LLAMA-3.2-3B-MathInstruct_LORA_SFT

This model is a fine-tuned version of meta-llama/Llama-3.2-3B-Instruct on the MathInstruct dataset. The fine-tuning process was designed to enhance the model's performance for mathematical instruction-following tasks, ensuring improved accuracy and precision when solving math-related problems.

It achieves the following results on the evaluation set:

• Loss: 0.6895

---

Model Description

This model is specifically fine-tuned for mathematical reasoning, problem-solving, and instruction-following tasks. Leveraging the LLaMA-3.2-3B-Instruct base model, it has been optimized to handle mathematical queries and tasks with improved efficiency and context understanding.

---

Training and Evaluation Data

The model was fine-tuned on the MathInstruct dataset.

• Dataset Source: TIGER-Lab.
• Dataset Focus: Mathematical instruction-following and reasoning tasks.
• Scope: A wide range of math topics, including arithmetic, algebra, calculus, and problem-solving.

The dataset was carefully curated to align with instructional objectives for solving mathematical problems and understanding step-by-step reasoning.

---

Training Procedure

Hyperparameters

• Learning rate: 0.0001
• Train batch size: 1
• Eval batch size: 1
• Gradient accumulation steps: 8
• Total effective batch size: 8
• Optimizer: AdamW (torch)
  • Betas: (0.9, 0.999)
  • Epsilon: 1e-08
• Learning rate scheduler: Cosine schedule with 10% warmup.
• Number of epochs: 3.0

Framework Versions

• PEFT: 0.12.0
• Transformers: 4.46.1
• PyTorch: 2.5.1+cu124
• Datasets: 3.1.0
• Tokenizers: 0.20.3

---

Training Results

• Loss: 0.6895
• Evaluation indicates strong performance on math instruction-following tasks. Further testing on specific use cases is recommended to assess the model’s generalizability.

---

Additional Information

• Author: Sri Santh M
• Purpose: Fine-tuned for educational and development purposes, particularly for math-related tasks.
• Dataset Link: MathInstruct Dataset

This model represents a focused effort to adapt the LLaMA-3.2-3B-Instruct model for specialized mathematical use cases. It can be further fine-tuned or extended for more specific mathematical domains or applications.