ssam17/phi3-industrial-anomaly
ssam17/phi3-industrial-anomaly is a 3.8 billion parameter Phi-3-mini-4k-instruct model, fine-tuned by ssam17 using QLoRA for industrial IoT anomaly detection. It specializes in analyzing industrial sensor data and network telemetry to identify anomalies, security threats, and provide real-time diagnostic recommendations. This model is optimized for interpretable AI responses and efficient deployment on edge devices for applications like predictive maintenance and smart manufacturing.
Loading preview...
Overview
This model is a fine-tuned version of Microsoft's Phi-3-mini-4k-instruct, specifically adapted for industrial anomaly detection and interpretable diagnostics. Utilizing QLoRA (Quantized Low-Rank Adaptation), it processes industrial sensor data and network telemetry to provide actionable insights for automation systems. The model has 3.8 billion parameters and a 4096-token context length, making it suitable for detailed industrial logs.
Key Capabilities
- Industrial Anomaly Classification: Detects unusual patterns in sensor data.
- Security Threat Detection: Identifies potential security threats within industrial network telemetry.
- Sensor Data Interpretation: Analyzes and interprets complex sensor readings.
- Real-time Diagnostic Recommendations: Provides immediate insights and recommendations for detected issues.
- Explainable AI Responses: Aims to offer clear, understandable reasons for its detections.
Training and Performance
The model was fine-tuned for 5 epochs on the ssam17/Edge-Industrial-Anomaly-Phi3 dataset, comprising 10,749 training samples. It uses 4-bit NF4 quantization for memory efficiency and achieves an evaluation loss of 2.3992 with a token accuracy of 54.51%. It is designed for efficient inference, supporting PyTorch, GGUF, and ONNX formats, and can be deployed on edge devices like Jetson Nano or Raspberry Pi 5.
Good For
- Industrial IoT Monitoring: Real-time anomaly detection in manufacturing environments.
- Predictive Maintenance: Early warning systems for equipment failure.
- Security Operations: Network intrusion detection in operational technology (OT) and information technology (IT) environments.
- Edge Deployment: Lightweight inference on industrial gateways and edge devices.
- Smart Manufacturing: Quality control and process optimization.