Whether the inference speed can be optimized

Parskatt / DKM

[CVPR 2023] DKM: Dense Kernelized Feature Matching for Geometry Estimation

https://parskatt.github.io/DKM/

Other

378 stars 28 forks source link

Whether the inference speed can be optimized #50

Open NickJony opened 5 months ago

NickJony commented 5 months ago

Great job! I do inference on 3090, it takes about 0.7 seconds to calculate the time of two images, and what other operations can be used for inference acceleration, in addition to reducing the image resolution.

Parskatt commented 5 months ago

You can try fp16, see my code in roma for some examples.

I haven't gotten torch.compile to work, but it could improve things.