Hi, thanks for your awesome work about contrastive learning in semantic segmentation. I notice that the `decoder_contrast.py` does not contain the code of dynamic sampling, can you provide the full code of your paper? :-)