Training setup details for EventSeqFormer?

[Yutaro-Aze]

Hi, thanks a lot for the great work and the release of the MouseSIS dataset and code — it's super helpful.

I’ve been experimenting with the EventSeqFormer model mentioned in the paper and was wondering if you could share more details about the training setup. Specifically, I’m trying to figure out:

• train epochs
• lr, lr drop
• batch size
• number of frames per input during training
• backbone used in Seqformer

I couldn’t find these details in the paper or codebase (apologies if I missed them). Any info would be greatly appreciated — even rough numbers or a config pointer would help!

Thanks again!