acceleration version. The new version has undergone significant optimization at the source code level, resulting in improved performance, ease of use, and GPU compatibility. Please update and feel free to try it out.