metadata
license: mit
Base model: westlake-repl/SaProt_650M_AF2
Task type: protein-level classification
Label meaning: Label corresponds to the structural class of given protein and includes 3,042 categories. This task is built on the SCOP database.
Dataset: SaProtHub/Dataset-Structural_Class-ProteinShake
Model input type: SA(Structure-aware) sequence
Performance (on test set): 80.16% Accuracy
LoRA config:
- r: 8
- lora_dropout: 0
- lora_alpha: 16
- 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