---
license: apache-2.0
base_model: unsloth/Qwen2.5-Coder-14B
pipeline_tag: text-generation
library_name: peft
tags:
- code
- energy-efficient-code-generation
- green-software-engineering
- lora
---

# Qwen2.5-Coder-14B Runtime-SFT

Part of the **Green Tea** replication package for *Beyond the Need for Speed: Energy-Aware Code Generation via Simulation-Guided Reinforcement Learning* (Rajput and Sharma). Runtime-contrastive supervised fine-tuning (LoRA r=64). On the 143-problem held-out benchmark it achieves **7.64% CARET** (Correctness-Adjusted Reduction in Energy Total).

- Base model: `unsloth/Qwen2.5-Coder-14B`
- Training: Runtime-contrastive supervised fine-tuning (LoRA r=64)
- Code and full replication: https://github.com/SMART-Dal/green-tea
- Dataset (Zenodo): https://doi.org/10.5281/zenodo.21210099

## Usage
```python
from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer
b = AutoModelForCausalLM.from_pretrained('unsloth/Qwen2.5-Coder-14B', device_map='auto')
m = PeftModel.from_pretrained(b, 'saurabh-singh-rajput/green-tea-qwen2.5-coder-14b-runtime-sft')
t = AutoTokenizer.from_pretrained('saurabh-singh-rajput/green-tea-qwen2.5-coder-14b-runtime-sft')
```

## Citation
```bibtex
@misc{rajput2026greentea,
  title={Beyond the Need for Speed: Energy-Aware Code Generation via Simulation-Guided Reinforcement Learning},
  author={Rajput, Saurabhsingh and Sharma, Tushar},
  year={2026},
  note={Preprint}
}
```