FP16 Inference with YOLOX Doesn't Provide Speedup

Troubleshooting
2

Hello,

This thread may be useful. TLDR unless you use Tensor Core FP16 will not give you the performance improvements seen in TRT.