Large Language Model Playground: experiment with Llama2, Falcon, GPT, and Flan chat LLM and tune LLM parameters
Model | Source | Parameters | Fine tuned for dialog task | Open-source | Model Link |
---|---|---|---|---|---|
GPT | OpenAI | 14.8 Billion | No | No | Click He |
llama-2-70b-chat | Meta | 70 Billion | Yes | Yes | Click He |
llama-2-13b-chat | Meta | 13 Billion | Yes | Yes | Click He |
llama-2-7b-chat | Meta | 7 Billion | Yes | Yes | Click He |
falcon-7b-instruct | TII | 7 Billion | Yes | Yes | Click He |
flan-ul2 | 20 Billion | Yes | Yes | Click He |
- Six LLM models: Change the used model
- Temperature range [0.1-1]: Control the creativity of the answer
- Output length range [1-1000]: Control the output's maximum number of tokens
- Interacting with conversational agents via simple UI
- Fast load and delete models for assessing the LLM models' performance on specific
- API keys for all models are provided except for GPT model (provided by users)
For live demo visit https://llm-playground-etp4.onrender.com/
LLM_Demo.mp4
Make sure all libraries included in the requirements.txt
file and their dependencies are installed using pip or conda command on your virtual environment.
pip install langchain==0.0.345
pip install python-dotenv==1.0.0
pip install streamlit==1.29.0
pip install replicate==0.20.0
pip install huggingface_hub==0.19.4
pip instal
https://github.com/shaimaaK/LLM-playground/assets/54285485/e18bbeaa-4b81-4f1e-84e9-2c100ff64491
l openai==1.3.5
To run the application, open the terminal in the root directory and execute the following command
streamlit run frontend.py