Basic Usage

An mtenn.Model object is build using a config object, and then can be used as a standard pytorch model. More details on the config system are in the mtenn.combination docs, and more details on the expected model inputs are in Model. Below, we detail a basic example of building a default Graph Attention model and using it to make a prediction on a SMILES string.

from dgllife.utils import CanonicalAtomFeaturizer, SMILESToBigraph
from mtenn.config import GATModelConfig

# Build model with GAT defaults
model = GATModelConfig().build()

# Build graph from SMILES
smiles = "CCCC"
g = SMILESToBigraph(
    add_self_loop=True,
    node_featurizer=CanonicalAtomFeaturizer(),
)(smiles)

# Make a prediction
pred, _ = model({"g": g})