search

zylo117 / Yet-Another-EfficientDet-Pytorch

The pytorch re-implement of the official efficientdet with SOTA performance in real time and pretrained weights.
GNU Lesser General Public License v3.0
5.2k stars 1.27k forks source link

Not able to do inference after training #711

Open NamanMakkar opened 2 years ago

NamanMakkar commented 2 years ago

Hello, I trained the Efficientdet model on the FLIR dataset for 12 epochs and for some reason after the training is complete, the model is not able to do inference. When I try to run inference, all the elements of the 'classification' tensor are zeros due to which the 'preds' are all empty. I don't run into this issue if the model isn't trained or is trained only for 1-2 epochs. I checked my model architecture and the data, everything is fine but the model simply isn't able to run inference after training for 9-12 epochs it can only do so if trained for 1-2 epochs. Can someone suggest me what to do ?

zylo117 commented 2 years ago

maybe you just have to train longer