Team5_NLP_Upstage

[25-2 NLP] Term Project. Enhancing a QA performance of LLM (Solar Pro 2) using RAG and Prompt engineering.

Clone this repository to your local.

git clone https://github.com/juminsuh/Team5_NLP_Upstage.git

Please download faiss_vectorstore folder from google drive. (😂 It might take a few minutes due to it's large size (i.e., 1.1GB))
Unzip faiss_vectorstore.
Make sure to set the faiss_vectorstore directory such as ➡️ ./Team5_NLP_Upstage/faiss_vectorstore
Make .env file and write UPSTAGE_API_KEY="your_upstage_api_key" to .env file. (The directory of your .env file is ./Team5_NLP_Upstage/.env)

conda create -n team5 python=3.10 -y # create
conda activate team5 # activate

pip install -r requirements.txt

🥳 You are ready to run the code!

python run.py --data_path ./datasets/testset.csv

➡️ If you want to evaluate other testset, then just change the value of --data_path.

python run.py --data_path <your_testsets_directory>

✅ You can check total score by running the code below (You should modify the directory of score.py appropriately before you run the code):

python score.py

We utilized textbook and QA datasets from hugging face.

Navigation Menu