AUCOHL/RTL-Repo

RTL-Repo: A Benchmark for Evaluating LLMs on Large-Scale RTL Design Projects - IEEE LAD'24

29
/ 100
Experimental

This benchmark helps hardware design engineers evaluate how well large language models (LLMs) can generate Verilog code completions within complex, multi-file projects. It takes a trained LLM and a dataset of real-world Verilog code samples as input, then outputs metrics like 'Edit Similarity' and 'Exact Match' to show the LLM's performance. The ideal end-user is a hardware design engineer or an LLM researcher focused on hardware description languages.

No commits in the last 6 months.

Use this if you need to assess how accurately an LLM can generate logically consistent and syntactically correct Verilog code within the context of large digital design projects.

Not ideal if you are looking for an LLM to generate entire RTL designs from high-level specifications, as this focuses on code completion within existing projects.

hardware-design Verilog-development RTL-design digital-logic LLM-evaluation
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 7 / 25
Maturity 16 / 25
Community 6 / 25

How are scores calculated?

Stars

34

Forks

2

Language

Python

License

Apache-2.0

Last pushed

Jun 05, 2024

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/transformers/AUCOHL/RTL-Repo"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.