Hrrformer: A Neuro-symbolic Self-attention Model (ICML23)

A diabolic implementation of mcSAT 😈

本项目探索并实现超越传统 Transformer 架构的高效序列建模方法，重点关注状态空间模型（SSM）Mamba 和线性注意力机制等新型架构。项目基于Pytorch框架，从零设计实现了一套完整的模型训练、评估、记录和可视化方案，并完成 GLUE Benchmark 和 LRA 的适配工作。

LRA (XYZ to LRA Converter) is a tool that processes tube bending data and converts it into LRA format with visualization support. This Python-based project aims to streamline the manufacturing workflow with capabilities such as profile loading, data conversion, and animation.