Anemll

Anemll / anemll-profile

Public

ANE (Apple Neural Engine) CostModel profiler for CoreML models

23
2
100% credibility
Found Mar 09, 2026 at 19 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Objective-C
AI Summary

A macOS tool that profiles CoreML AI models to analyze operation costs, device usage, and real-world prediction speeds on Apple's Neural Engine.

How It Works

1
🔍 Discover the profiler

You hear about a handy tool that checks how well your AI models run on your Mac's special chip.

2
📥 Get it set up

You easily add the tool to your Mac with a quick install command.

3
📁 Pick your model

You choose the folder or file where your AI model lives on your computer.

4
🚀 Run the check

You tell the tool to analyze your model, and it quickly examines speed and what parts use the most power.

5
📊 See the breakdown

A clear report pops up showing slow spots, fast predictions, and tips on what's using CPU or graphics.

6
💡 Spot improvements

You learn which operations are the slowest and why some parts can't use the fast chip.

✅ Optimize your AI

Now you know exactly how to make your model faster and smoother on your Mac.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 19 to 23 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is anemll-profile?

anemll-profile is an Objective-C CLI tool that profiles CoreML models on Apple Silicon, diving into ANE Apple Neural Engine cost estimates, per-op device placement, and real-world prediction throughput. It tells you exactly why operations fall back to CPU or GPU—think unsupported tensor types or compiler quirks—and breaks down compute/memory bottlenecks with GFLOP/s, GB/s metrics. Developers get a clear view of ANE Apple Silicon efficiency without digging through Apple ANE API logs themselves.

Why is it gaining traction?

Unlike generic CoreML tools, it combines costmodel analysis from the Apple ANE compiler service with actual throughput benchmarks, spotlighting top expensive ops and conv details for quick optimizations. The brew install and simple CLI—like `anemll-profile model.mlmodelc -a` for GPU inclusion—make it dead simple for PyTorch Apple ANE exports or ANE Apple LLM tuning. It surfaces fallback reasons (e.g., "Cannot support standalone slice_update") that stump most devs chasing Apple one vs GPU balance.

Who should use this?

ML engineers deploying CoreML models to iOS/macOS apps who need to maximize ANE usage and minimize CPU fallbacks. iOS devs optimizing on-device neural models, especially from PyTorch Apple ANE conversions, profiling throughput before App Store submission. Teams benchmarking Apple Neural Engine vs GPU for costmodel-driven tweaks.

Verdict

Grab it if you're deep in ANE CoreML work—solid docs and MIT license make the 19 stars and 1.0% credibility score forgivable for this niche profiler. Still early; watch for macOS updates beyond Sonoma.

(178 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.