Model Card for beans-0-1.5B

An LLM trained to reason about legal chess moves.

Model Description

• Developed by: Stephen Shkeda
• License: MIT
• Finetuned from model: deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B

Model Sources

• Repository: https://github.com/sshkeda/beans
• Training data: https://huggingface.co/datasets/sshkeda/beans-0-dataset.json

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 1
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 16
• total_train_batch_size: 32
• total_eval_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• num_epochs: 3.0

Framework versions

• Transformers 4.49.0
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0