WARNING:Shape of head.cls_preds.0.weight in checkpoint is torch.Size([80, 96, 1, 1]), while shape of head.cls_preds.0.weight in model is torch.Size([2, 96, 1, 1]).

Megvii-BaseDetection / YOLOX

YOLOX is a high-performance anchor-free YOLO, exceeding yolov3~v5 with MegEngine, ONNX, TensorRT, ncnn, and OpenVINO supported. Documentation: https://yolox.readthedocs.io/

Apache License 2.0

9.48k stars 2.21k forks source link

WARNING:Shape of head.cls_preds.0.weight in checkpoint is torch.Size([80, 96, 1, 1]), while shape of head.cls_preds.0.weight in model is torch.Size([2, 96, 1, 1]). #233

Closed haileykwak closed 3 years ago

haileykwak commented 3 years ago

Hi Im training my custom datasets in voc format. training well now but have warning at start.

could this warning have bad result? or just ignore it?

my command and waning below, $ python3 tools/train.py -f exps/example/yolox_voc/yolox_voc_tiny.py -d 1 -b 64 --fp16 -o -c yolox_tiny.pth.tar

F0xZz commented 3 years ago

you use the custom datasets you should change the code in the yolox/data/datasets/voc.py data/datasets/voc_classes.py yolox/exp/yolox_base.py

haileykwak commented 3 years ago

I changed the code in voc_classes.py , yolox_base.py. which portion of the code should be changed in voc.py?

GOATmessi8 commented 3 years ago

Hi Im training my custom datasets in voc format. training well now but have warning at start.

could this warning have bad result? or just ignore it?

my command and waning below, $ python3 tools/train.py -f exps/example/yolox_voc/yolox_voc_tiny.py -d 1 -b 64 --fp16 -o -c yolox_tiny.pth.tar

It's normal, as the coco pretrained weight has 80 classes but voc only has 20