SHENHAIBB

> That's wired, code [here](https://github.com/Megvii-BaseDetection/cvpods/blob/master/tools/train_net.py#L148-L149) add sys path to import config. maybe the dir is wrong. In my linux, i must input the whole dir not "../"

2022-04-29 13:20:12.044 | INFO | cvpods.evaluation.evaluator:inference_on_dataset:112 - Start inference on 399 data samples 2022-04-29 13:20:13.201 | INFO | cvpods.evaluation.evaluator:inference_on_dataset:142 - Inference done 12/399. 0.0739 s / sample. ETA=0:00:28 2022-04-29 13:20:18.268...

> 0.0736 second per sample per device. yeah, i have only one gpu then my inference time is 0.1519s or 0.0736s

import shapely.geometry as shgeo

I use pytorch1 9.0+cu111, what is the error you encountered?

Every time the test reached 700, the program would break. Why？