SenseTime

/

deformable-detr-with-box-refine-two-stage

like 2

Follow

SenseTime 22

I'm trying to fine-tune the SenseTime/deformable-detr-with-box-refine-two-stage model using only a single feature level (num_feature_levels=1). However, during training, I encounter the following error during the forward pass:

RuntimeError: Given groups=1, weight of size [256, 2048, 1, 1], expected input[8, 512, 94, 167] to have 2048 channels, but got 512 channels instead

If I want to use only one feature level, besides setting num_feature_levels=1, are there other components (e.g., backbone output layer, input_proj, etc.) that also need to be manually modified?

Thanks