Question about the Stage-1 K² / outline prediction model

[ambitious-rice]

Hi, thanks for releasing the code.

In the paper, Algorithm 1 and Fig. 4 describe a two-stage diffusion framework:

1. Stage 1: train DM1 to predict the K² / map-outline from environmental inputs.
2. Stage 2: use the predicted K² / map-outline as a condition to generate the final radio map.

I found caculate_k.py, which computes the K² / outline maps offline, and the DPMK / DPMCARK / IRT4K configs seem to use these computed K² maps as conditions for the final diffusion model.

Could you please point me to the code/config/checkpoint for training the Stage-1 model that predicts the K² / outline map itself?

Thanks!