coreml-diag

What problem does it solve?

This Skill helps developers diagnose and resolve common issues encountered with CoreML models, including load failures, performance bottlenecks, memory problems, accuracy degradation after compression, compute unit conflicts, and conversion errors.

Core Features & Use Cases

• Load Failure Diagnosis: Identifies reasons for models not loading, such as unsupported versions or compute plan issues.
• Performance Optimization: Analyzes slow inference and first-load times, suggesting fixes related to compute units and caching.
• Accuracy & Compression Issues: Provides strategies for addressing accuracy loss after model compression techniques like palettization, quantization, and pruning.
• Conversion Troubleshooting: Helps resolve errors during model conversion and identifies causes for output discrepancies.
• Use Case: A developer's CoreML model is crashing on a specific device. This Skill can guide them through checking deployment targets, compute unit compatibility, and profiling on the actual device to pinpoint the cause.

Quick Start

Use the coreml-diag skill to troubleshoot a CoreML model that is failing to load.