YOLOX Model conversion error reported: [TensorRT] ERROR: runtime.cpp (25) – Cuda Error in allocate: 2 (out of memory)
Configuration:
ubuntu 18.04
pytroch 1.9.0
cuda10.0
cudnn 7.5.0
TensorRT 5.1.5
torch2trt 0.3.0
Solution:
Change the max_workspace_size
in trt.py