result after running eval.py

[oliver0922]

Hi Liu

I applied weight pruning on pretrained detector.core.backbone(resnet34) after training 80 epochs.

The process was training -> pruning -> finetuning.

However when I ran the eval.py.

The result was like this below

Can you guess what's the problem?

The result after 59 epochs training was like this below

Thanks

[Owen-Liuyuxuan]

Could you please provide some scripts to reproduce the experiment?