Skip to content

DIFFERENCES.md

Latest commit

 

History

History
19 lines (13 loc) · 1.31 KB

DIFFERENCES.md

File metadata and controls

19 lines (13 loc) · 1.31 KB

Differences from Original Decimer Packages

This document explains the changes made compared to the original DECIMER V2 and DECIMER Image Classifier packages.

Changes vs. Decimer Transformer Package

  • Mac Silicon Tensorflow support - Installation handles Mac M1/M2/M3 GPU acceleration via tensorflow-macos and tensorflow-metal (local install only, not Docker)
  • Server-based API - The server version uses JSON calls; the standalone version overrides the original predict_smiles()
  • Python API provided - Easy integration with existing Python projects
  • Note: You can still use the original predict_smiles() function by installing the decimer packages directly

Changes vs. Classifier Package

  • Adjusted threshold - Uses classifier score < 0.3 instead of true/false readout, allowing reaction schemes to be classified as structures
  • Reaction scheme support - The original classifier wasn't designed for reaction schemes, but this threshold adjustment makes it work
  • Modified keras calls - Changed to tf.keras instead of standalone keras
  • Removed image saving - Removed unnecessary save_image command

To adjust the threshold, edit IC_THRESHOLD in decimer_server.py.