Open nkempynck opened 1 week ago
It shouldn't be necessary to change anything else besides the n_tasks
parameter. If the model doesn't learn well, you might consider increasing its capacity by increasing the number of layers / filters.
Alright thanks I will try that out.
Hi
I want to train a multitask model from scratch for multiple scATAC-tracks. I just followed the 3_train notebook, and updated the notebook at the relevant parts to take in these multiple tasks instead of 1 task. Are there any other things you would recommend changing? Do I still use the same loss function?
Thanks Niklas