Can I use this for both training and inference?

Yes, it covers training-centric frameworks like Burn and Candle as well as inference-only tools like tract.

Does it support GPU acceleration?

Yes, it includes specific patterns for CUDA and Metal support, as well as async data loading to keep the GPU utilized.

Which Rust ML frameworks does this skill support?

It provides specialized guidance for Candle, Burn, tch-rs, tract (ONNX), and ndarray for tensor manipulation.

How does domain-ml handle memory efficiency?

It enforces zero-copy operations, tensor views, and singleton model loading patterns to minimize memory bandwidth bottlenecks.

How does it help with Python interoperability?

It focuses on standard model formats like ONNX and PyTorch bindings via tch-rs to bridge the gap between Python training and Rust deployment.

Machine Learning for Rust

Name: Machine Learning for Rust
Author: actionbook

byactionbook

•

596

データサイエンスとML

Accelerates the development of high-performance ML and AI applications in Rust by providing domain-specific design patterns and framework guidance.

概要

The domain-ml skill optimizes Claude's performance when assisting with Rust-based machine learning projects, focusing on the critical trade-offs between memory efficiency, GPU acceleration, and model portability. It provides structured guidance for using major Rust ML crates like Candle, Burn, tch-rs, and tract, ensuring developers implement zero-copy tensor operations, efficient batched inference, and lazy model initialization. By bridging the gap between Python-heavy research and Rust-based production environments, this skill helps users build robust, deterministic, and highly performant AI systems.

主な機能

Standardizes model loading and portability using ONNX and the tract runtime
Implements efficient batch processing and streaming for high-throughput data pipelines
Provides implementation patterns for GPU-accelerated inference using Candle and tch-rs
Optimizes memory usage through zero-copy and in-place tensor operations
Enforces best practices for numerical precision and reproducibility in Rust ML code
596 GitHub stars

ユースケース

Deploying pre-trained ONNX models for production inference in Rust
Building high-performance neural networks with pure Rust frameworks like Candle or Burn
Creating efficient data processing pipelines for large-scale ML datasets using Polars and ndarray

概要

主な機能

Standardizes model loading and portability using ONNX and the tract runtime
Implements efficient batch processing and streaming for high-throughput data pipelines
Provides implementation patterns for GPU-accelerated inference using Candle and tch-rs
Optimizes memory usage through zero-copy and in-place tensor operations
Enforces best practices for numerical precision and reproducibility in Rust ML code
596 GitHub stars

ユースケース

Deploying pre-trained ONNX models for production inference in Rust
Building high-performance neural networks with pure Rust frameworks like Candle or Burn
Creating efficient data processing pipelines for large-scale ML datasets using Polars and ndarray