jiosephlee/intern-s1-mini-assay-transfer-record-level-v27-bbb-martins-l2-step1575

TEXT GENERATIONPricing:Input $0.468 / Output $1.82Concurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 23, 2026Architecture:Transformer Featherless Exclusive Cold

The jiosephlee/intern-s1-mini-assay-transfer-record-level-v27-bbb-martins-l2-step1575 is a 8 billion parameter, full fine-tuned checkpoint derived from the Intern-S1-mini-lm base model, with a context length of 32768 tokens. It is specifically optimized for assay transfer tasks at the record level, demonstrating improved performance over vanilla Morgan methods in joint ID/OOD error and Spearman correlation. This model is designed for applications requiring precise chemical property prediction and transfer learning within drug discovery contexts.

Loading preview...

Model Overview

This model, intern-s1-mini-assay-transfer-record-level-v27-bbb-martins-l2-step1575, is a fine-tuned checkpoint of the Intern-S1-mini-lm base model. It was developed by jiosephlee through a specialized assay-transfer run, focusing on record-level data. The training involved 5 epochs with a batch size of 32 per GPU across 4 GPUs, using a learning rate of 2e-5.

Key Capabilities

  • Enhanced Assay Transfer: Specifically fine-tuned for transferring knowledge across assays at a record level.
  • Optimized Checkpoint: The model represents checkpoint step 1575, selected based on a validation selector value of 0.5000490216050504 (lower is better) for ranking_validation/overall/knn_id_ood_level_macro_mae_f1_error_at_3.
  • Improved Performance: Demonstrates superior performance compared to the vanilla Morgan method on key metrics:
    • Joint ID/OOD error @3: 0.5000 (Model) vs. 0.7055 (Vanilla Morgan)
    • Spearman correlation: 0.5007 (Model) vs. 0.0739 (Vanilla Morgan)
    • Top-1 hit @3: 0.2376 (Model) vs. 0.1881 (Vanilla Morgan)

Good For

  • Chemical Property Prediction: Ideal for tasks involving the prediction of chemical properties where assay transfer is crucial.
  • Drug Discovery Research: Suitable for researchers and developers in drug discovery needing to leverage existing assay data for new predictions.
  • Transfer Learning Applications: Excellent for scenarios requiring robust transfer learning capabilities in scientific domains.

Finetunes
1 models