Increase the inference speed #2851
Description
📚 Documentation
The speed when i use a faster_rcnn model to do inference on 2cpu is very slow(20s per pic). I try to find so ways to increase the speed and i find somebody says that, there exists a script can convert torch model to caffe2 to increase the inference speed in Tool. But i can't find it.
So does detectron2 support some speed increasing ways on cpu?
- Tool/caffe2_convert.py