Fine-tuning of ResNet32 for different task with less datapoints

ml4bio / RNA-FM

Nature Methods: RNA foundation model (together with RhoFold)

https://ml4bio.github.io/RNA-FM/

MIT License

215 stars 23 forks source link

Fine-tuning of ResNet32 for different task with less datapoints #4

Open guanjue opened 1 year ago

guanjue commented 1 year ago

Hi, I'm reaching out with a question related to the applicability of your models for tasks with a smaller dataset. Specifically, I'm interested in whether there are available scripts for fine-tuning the RNA-FM + ResNet model with few data points, similar to how the RNA-FM (TL) model was adjusted in the paper's RNA 3D closeness prediction task.

Any guidance or resources would be greatly appreciated.

mydkzgj commented 1 year ago

Hi, sorry for the late reply!

Sure. You can directly use the weight files we offered for RNA secondary structure prediction as the initial weight for your related tasks.

As for the scripts, you can check the tutorial for feature-based training. If you want to fine-tune your entire model, the weights of the backbone should also be optimized .