Very brilliant job and great contribution to the community!
Given that Nicheformer is pretrained on a combination of single-cell and spatial data, I assume its imputation performance likely surpasses that of scGPT. Has the Nicheformer team explored this downstream task? If so, could you kindly provide any tutorials related to it? I'd like to apply this wonderful model :)
Thanks!