#12 已经帮忙给出了输出的P-E的设置。不知道类似的设置能不能输出SLP，如果不能设置输出SLP辨别台风，应该可以换个思路吧：用最底层输出的u,v风来辨别。
Please refer to #12 for P_E. If we can't set output SLP like #12, maybe we could use u,v at the lowest layer to detect cyclones.

I didn't figure out how to output both SLP and P_E (or precipitation) with the method in #12 . Any thoughts?

I don't know. @yaniyuval Thank you for helping me before again and again. Would you please give more detailed instructions to set "new_model_config_str"? Which variables could be set to output? Thank you.

Hi,
Currently it is a bit tricky to output both.
I will try to make some changes so it would be possible in the near future.

Hi, Currently it is a bit tricky to output both. I will try to make some changes so it would be possible in the near future.

Looking forward to the new features!

@yaniyuval Thank you! And you authors once mentioned to separate the dynamical part tendencies and learned tendencies as in your first NeuralGCM paper on Nature. Would you please show us how does it work now?

@yaniyuval Thank you! And you authors once mentioned to separate the dynamical part tendencies and learned tendencies as in your first NeuralGCM paper on Nature. Would you please show us how does it work now?

Is it related to the state.memory? I'm not sure if "memory" refers to tendency here.