---
library_name: transformers
tags:
- chess
- llama
- ChessLlama
- chess-engines
license: apache-2.0
datasets:
- Q-bert/Elite-Chess-Games
---

# ChessLlama


![image/png](https://cdn-uploads.huggingface.co/production/uploads/63da3d7ae697e5898cb86854/Px2blZin1iA_GT8nPah4J.png)

Generated by DALL-E 3.



## Model Details

This pre-trained model has been trained on the Llama architecture with the games of grand master chess players.

### Model Description

- **Developed by:** [Talha Rüzgar Akkuş](https://www.linkedin.com/in/talha-r%C3%BCzgar-akku%C5%9F-1b5457264/)
- **Data Format:** [Universal Chess Interface (UCI)](https://en.wikipedia.org/wiki/Universal_Chess_Interface)
- **Model type:** [Llama Architecture](https://huggingface.co/docs/transformers/main/model_doc/llama)
- **License:** [apache-2.0]()

## How to Get Started with the Model

This notebook is created to test the model's capabilities. You can use it to evaluate performance of the model.

[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1guqb9xjvOalFQV7AKucaFN0D3Kd1SSzC?usp=sharing)

### Challenge

You can use this model or dataset to train your own models as well, and challenge me in this new field.

# Training Details

### Training Data

[Q-bert/Elite-Chess-Games](https://huggingface.co/datasets/Q-bert/Elite-Chess-Games)

### Training Procedure

This model was fully trained from scratch with random weights. It was created from the ground up with a new configuration and model, and trained using the Hugging Face Trainer for 1200 steps. There is still potential for further training. You can see the training code below.
[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1VYtxJ2gYh-cXZbk1rOMlOISq8Enfw_1G#scrollTo=z2dj2aXALbc5)

**Training Loss Graph:**
![image/png](https://cdn-uploads.huggingface.co/production/uploads/63da3d7ae697e5898cb86854/GFurIWI_FIcfJNlER05RS.png)