jiosephlee/intern-s1-mini-molecule-property-transfer-v1.2-bioavailability-ma-best-knn-mae5

TEXT GENERATIONPricing:Input $0.468 / Output $1.82Concurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 18, 2026Architecture:Transformer Featherless Exclusive Cold

The jiosephlee/intern-s1-mini-molecule-property-transfer-v1.2-bioavailability-ma-best-knn-mae5 model is a specialized 8-billion parameter model developed by jiosephlee, focused on molecule property transfer for bioavailability prediction. It was trained for 25 epochs with a degree cap of 16 on the jiosephlee/molecule-property-transfer-v1.2-bioavailability-ma-vote-mean-intern dataset. This model is optimized for predicting molecular bioavailability, achieving a held-out test KNN MAE@5 of 0.498376, making it suitable for cheminformatics and drug discovery applications.

Loading preview...

Model Overview

This model, developed by jiosephlee, is a specialized 8-billion parameter model designed for molecule property transfer, specifically targeting bioavailability prediction. It represents the best checkpoint from a 25-epoch training run, utilizing a degree-16 assay-transfer method.

Key Training Details

  • Dataset: Trained on jiosephlee/molecule-property-transfer-v1.2-bioavailability-ma-vote-mean-intern.
  • Training Configuration: Employed a degree cap of 16 (8 outgoing and 8 incoming pairs per molecule), trained for 25 epochs with a learning rate of 2e-5, batch size of 32, and gradient accumulation of 1.
  • Validation Metric: Selection was based on validation KNN MAE@5, achieving a best score of 0.604953.

Performance

  • The model demonstrated a held-out test KNN MAE@5 of 0.498376, indicating its effectiveness in predicting molecular bioavailability.

Use Cases

This model is particularly well-suited for:

  • Cheminformatics: Analyzing and predicting properties of chemical compounds.
  • Drug Discovery: Assisting in the early stages of drug development by predicting bioavailability, a crucial pharmacokinetic property.
  • Molecular Property Prediction: Transferring learned properties to new molecular structures for various applications.