AttributeError: 'CustomFastRCNNOutputs' object has no attribute '_log_accuracy'

[codeman-24]

AttributeError: 'CustomFastRCNNOutputs' object has no attribute '_log_accuracy'
when i run train "python projects/UniDet/train_net.py --config-fil│
e projects/UniDet/configs/Base-CRCNN-COCO.yaml --num-gpus 8"

[LigZhong]

I switched to Detectron 2.0.2 as I remember, which resolved this problem.

[xingyizhou]

Hi,
Sorry for the inconvenient, this is a detectron2 version mismatch, you can change the line with self._log_accuracy to _log_classification_stats. After from detectron2.modeling.roi_heads.fast_rcnn import _log_classification_stats.

[ahong007007]

sorry, I have a problem .

_log_classification_stats is a function, what parameters need to be passed in? How to calculate the accuracy and recall rate without parameters?
Thank you very much for your reply and I wish you a happy work.