v0.1.0 - Initial public-ready release

Overview

Initial public-ready release of AI_RFIC_workflow, a research workflow for RFIC pixel-layout generation, Keysight ADS/RFPro EM automation, dataset construction, and CNN-based surrogate modeling.

Included in this release

• public-facing repository structure and documentation
• de-localized configuration examples
• validated parallel_version ADS/RFPro batch flow
• HDF5 dataset construction pipeline
• PyTorch training and verification pipeline
• clarified layout-template generation and GUI role in the workflow
• noncommercial public research licensing

Validated workflow

The following path has been runtime-validated on Windows with ADS/RFPro available:

1. validate ADS batch configuration
2. run the small parallel ADS flow
3. export Touchstone results
4. build HDF5 from generated .sNp files
5. train a CNN checkpoint
6. verify a trained model on stored samples

Notes

• parallel_version/ is the recommended execution line
• serial_version/ is retained as legacy/reference material
• the layout generator under Data_process/JSON_layout_create/ is the upstream JSON template generation stage
• large local assets such as trained checkpoints, HDF5 datasets, ADS workspaces, and simulation outputs are intentionally not tracked by Git

Environment model

This repository uses separate runtime contexts:

• a normal Python environment for orchestration
• ADS-bundled Python for ADS/RFPro worker execution
• a separate normal Python environment for HDF5 and PyTorch work

Known constraints

• Windows-focused workflow
• requires Keysight ADS / RFPro, valid licenses, and accessible PDK or reference technology libraries
• GUI utilities require a regular desktop Python with tkinter, not ADS internal Python

Recommended starting points

• README.md
• docs/core/SETUP.md
• docs/core/QUICKSTART.md
• docs/core/CONFIG_REFERENCE.md