/playLM

Primary LanguageJupyter Notebook

How strong is reasoning of LLMs in Video Games?

Requirements

  • Python

Installing Python dependencies

  1. (recommended) Create a virtual environment

  2. Install dependencies from requirements.txt file

    pip install -r requirements.txt

Getting the model

There are two models, LLama-based model fine-tuned for instruction tuning and a replication of OpenAI’s GPT-3

High-level API for LLama-based model

We are using a high-level API for the LLama-based model. The API is downloaded with all the dependencies when running the code for the first time. The reposority is located at https://github.com/abetlen/llama-cpp-python

Downloading the LLama-based models

Create a folder model in the root directory

For the wizardLM model

  1. Go to https://huggingface.co/TheBloke/wizardLM-7B-GGML/tree/main

  2. Download wizardLM-7B.ggmlv3.q4_1.bin model.

  3. Place the downloaded model in the model folder.

For the LLama model

  1. Go to https://huggingface.co/TheBloke/LLaMa-7B-GGML/tree/main

  2. Download llama-7b.ggmlv3.q4_1.bin model

  3. Place the downloaded model in the model folder.