README.md

---
base_model: westlake-repl/SaProt_650M_AF2
library_name: peft
---
# Base model: [westlake-repl/SaProt_650M_AF2](https://huggingface.co/westlake-repl/SaProt_650M_AF2)

# Model Card for Model ID

<!-- Provide a quick summary of what the model is/does. -->
This model is used to predict signal peptides on each site of amino acid sequences.

### Task type
Residue level clssification

### Dataset description
The dataset is from [SignalP 6.0 predicts all five types of signal
peptides using protein language models](https://www.nature.com/articles/s41587-021-01156-3).
This dataset contains 7 classes:

S (0): Sec/SPI signal peptide | T (1): Tat/SPI or Tat/SPII signal peptide | L (2): Sec/SPII signal peptide |
P (3): Sec/SPIII signal peptide | I (4): cytoplasm | M (5): transmembrane | O (6): extracellular
### Model input type
Amino acid sequence

### Performance
test_acc: 0.96

### LoRA config
lora_dropout: 0.0

lora_alpha: 16

target_modules: ["query", "key", "value", "intermediate.dense", "output.dense"]

modules_to_save: ["classifier"]

### Training config
class: AdamW

betas: (0.9, 0.98)

weight_decay: 0.01

learning rate: 1e-4

epoch: 10

batch size: 100

precision: 16-mixed