coreml-diag

Diagnose and remediate CoreML model load, performance, memory, and accuracy issues.

Updated Dec 23, 2025

Search Tags:#performance #memory #diagnostics #conversion #machine-learning #model-loading #coreml

Ready to run

npx skills add https://github.com/pradeepmouli/swift-template --skill coreml-diag-pradeepmouli

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: coreml-diag
Source: https://github.com/pradeepmouli/swift-template/tree/main/.agents/skills/axiom-ios-ml/coreml-diag
Command: npx skills add https://github.com/pradeepmouli/swift-template --skill coreml-diag-pradeepmouli

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

CoreML diagnostics identify and remediate issues that prevent CoreML models from loading, running efficiently, or producing accurate results on iOS and macOS devices.

Core Features & Use Cases

Load failure diagnosis: detect common causes like unsupported model versions or compute plan errors.
Performance and memory profiling: analyze first-load latency, per-operation compute units, memory growth, and device-specific bottlenecks.
Accuracy and conversion guidance: troubleshoot palettization, quantization, pruning, and cross-framework conversion issues.
Practical guidance: pressure scenarios for device vs simulator, and deployment considerations.

Quick Start

Run a diagnostic pass on your CoreML model by providing your model file and deployment target; the tool returns a prioritized action plan.

Frequently Asked Questions about coreml-diag

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I diagnose CoreML model load failures on iOS?▼

Diagnose CoreML load failures by providing the MLModel file, deployment target, and device profile to detect unsupported model versions and compute plan errors. The tool then returns a prioritized action plan to remediate the specific loading issues.

How do I profile CoreML performance and memory usage on device?▼

Profile CoreML performance and memory by analyzing first-load latency, per-operation compute units, and memory growth across device profiles. This identifies device-specific bottlenecks and returns targeted suggestions for optimizing on-device inference.

Does CoreML diagnostics work with quantized and palettized models?▼

CoreML diagnostics supports quantized and palettized models by troubleshooting accuracy and conversion issues related to palettization, quantization, pruning, and cross-framework conversion. It requires the MLModel file and deployment target to guide the analysis.

Can I use CoreML diagnostics for both simulator and on-device inference?▼

CoreML diagnostics applies to both simulator and on-device inference across diverse compute units. It provides practical guidance for pressure scenarios and deployment considerations specific to iOS and macOS devices.

What's the best way to troubleshoot CoreML accuracy issues after conversion?▼

Troubleshoot CoreML accuracy issues by analyzing the MLModel file alongside deployment target and device profile. The tool identifies conversion, palettization, quantization, and pruning problems to produce a prioritized remediation plan.

Why does CoreML inference run slowly on certain compute units?▼

CoreML inference slows down due to device-specific bottlenecks and suboptimal per-operation compute unit assignments. Diagnostics analyze the compute plan and memory growth to pinpoint the exact operations causing latency during on-device execution.