Overview

STEM-AI-mtl/phi-2-electrical-engineering is a specialized 2.7 billion parameter language model developed by STEM.AI, built upon Microsoft's phi-2 architecture. It is uniquely fine-tuned using LoRa (Low-Rank Adaptation) for the domain of electrical engineering, focusing on question-answering and code generation.

Key Capabilities

• Electrical Engineering Q&A: Excels at answering questions related to general electrical engineering concepts.
• Kicad Software Expertise: Provides Q&A support specifically for Kicad, an open-source electronic design automation (EDA) software.
• Python Code Generation: Capable of generating Python code, particularly for Kicad's scripting console.

Training Details

The model was fine-tuned on a combination of two datasets:

• STEM-AI-mtl/Electrical-engineering: A custom dataset with queries covering 65% general electrical engineering, 25% Kicad, and 10% Python code for Kicad scripting.
• garage-bAInd/Open-Platypus: An additional dataset related to STEM and NLP.

The LoRa PEFT training was conducted on a 48 Gb A40 Nvidia GPU.

Direct Use Cases

This model is ideal for developers and engineers seeking assistance with electrical engineering queries, Kicad software usage, and generating Python scripts for Kicad automation.