How does this skill help with mobile deployment?

It provides specific patterns for converting models to TFLite, optimizing them for mobile kernels, and benchmarking performance on Android devices using adb.

Can I use this for JAX models?

Yes, the skill includes specialized workflows for converting JAX modules to SavedModel and TFLite formats using Orbax export tools.

Does it cover high-performance serving?

Yes, it includes guidance for TensorFlow Serving, graph transformations for production optimization, and distributed strategies using MirroredStrategy.

What types of quantization are supported?

It supports Post-Training Float16, Dynamic Range, and Full Integer quantization, including calibration with representative datasets for precision.

TensorFlow Model Deployment

Name: TensorFlow Model Deployment
Author: TheBushidoCollective

byTheBushidoCollective

•

Data Science & ML

Deploy, optimize, and serve TensorFlow and JAX models across production, mobile, and edge environments.

About

This skill provides Claude with specialized capabilities for the end-to-end deployment lifecycle of TensorFlow models. It includes expertise in exporting to SavedModel format, converting models for mobile and edge use via TFLite, and applying advanced quantization techniques to reduce model size and improve latency. Additionally, it covers JAX-to-TFLite conversion, graph transformations for mobile kernels, and distributed training strategies, ensuring models are production-ready and optimized for their target hardware or cloud infrastructure.

Key Features

SavedModel and Keras model export with custom serving signatures
Advanced quantization (Float16, Dynamic Range, Full Integer) for optimization
39 GitHub stars
Comprehensive TFLite conversion for mobile and edge devices
Model graph transformations and distributed training deployment
JAX to TFLite conversion via Orbax and SavedModel workflows

Use Cases

Implementing distributed serving architectures for high-throughput AI applications
Optimizing large machine learning models for deployment on Android or IoT devices
Converting research-grade JAX models into production-ready TFLite formats

About

Key Features

SavedModel and Keras model export with custom serving signatures
Advanced quantization (Float16, Dynamic Range, Full Integer) for optimization
39 GitHub stars
Comprehensive TFLite conversion for mobile and edge devices
Model graph transformations and distributed training deployment
JAX to TFLite conversion via Orbax and SavedModel workflows

Use Cases

Implementing distributed serving architectures for high-throughput AI applications
Optimizing large machine learning models for deployment on Android or IoT devices
Converting research-grade JAX models into production-ready TFLite formats