罗崚骁 (Luo, Lingxiao)

> once the discriminator kicks in everything goes crazy, does somebody have an advice? Perhaps because the discriminator is not trained at all until the `g_loss` is introduced to the...

@Yuxin-CV Hello, I have a question related to the application of EVA-02 for semantic segmentation. Since ViT-Adapter is not used, does this imply that all feature maps received by the...

`Box` 或许应该理解为 C++ 中的 `std::unique_ptr` 更合适？

@mingxin-zheng Yes, maybe we can let the user decide whether to filter and what suffix to filter.

Thanks, I feel like using a regex pattern is a good idea. On the other hand, reading a list of files may not work as expected because they will be...

@teituroli Hello, I'm also interested in this problem. May I ask have you tried that solution? If so, could you please share your results here? Thank you very much!

Same here, really looking forward to a way that could "inherit" from another config file and modify the arguments, or even compose multiple config files inside a single config file...

Seemingly related to https://github.com/omni-us/jsonargparse/issues/221.

> `RandCropByPosNegLabelD` needs `label_key='seg'` to be up to date I agree. Fortunately, such crop operation is usually performed in early stage, e.g., for nnU-Net, after resampling to a target spacing...

> 1. The transform copies and then internally resamples the data to be up to date, creating a temporary tensor for the results > 2. The transform performs its calculations...