Port-LLM

Yali Zhang, Haifan Yin, Weidong Li, Emil Björnson, Mérouane Debbah, "Port-LLM: A Port Prediction Method for Fluid Antenna based on Large Language Models"，available online: [paper].

Dependencies and Installation

• Python 3.9
• Pytorch 2.1.0
• NVIDIV GPU + CUDA 12.1
• Anaconda (conda 24.5.0)

Dataset

The dataset used in this code is derived from the CDL-D channel model. Comprehensive information regarding the parameter configurations for dataset generation is available in our paper.

Get started

• The code for model training is in the train, while the code utilized for testing the model across multiple antennas at the base station can be found in the Multiantenna_test. other_NN_based_models contains the other neural network-based models compared in our paper and plot_figs includes some data file processing and graphing code utilized in our research.
• The trained model obtained by utilizing the codes in train is employed for the purposes of performance evaluation and comparative analysis in Multiantenna_test and other_NN_based_models.
• Please be advised that when utilizing this code, it is essential to modify the file paths within the code to correspond with your specific file locations.
• The pre-trained GPT-2 model utilized in the code is available for download from the official website [Hugging Face].

Attention!

• To reduce the number of our model parameters, we have further refined the Port-LLM model described in the original paper. The code for the improved Port-LLM model, based on LoRA fine-tuning, is presented in Updata_code\model_lore.py.
• To more fully leverage the powerful natural language processing capabilities of large language models and further compare the effectiveness of prompt fine-tuning versus LoRA fine-tuning in the models studied here, we propose a prompt-based fine-tuning model named Prompt-Port-LLM in the updated version of our paper. The specific implementation code for this model can be found at: Updata_code\model_prompt.py.

Citation

If you find this repo helpful, please cite our paper.

@article{zhang2025portllm,
  title={Port-LLM: A Port Prediction Method for Fluid Antenna based on Large Language Models},
  author={Yali Zhang, Haifan Yin, Weidong Li, Emil Björnson, Mérouane Debbah},
  journal={arXiv preprint arXiv:2502.09857},
  year={2025}
}