CAR-T Responder: attention-MIL weights
Inference artifacts for the CAR-T Responder demo in the Immune Research Space (https://huggingface.co/spaces/AdamSion74/immune-research).
model.npz: gated attention-MIL weights (numpy arrays)pca.pkl: fitted scikit-learn PCA (2,000 genes → 50 PCs), scikit-learn 1.9.1genes.txt: the 2,000 input genes, in ordermodel_meta.json: layer sizes
Trained on GEO GSE267097 (pre-treatment blood scRNA-seq, CD19 CAR-T, 55 patients). For research use only.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support