metadata
language:
- en
license: cc-by-nc-4.0
base_model:
- google/flan-t5-large
EPlus-LLM
Natural Language Interface for Automated Building Energy Modeling via LLMs
A prototype project exploring the use of fine-tuned large language models to automate building energy modeling from natural language input.
🎉 News
- ⚡️ [2025/01/01]: A prompting-based method for auto-building energy modeling has been released. Paper here.
- 🔥 [2024/05/016]: We first successfully implement natural language-based auto-building modeling by fine-tuning a large language model (LLM). Paper here.
🚀 Key Features
- Scalability: Auto-generates EnergyPlus models, including varying geometry sizes and internal loads.
- Accuracy & Efficiency: Achieves 100% modeling accuracy while reducing manual modeling time by over 95%.
- Interaction & Automation: A user-friendly human-AI interface for seamless model creation and customization.
🏗️ Target Users
This current platform is designed for engineers, architects, and researchers working in building performance, sustainability, and resilience. It is especially useful during early-stage conceptual design when modeling decisions have the greatest impact.
🚀 Quick Start
Here provides a code snippet to show you how to load the EPlus-LLM and auto-generate building energy models.
import torch
from transformers import (
AutoModelForSeq2SeqLM,
AutoTokenizer,
)
tokenizer = AutoTokenizer.from_pretrained("google/flan-t5-large")
model = AutoModelForSeq2SeqLM.from_pretrained("EPlus-LLM/EPlus-LLMv1")
generation_config = model.generation_config
generation_config.max_new_tokens = 1300
generation_config.temperature = 0.1
generation_config.top_p = 0.1
generation_config.num_return_sequences = 1
generation_config.pad_token_id = tokenizer.eos_token_id
generation_config.eos_token_id = tokenizer.eos_token_id
input="<Your input, description of the desired building.>"
input_ids = tokenizer(input, return_tensors="pt", truncation=False).to(device)
generated_ids = model.generate(input_ids = input_ids.input_ids,
attention_mask = input_ids.attention_mask,
generation_config = generation_config)
generated_output = tokenizer.decode(generated_ids[0], skip_special_tokens=True)
generated_output = new_tokens.replace("_", " ")
generated_ooutput = new_tokens.replace("|", "\n")
print(generated_output)
📝 Citation
If you find our work helpful, feel free to give us a cite.
@article{jiang2025EPlus-LLM,
author = {Gang Jiang and Zhihao Ma and Liang Zhang and Jianli Chen},
title = {Prompt engineering to inform large language models in automated building energy modeling},
journal = {Applied Energy},
volume = {367},
pages = {123431},
year = {2024},
month = {Aug},
doi = {https://doi.org/10.1016/j.apenergy.2024.123431}
}
@article{jiang2025prompting,
author = {Gang Jiang and Zhihao Ma and Liang Zhang and Jianli Chen},
title = {Prompt engineering to inform large language models in automated building energy modeling},
journal = {Energy},
volume = {316},
pages = {134548},
year = {2025},
month = {Feb},
doi = {https://doi.org/10.1016/j.energy.2025.134548}
}