Best way to finetune a classifier from EVO2 7b #181
Description
Description
We’d like to evaluate the model on downstream classification tasks (both Token and Sequence level). This requires adding a classification head and finetuning.
Context
The README notes that training/finetuning should use Savanna or BioNeMo. I haven’t found an example that adds a task head and runs supervised finetuning for classification/regression.
What I’m looking for:
- Recommended way to attach a classification (and/or regression) head to the base model using Savanna or BioNeMo
- Example scripts for supervised finetuning
Overall ,any help on this issue would be appreciated.