README.md

---
tags:
- chemistry
- cheminformatics
- materials-science
- astra
- transformers
- bert
- property-prediction
- regression
license: mit
---

# ASTRA: Initial LUMO Predictor

This model is part of the **ASTRA (Advanced Solvation Transformer for Rational Additives)** framework. 
It is a BERT-based regression model fine-tuned from the pre-trained `astra-bert-mlm`. This specific model is designed to predict the **Binding Energy (Eb)** of given additive molecules based on their SMILES. 

This is the "Initial" version of the predictor, meaning it was trained on the baseline dataset before the ASTRA Active Learning loop began.

## Model Details
- **Architecture:** BERT (Sequence Classification / Regression)
- **Stage:** Initial (Before Active Learning)
- **Task:** Property Prediction (Eb)
- **Base Model:** `wkdghdus23/astra-bert-mlm`

## Usage
```python
from astra.tokenizer import initial_bert_tokenizer_with_vocabulary
from astra.model import BertForDownstream, EmbeddingTunedModel

# Load the tokenizer and model
vocab_file = "./vocab.txt"

target_name = ["Eb"]

tokenizer = initial_bert_tokenizer_with_vocabulary(path=vocab_file)
model = EmbeddingTunedModel.from_pretrained_embtune_model(pretrained_path=pretrained_model_path,
		tokenizer=tokenizer,
		target_name=target_name)
```

## More Information
For more details on data preparation, downstream fine-tuning, and the full active learning loop, please visit our [GitHub Repository](https://github.com/wkdghdus23/astra).