- Clone the repository
git clone https://github.com/YUECHE77/Syllabus-Agent.git
- Install Package
conda create -n agent python=3.10 -y
conda activate agent
pip install -r requirements.txt
You need to download the following models from huggingface:
- google-bert/bert-base-uncased
- BAAI/bge-base-en-v1.5
- BAAI/bge-reranker-base
- NousResearch/Hermes-2-Pro-Llama-3-8B (Not required)
Also, our trained BERT model: BERT model weight
You can also find our datasets, experiments results, as well as the processed syllabus (already contained in this repo) from the link above.
It's also necessary to apply an API key from togetherAI: TogetherAI API
Recommend: Put the API key in .env file for safety.
Then, change the model path in python scripts.
We provide:
demo.py
to run single-turn inference. Usepython demo.py
multi_turn_demo.py
to run multi-turn conversation in the command line. Usepython multi_turn_demo.py
user_interface.py
as our final and ultimate version of work. You are able to interact with our agent in a well-designed UI (by us). Usepython user_interface.py
Additionally, we also have localized version and multiple-functions version.
Example:
- Use GLM, GPT, LLaMA, and Qwen as agent: ./agents/
- Our binary classifier using BERT and customer model. Also the code to construct dataset: ./binary_classifier/
- Evaluate the performance of each component: ./evaluations/
- Useful functions: ./utilities/