Overview

This model, Plaban81/codegen-finetuned-python, is a 7 billion parameter variant of Meta's Llama-2 architecture. It has been specifically fine-tuned using the QLoRA method with 4-bit quantization and the PEFT library to excel at generating Python code. The training leveraged the python_code_instructions_18k_alpaca dataset, which comprises problem descriptions paired with Python code solutions, formatted in an Alpaca-style instruction format.

Key Capabilities

• Python Code Generation: Highly optimized for generating Python code based on given instructions.
• Instruction Following: Fine-tuned on an instruction-based dataset, enabling it to understand and respond to coding prompts effectively.
• Efficient Deployment: Trained with 4-bit QLoRA, making it suitable for environments with limited computational resources.

Good for

• Developers seeking an efficient, smaller-scale model for Python code generation tasks.
• Applications requiring code snippets or full function implementations in Python.
• Experimentation with fine-tuned Llama-2 models for specialized programming tasks.