Seungmin0825/AnalogToBi-checkpoint

AnalogToBi

AnalogToBi: Device-Level Analog Circuit Topology Generation via Bipartite Graph and Grammar-Guided Decoding

AnalogToBi is a framework for automatic generation of device-level analog circuit topologies. It trains a compact decoder-only Transformer (11.3M parameters) from scratch and generates electrically valid, novel circuit topologies.

Key Features

• Circuit type conditioning: Divide datasets as 15 circuit categories (OpAmp, LDO, Comparator, etc.)
• Device renaming augmentation: Randomizes device numbering to prevent memorization while preserving topology
• Bipartite graph representation: Decouples devices and nets into distinct node types for compact structural description
• Grammar-guided decoding: State machine-based constrained decoding enforces electrical validity during generation

---

Paper

AnalogToBi: Device-Level Analog Circuit Topology Generation via Bipartite Graph and Grammar Guided Decoding

arXiv: https://arxiv.org/abs/2603.08720

---

Code

Official implementation:
https://github.com/Seungmin0825/AnalogToBi

---

Model Weights

This repository provides pretrained model checkpoints for AnalogToBi.