Request for Checkpoints for Each Downstream Task (e.g., Perturbation)

[JunyanReplicant]

Hello, and thank you for making this project available!

I’m currently working through your code and running the provided tutorials, including the perturbation example. If I understand correctly, after loading the scGPT parameters with load_pretrained, the next step involves further training and fine-tuning these parameters on the perturbation (or other downstream task) dataset.

Since I am testing this on a local machine with limited computing resources (just to confirm everything works before moving to more powerful servers), having the checkpoints for each fine-tuning example would be incredibly helpful. That way, I could skip the training/fine-tuning step (since I'm using the tutorial sample dataset) and directly continue following along with the tutorial code (to see if I can obtain the corresponding results).

If it’s possible, could you please share the saved checkpoints resulting from the fine-tuning process for each example? Your assistance would greatly streamline my workflow and is much appreciated.

Thank you very much for your time and support!