Ruckus-PyAssi-13b: Python Code Generation Model

flytech/Ruckus-PyAssi-13b is a 13 billion parameter model derived from meta-llama/Llama-2-13b-hf. It has been specifically fine-tuned by flytech using Supervised Fine Tuning (SFT) and Low-Rank Adaptation (LoRA) techniques.

Key Capabilities & Training

• Code Generation: Primarily focused on generating Python code.
• Instructional Tasks: Excels at understanding and executing instructional tasks related to code.
• Executional Layer: Designed to function as an underlying layer for code execution.
• Training Data: Fine-tuned on 10,000 examples from the flytech/llama-python-codes-30k dataset.
• Training Efficiency: Trained for 13 hours on a single A6000 48GB vRAM GPU, utilizing a 4-bit architecture.

Intended Uses & Formatting

This model is ideal for applications requiring robust Python code generation. It is formatted for chat-based interactions, requiring prompts to be enclosed within [INST]...[/INST] tags for optimal performance. Further fine-tuning is explicitly supported due to its LoRA-based training.