SaProtHub/Model-Subcellular_Localization-650M

Base model: westlake-repl/SaProt_650M_AF2

Task type: protein-level classification

The digital label means:

0: Nucleus

1: Cytoplasm

2: Extracellular

3: Mitochondrion

4: Cell.membrane

5: Endoplasmic.reticulum

6: Plastid

7: Golgi.apparatus

8: Lysosome/Vacuole

9: Peroxisome

Dataset: SaProtHub/Dataset-Subcellular_Localization-DeepLoc

Model input type: SA(Structure-aware) sequence

Performance (on test set): 85.75% Accuracy

LoRA config:

• r: 16
• lora_dropout: 0
• lora_alpha: 32
• 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