Training

[bear-rabbit]

Hello, I have used your semantic mapping solution, starting with pairing to create data. However, the F1 score on my first training attempt was 0.9. May I know what could be the reason? I have replaced the original clustering method.

[kxhit]

Hi, could you give more details? I do not fully understand.

[bear-rabbit]

I used the semantic graph generation method mentioned in data_process/readme.md to regenerate the semantic graph and pairing from the kiiti sequence, and trained the model through main_sg.py. But I can get the effect of F1max=0.92 only by epoch=1. Why is this?

[bear-rabbit]