How does this skill help with model conversion?

It provides optimized implementation patterns and CLI guidance for converting PyTorch and TensorFlow models to CoreML using coremltools.

Can it help reduce model size?

Yes, it includes specific routines for model compression techniques like quantization, palettization, and pruning to ensure models fit on mobile devices.

Does it support Large Language Models (LLMs)?

Absolutely, it provides patterns for implementing KV-caching and stateful models essential for running transformers on-device.

How do I handle speech transcription?

The skill guides you through setting up SpeechAnalyzer and SpeechTranscriber for both live and recorded audio processing on modern iOS versions.

iOS Machine Learning & Speech

Name: iOS Machine Learning & Speech
Author: CharlesWiltgen

byCharlesWiltgen

•

233

Mobile Development

Streamlines the deployment of machine learning models and speech-to-text features on Apple platforms through CoreML optimization and implementation guidance.

概要

This skill provides a comprehensive framework for integrating advanced machine learning and speech recognition into iOS, iPadOS, watchOS, and tvOS applications. It offers specialized routing for CoreML model conversion from PyTorch or TensorFlow, advanced model compression techniques like quantization and pruning, and the implementation of state-of-the-art features such as KV-cache for LLMs and MLTensor for pipeline stitching. Additionally, it simplifies the integration of Apple's Speech framework, enabling both live and file-based transcription with modern APIs like SpeechAnalyzer and SpeechTranscriber.

主な機能

Automated CoreML conversion and model stitching via MLTensor
Advanced model compression including quantization, palettization, and pruning
Implementation patterns for stateful models and KV-cache for LLMs
Comprehensive Speech-to-Text integration using SpeechAnalyzer
Diagnostic workflows for debugging inference performance and memory pressure
233 GitHub stars

ユースケース

Converting and optimizing PyTorch or TensorFlow models for on-device execution
Implementing real-time speech transcription and audio analysis in mobile apps
Deploying Large Language Models (LLMs) to iOS with efficient state management

概要

主な機能

Automated CoreML conversion and model stitching via MLTensor
Advanced model compression including quantization, palettization, and pruning
Implementation patterns for stateful models and KV-cache for LLMs
Comprehensive Speech-to-Text integration using SpeechAnalyzer
Diagnostic workflows for debugging inference performance and memory pressure
233 GitHub stars

ユースケース

Converting and optimizing PyTorch or TensorFlow models for on-device execution
Implementing real-time speech transcription and audio analysis in mobile apps
Deploying Large Language Models (LLMs) to iOS with efficient state management