How to get a tensorrt model?
muse1835 opened this issue · comments
Hi,
I want to get the RT-DETR tensorrt models.
But I can't find any code about loading model or custom trained weights.
(I use pytorch.)
Could you explain?
You find some resources for deployment in #95
star this repo to follow updates
thanks for reply.
Is there any code about converting torch model to engine??...
I already saw the link but I couldn't convert my .pth model to .engine.
I want to convert my .pth to .engine.
could you give me some refer or python codes?
(I tried converting but some error about 'state_dict' was ocurred after torch2trt, torch.load(".pth")
I will upload my error message 12h later.)
You can use nv tensorrt tool trtexec
.
trtexec --onnx=xxx.onnx --saveEngine=xxx.engine --buildOnly --fp16