Update README.md

README.md

@@ -1,3 +1,14 @@
+---
+title: MultiTaskTox
+emoji: 📈
+colorFrom: purple
+colorTo: blue
+sdk: docker
+pinned: false
+license: apache-2.0
+short_description: MultiTaskTox is a two-stage Gradient Boosting workflow
+---
+
 # MultiTaskTox – LightGBM Fingerprint Classifier for Tox21
 
 MultiTaskTox is a two-stage Gradient Boosting workflow purpose-built for the [Tox21](https://huggingface.co/datasets/ml-jku/tox21) benchmark. It ingests molecular SMILES strings, converts them into high-dimensional fingerprints (ECFP or MAP4), and trains a set of LightGBM classifiers that leverage cross-task signal to improve toxicity prediction across all 12 Tox21 targets.
@@ -112,4 +123,4 @@ The function:
 - MAP4 fingerprints are more expensive to compute; enable the cache directory to avoid recomputation across runs.
 - Use the saved metrics files to compare stage-one vs. stage-two AUCs and to trace which configuration produced a set of checkpoints.
 
-Happy modeling! If you extend MultiTaskTox (new fingerprints, alternative learners, etc.), keep the `predict(smiles)` contract intact so your Space remains leaderboard compatible.
+Happy modeling! If you extend MultiTaskTox (new fingerprints, alternative learners, etc.), keep the `predict(smiles)` contract intact so your Space remains leaderboard compatible.