huggingface

huggingface / transformers-to-mlx

Public

Agent Skill to help convert transformer LLMs to mlx-lm

14
1
100% credibility
Found Apr 23, 2026 at 14 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

A skill for AI agents to convert language models from Hugging Face Transformers to MLX format, with scripts for verification and testing.

How It Works

1
🔍 Discover the Helper

You learn about a handy tool that lets AI assistants convert language models to run super fast on Macs.

2
đź“Ą Add to Your AI Friend

You easily add this special skill to your AI assistant so it can handle conversions for you.

3
đź’¬ Ask for a Conversion

You simply tell your AI something like 'Please convert this model,' and it gets to work right away.

4
⚙️ It Does the Magic

Your AI sets up a private workspace, grabs the model, studies it, builds a new version, and runs tests to make sure everything matches perfectly.

5
âś… Check the Results

You review the detailed comparisons showing the new model behaves just like the original.

6
Decide Next
đź”’
Keep Private

Save the converted model for your own fast AI projects on Mac.

📤
Share It

Let your AI prepare everything to suggest adding it to the shared collection.

🎉 Faster AI Ready

You now have a speedy version of the model running smoothly on your Mac, feeling the boost in performance.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 14 to 14 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 transformers-to-mlx?

This Python-based agent skill helps convert Hugging Face Transformers LLMs to the mlx-lm format for Apple's MLX framework. Tell your agent—like Claude via agent skills on GitHub—"convert olmo_hybrid to MLX," and it spins up a virtual environment, analyzes the model config, ports the architecture, runs equivalence tests, and preps a PR for mlx-lm. It tackles the tedious manual porting of complex models from PyTorch to efficient Apple Silicon inference.

Why is it gaining traction?

Unlike manual conversions or basic scripts, it automates discovery, dtype resolution, layer-wise comparisons, and generation checks—handling MoEs, hybrid SSM+attention, and more—while bundling verification tools for predictions and logits. Developers dig the agent github claude integration, akin to agent skills anthropic or agent github copilot vscode, turning prompts into tested ports. Low stars haven't slowed shares on agent skills reddit and agent skills github.

Who should use this?

ML engineers porting HF models to MLX for Mac optimization, mlx-lm contributors validating new architectures, or agent skills io experimenters building agent github copilot cli workflows. Ideal for those tired of debugging dtype mismatches or layer drifts in transformers to mlx migrations, especially with agent skills vs mcp debates.

Verdict

Grab it if you're in the MLX ecosystem—installation is a one-liner uv command, docs are solid, and tests boost confidence—but at 14 stars and 1.0% credibility, it's early-stage; review outputs before PRs. Promising agent skills library entry, worth forking for custom tweaks.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.