DingHing

The trend of the loss curve in the BYOL method is always decreasing, or does it decrease first and then increase? Because as I understand it, initially the target_network and...

DeepLabV3+, the backbone can only use Xception, and other backbones cannot be selected

1

DeepLabV3+, the backbone can only use Xception, and other backbones cannot be selected 

代码中实现的，是不是多卡并行时，每个GPU的queue独立更新？，我感觉应该像moco一样，先keys = concat_all_gather(keys)，再统一进行更新，是不是更合理一点？

I would like to raise a GitHub issue to ask when the LangMem section will be added.