LLAMA-Tree-of-Thoughts Colab Notebook

This Colab notebook is a small experimentation based on the code from rhohndorf, and you can find the original repository here.

Requirements

To run this notebook, you will need the following:

• A Colab environment (either free or paid version)
• Internet connectivity

Installation

To get started, we need to install the llama-cpp-python library. Run the following command:

!pip install llama-cpp-python

Model Download

Next, we will download the LLAMA model file. The model file contains the pre-trained LLAMA reasoning model. Run the following code block:

import requests

url = "https://huggingface.co/TheBloke/Wizard-Vicuna-7B-Uncensored-GGML/resolve/main/Wizard-Vicuna-7B-Uncensored.ggmlv3.q4_1.bin"
filename = "modelFile.bin"

response = requests.get(url)
response.raise_for_status()

with open(filename, "wb") as file:
    file.write(response.content)

print("File downloaded successfully!")

LLAMA Settings

Before running the LLAMA experiment, we need to configure the LLAMA settings. Modify the settings according to your preferences in the following code block:

cfg_model_path = "/content/modelFile.bin"  # Path to the downloaded model file
cfg_temperature = 0.7  # Temperature parameter for LLAMA's text generation
cfg_context_size = 2048  # Context size for LLAMA's reasoning

cfg_branching_factor = 3  # Branching factor for LLAMA's planning phase
cfg_max_plan_length = 3  # Maximum plan length for LLAMA's planning phase

Running the LLAMA Experiment

To run the LLAMA experiment, execute the code block below:

from llama_cpp import Llama

llm = Llama(cfg_model_path, n_ctx=cfg_context_size)

# Define helper functions and solve the problem
# ...

# Main execution
# ...

Ensure that you have completed the necessary modifications in the code block above before running the experiment. Replace the placeholder goal with your desired problem statement:

goal = "Earn money online without finding clients"

The notebook will then proceed to find a plan that solves the given goal. The plan will be displayed as the output.

Enjoy experimenting with LLAMA-Tree-of-Thoughts in this Colab notebook!