The Backend for Project Netherworld. For more information about the Project, see this homepage.
- Efficient Communication via Serialization of Tensors and Chat History(using FastAPI and Uvicorn) to Interface with the Phantasmagoria Front End.
- Implementation of Experimental Warpers such as top_a and Tail Free Sampling to allow more options for creatively generating text via the conversational agent (chatbot).
- Implementation of Experimental Processors such as Logit_bias(es) to customize their conversation agent to say certain words/phrases more or less often
- Quick and Efficient text generation via the Transformers and Pytorch libraries to leverage GPT-J/Neo/NeoX based Open Source Models such as ConvoGPT.
- Customizable Port Hosting.
- Have Python installed. This project requires Python 3.10 or higher.
- Either clone this repository, download it via zip, or download the release.
- Using your favorite CLI (command prompt, bash, etc.), use
cdto change the directory to where you downloaded the repository. - Run the following command:
pip install -r requirements.txt - Alternatively, should you prefer not using pip and want to use conda instead, run the following command:
conda install --file requirements.txt
There is not much to go over the backend, as the user all you are concerned with is setting it up, whether it be locally or on a virtual machine. If you are hosting non-locally, be sure that you have port forwarded a specific port to ping. Otherwise, this project will not work.
- To run the program, you will need to run the following command:
python backend_server.py. Some installations of Python 3 utilizepython3as a prefix to run python commands instead, so if this is the case, run the command as such instead:python3 backend_server.py”By default, this will make the backend server run on port 8000. - Optionally, you can set the port to a custom value simply by passing a number to the run command shown above. For example, you can run the following command to set the backend server to run on port 8080:
python backend_server.py 8080. - You will be aware that the server works if you see the following (or similar) text outputted in your CLI or console:
