How strong is reasoning of LLMs in Video Games?
Requirements
- Python
Installing Python dependencies
-
(recommended) Create a virtual environment
-
Install dependencies from
requirements.txt
filepip 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
-
Go to https://huggingface.co/TheBloke/wizardLM-7B-GGML/tree/main
-
Download
wizardLM-7B.ggmlv3.q4_1.bin
model. -
Place the downloaded model in the
model
folder.
For the LLama model
-
Go to https://huggingface.co/TheBloke/LLaMa-7B-GGML/tree/main
-
Download
llama-7b.ggmlv3.q4_1.bin
model -
Place the downloaded model in the
model
folder.