Saved in:
Bibliographic Details
Main Author: amiyax
Format: Recurso digital
Language:
Published: Zenodo 2026
Online Access:https://doi.org/10.5281/zenodo.18899990
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866901060776361984
author amiyax
author_facet amiyax
contents <h2>OptoFlow-ML v1.0.0</h2> <p>An MLOps-driven platform for automated high-throughput prediction of organic π-conjugated molecular properties, implementing the hierarchical machine learning framework described in Bhat et al. (2022, <em>Chem. Sci.</em>, DOI: 10.1039/d2sc04676h).</p> <h3>Key Features</h3> <ul> <li><strong>Production-grade model versioning</strong>: Timestamp-based immutable snapshots with provenance tracking via <code>metadata.json</code></li> <li><strong>One-command reproducibility</strong>: Deterministic dependency resolution through <code>uv</code> package manager</li> <li><strong>Automated batch training</strong>: Single CLI invocation trains all 15 properties across multiple model generations</li> <li><strong>Architecture-aware routing</strong>: Dynamic model selection based on <code>model_name</code> keywords</li> <li><strong>Uncertainty quantification</strong>: Evidential regression outputs calibrated variance via Normal-Inverse-Gamma posteriors</li> </ul> <h3>Supported Properties</h3> <p>15 electronic, redox, and optical properties including HOMO, LUMO, ionization potential, electron affinity, reorganization energies, and absorption/emission wavelengths.</p> <h3>Model Hierarchy</h3> <ul> <li><strong>2nd Generation</strong>: MLP baseline with RDKit descriptors + Morgan fingerprints</li> <li><strong>3rd Generation</strong>: MPNN with Set2Set readout on molecular graphs</li> <li><strong>4th Generation</strong>: MPNN + RDKit hybrid features</li> <li><strong>4th Generation Evidential</strong>: MPNN with uncertainty quantification</li> </ul> <h3>Technical Stack</h3> <ul> <li>Python 3.10 (DGL compatibility requirement)</li> <li>PyTorch 2.0.1 + DGL 1.1.2 + DGLLife 0.3.2</li> <li>RDKit 2025.9.1 for molecular featurization</li> <li><code>uv</code> for deterministic environment management</li> </ul> <h3>Documentation</h3> <p>Complete bilingual documentation (Chinese/English) included:</p> <ul> <li>Architecture design (<code>docs/design/ARCHITECTURE.md</code>)</li> <li>Setup guide (<code>docs/guide/SETUP.md</code>)</li> <li>Training workflows (<code>docs/guide/TRAINING.md</code>)</li> <li>Prediction interfaces (<code>docs/guide/PREDICTION.md</code>)</li> <li>Versioning system (<code>docs/guide/VERSIONING.md</code>)</li> </ul> <h3>Citation</h3> <p>If you use this platform in your research, please cite the original algorithmic work:</p> <blockquote> <p>Bhat, V.; Sornberger, P.; Pokuri, B. S. S.; Duke, R.; Ganapathysubramanian, B.; Risko, C.<br> <em>Electronic, Redox, and Optical Property Prediction of Organic π-Conjugated Molecules through a Hierarchy of Machine Learning Approaches.</em><br> Chemical Science, 2022. DOI: 10.1039/d2sc04676h</p> </blockquote> <h3>License</h3> <p>MIT License</p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_18899990
institution Zenodo
language
publishDate 2026
publisher Zenodo
record_format zenodo
spellingShingle flashtenser/OptoFlow-ML: v1.0.0 - OptoFlow-ML: MLOps Platform for Organic Photoelectronic Property Prediction
amiyax
<h2>OptoFlow-ML v1.0.0</h2> <p>An MLOps-driven platform for automated high-throughput prediction of organic π-conjugated molecular properties, implementing the hierarchical machine learning framework described in Bhat et al. (2022, <em>Chem. Sci.</em>, DOI: 10.1039/d2sc04676h).</p> <h3>Key Features</h3> <ul> <li><strong>Production-grade model versioning</strong>: Timestamp-based immutable snapshots with provenance tracking via <code>metadata.json</code></li> <li><strong>One-command reproducibility</strong>: Deterministic dependency resolution through <code>uv</code> package manager</li> <li><strong>Automated batch training</strong>: Single CLI invocation trains all 15 properties across multiple model generations</li> <li><strong>Architecture-aware routing</strong>: Dynamic model selection based on <code>model_name</code> keywords</li> <li><strong>Uncertainty quantification</strong>: Evidential regression outputs calibrated variance via Normal-Inverse-Gamma posteriors</li> </ul> <h3>Supported Properties</h3> <p>15 electronic, redox, and optical properties including HOMO, LUMO, ionization potential, electron affinity, reorganization energies, and absorption/emission wavelengths.</p> <h3>Model Hierarchy</h3> <ul> <li><strong>2nd Generation</strong>: MLP baseline with RDKit descriptors + Morgan fingerprints</li> <li><strong>3rd Generation</strong>: MPNN with Set2Set readout on molecular graphs</li> <li><strong>4th Generation</strong>: MPNN + RDKit hybrid features</li> <li><strong>4th Generation Evidential</strong>: MPNN with uncertainty quantification</li> </ul> <h3>Technical Stack</h3> <ul> <li>Python 3.10 (DGL compatibility requirement)</li> <li>PyTorch 2.0.1 + DGL 1.1.2 + DGLLife 0.3.2</li> <li>RDKit 2025.9.1 for molecular featurization</li> <li><code>uv</code> for deterministic environment management</li> </ul> <h3>Documentation</h3> <p>Complete bilingual documentation (Chinese/English) included:</p> <ul> <li>Architecture design (<code>docs/design/ARCHITECTURE.md</code>)</li> <li>Setup guide (<code>docs/guide/SETUP.md</code>)</li> <li>Training workflows (<code>docs/guide/TRAINING.md</code>)</li> <li>Prediction interfaces (<code>docs/guide/PREDICTION.md</code>)</li> <li>Versioning system (<code>docs/guide/VERSIONING.md</code>)</li> </ul> <h3>Citation</h3> <p>If you use this platform in your research, please cite the original algorithmic work:</p> <blockquote> <p>Bhat, V.; Sornberger, P.; Pokuri, B. S. S.; Duke, R.; Ganapathysubramanian, B.; Risko, C.<br> <em>Electronic, Redox, and Optical Property Prediction of Organic π-Conjugated Molecules through a Hierarchy of Machine Learning Approaches.</em><br> Chemical Science, 2022. DOI: 10.1039/d2sc04676h</p> </blockquote> <h3>License</h3> <p>MIT License</p>
title flashtenser/OptoFlow-ML: v1.0.0 - OptoFlow-ML: MLOps Platform for Organic Photoelectronic Property Prediction
url https://doi.org/10.5281/zenodo.18899990