ASTRA: Updated LUMO Predictor (Final)

This model is part of the ASTRA (Advanced Solvation Transformer for Rational Additives) framework. It is an updated BERT-based regression model designed to predict the Lowest Unoccupied Molecular Orbital (LUMO) levels of additive molecules on thier solvation structure from SMILES.

Unlike the initial predictor, this "Final" version has been re-trained and updated using the datasets generated during the ASTRA Active Learning loop. Consequently, it exhibits improved predictive accuracy and robustness, particularly within the targeted chemical space of high-reward electrolyte additives.

Model Details

• Architecture: BERT (Sequence Classification / Regression)
• Stage: Final (After Active Learning Loop)
• Task: Property Prediction (LUMO)
• Base Model: Updated from wkdghdus23/astra-predictor-initial-lumo

Usage

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 = ["LUMO"]

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.