Fluorescence Landscape Prediction
Collection
This group contains all relative training datasets and model checkpoints trained on these datasets for Fluorescence Landscape Prediction. • 3 items • Updated
Base model: westlake-repl/SaProt_650M_AF2
Task type: protein-level regression
Label is the log-fluorescence intensity.
Dataset: SaProtHub/Dataset-Fluorescence-TAPE
Model input type: Amino acid sequence
Performance (on test set): 0.697 Spearman's ρ
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: 1e-4
- epoch: 50
- batch size: 64
- precision: 16-mixed
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support