README.md

LawBot - Model

💻 Getting Started

⚠️ How To install Requirements

Cuda install

1. Run the following code on your terminal.

wget https://developer.download.nvidia.com/compute/cuda/11.8.0/local_installers/cuda_11.8.0_520.61.05_linux.run
chmod +x cuda_11.8.0_520.61.05_linux.run
sh cuda_11.8.0_520.61.05_linux.run

2. Input accept to proceed.

3. Select the driver and install.

4. Run the following commands on your terminal.

$ export PATH=/usr/local/cuda-11.8/bin:$PATH
$ export LD_LIBRARY_PATH=/usr/local/cuda-11.8/lib64:$LD_LIBRARY_PATH
$ pip install -r requirements.txt

⌨️ How To Train

LLM (Large Language Model)

• Before training, place Legal QA data at following directory
  model/LLM/train/data
model/LLM/train/val_data
• HuggingFace Write Token should be filled at line #38 of the following file
• Run the following command on your terminal

$ python3 model/LLM/train/train.py

Question Filterering Model (Koelectra)

$ python3 model/Filter/train.py

⌨️ How To Infer

LLM (Large Language Model)

• peft model id should be changed after training at line #27 of the following file
• Run the following command on your terminal

$ python3 model/LLM/inference/infer.py

Sentence BERT Retrieval

• Before training, place Legal QA data at following directory
  model/Retrieval/bert_retrieval/data
  
• Run the following command on your terminal

$ python3 model/Retrieval/bert_retrieval/retrieval_main.py

BM25 Retrieval

• Before training, place Legal QA data at following directory
  model/Retrieval/bm25_retrieval/all_data
  
• Run the following command on your terminal

$ python3 model/Retrieval/bm25_retrieval/retrieval_main.py

Question Filtering Model (Koelectra)

• Run the following command on your terminal

$ python3 model/Filter/infer.py

⌨️ How To Evaluate

LLM (Large Language Model)

• model name and use parmater should be changed if needed
• Run the following command on your terminal

$ python3 model/LLM/evaluation/evaluate_metrics.py