ericzhang1122's picture
Update README.md
79e0dbe verified
metadata
license: mit

Base model: westlake-repl/SaProt_650M_AF2

Task type: protein-level classification

Binary label represented by 1 for the binding and 0 for the non-binding.

Dataset: SaProtHub/Dataset-Metal_Ion_Binding

Model input type: SA(Structure-aware) sequence

Performance (on test set): 79.37% Accuracy

LoRA config:

  • r: 32
  • lora_dropout: 0
  • lora_alpha: 64
  • target_modules: ["query", "key", "value", "intermediate.dense", "output.dense"]
  • modules_to_save: ["classifier"]

Training config:

  • optimizer:
    • class: AdamW
    • betas: (0.9, 0.98)
    • weight_decay: 0.01
  • learning rate: 5e-4
  • epoch: 100
  • batch size: 64
  • precision: 16-mixed