innofacisteven's picture
Update README.md
97f07d2 verified
---
license: mit
---
# EDA Log Dataset for IC Debugging
## 📌 專案背景
這份資料集是 2026 智慧創新大賞 (Best AI Awards) 參賽專案 **Agentic-EDA Copilot** 的核心元件,
專門用來解決 IC 設計流程中,RTL 到 GDS 迭代所產生的大量非結構化 Log 解析難題。
## 📊 資料內容
涵蓋模擬 EDA 工具(如 OpenROAD)執行日誌,分為以下三大類別:
| 類別 | 說明 |
|------|------|
| **Timing Violations** | WNS(Worst Negative Slack)與 TNS 違規樣本 |
| **Physical Design Rules** | Multi-driven net、Latch inferred 等電路錯誤 |
| **Hybrid Metadata** | 經 Regex 解析後萃取的結構化欄位(Severity、Line Number、Module Name)|
## 🔗 相關連結
- **GitHub Repository**:[Agentic-EDA Copilot](https://github.com/stevenaiguy/Agentic-EDA-Copilot)
- **核心技術**:n8n Workflow 搭配 Agentic RAG 架構
## 🛠️ 使用方式
搭配專案內建的 **Hybrid Parser** 載入:
```python
from datasets import load_dataset
ds = load_dataset("innofacisteven/eda-log-dataset-for-ic-debugging")
```
載入後建議先執行 Regex 預處理管線,將原始 Log 結構化,再送入 LightRAG 進行語意檢索。
完整使用流程請參考 GitHub Repository 的 `docs/pipeline.md`