search

facebookresearch / detectron2

Detectron2 is a platform for object detection, segmentation and other visual recognition tasks.
https://detectron2.readthedocs.io/en/latest/
Apache License 2.0
29.3k stars 7.32k forks source link

The ce_loss became negative when I was using the mask2former to do instance segmentation #5300

Open AIzealotwu opened 4 weeks ago

AIzealotwu commented 4 weeks ago

command you run: python train_net.py --num-gpus 8 --config-file configs/coco/instance-segmentation/swi n/maskformer2_swin_large_IN21k_384_bs16_100ep.yaml MODEL.WEIGHTS "weig hts/model.pkl"

The data form used in my case is coco. And the registration of this dataset was strictly followed with the document of our office

github-actions[bot] commented 4 weeks ago

You've chosen to report an unexpected problem or bug. Unless you already know the root cause of it, please include details about it by filling the issue template. The following information is missing: "Instructions To Reproduce the Issue and Full Logs"; "Your Environment";

Programmer-RD-AI commented 3 weeks ago

Hi, To resolve the issue you are facing, it is usually a case in the training step.. Try and change the LR (learning rate) and adjust your training process and check if it fixes the issue. Thank you

AIzealotwu commented 2 weeks ago

Well, Thanks. Maybe this issue was caused by so large of my picture. The learning rate was followed with the default in the mask2former source paper.

Programmer-RD-AI commented 2 weeks ago

Hi, Glad you found it useful! :) Try adjusting parameters related to image resolution, batch size, augmentation techniques, or even model architecture to fine-tune your instance segmentation setup. If you need further guidance or have specific questions, feel free to ask!