search

ml4bio / e2efold

pytorch implementation for "RNA Secondary Structure Prediction By Learning Unrolled Algorithms"
MIT License
108 stars 17 forks source link

Question about the training phases #14

Closed L0-zhang closed 2 years ago

L0-zhang commented 2 years ago

Hi. Your model is fantastic,especially the post processing part.

Do you have a try whether it's better to save the weight file and test short(<600)and long (600~1800) seqs independantly?

Thank you very much.

liyu95 commented 2 years ago

Thank you very much for your interest!

I split the data into two parts mainly for save the memory and accelerate the program. In fact, one model for all the data also work well.

By the way, our group will release a new method very soon, which can outperform E2Efold by 25% on cross-dataset validation.

L0-zhang commented 2 years ago

Thank you very much.