cccrust

cccrust / rutorch

Public

A neural network framework for rust just like pytorch.

14
2
100% credibility
Found Mar 13, 2026 at 14 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Rust
AI Summary

Rutorch is an educational Rust project implementing a minimal tensor library with automatic differentiation, supporting CPU and Metal acceleration, and featuring interactive demos for neural networks on tasks like XOR, classification boundaries, and character-level text generation.

How It Works

1
🔍 Discover Rutorch

You stumble upon Rutorch, a playful sandbox where you can watch tiny AI brains learn simple tricks like logic puzzles and pattern spotting.

2
💻 Set it up on your computer

Grab the files and get everything ready with a quick setup so you can start experimenting right away.

3
Pick your first learning adventure
🧩
Logic puzzle (XOR)

Train it to figure out exclusive-or logic from examples.

🌀
Pattern boundaries

Help it draw invisible lines separating swirly or clumpy dots.

📝
Math expression generator

Teach it to predict and create math formulas character by character.

4
⚡ Watch the AI learn

Hit go and see the loss drop as your little network gets smarter over many practice rounds, feeling the excitement build.

5
📊 Check the amazing results

Review how perfectly it solves your puzzle, spots patterns, or writes new math, with optional pictures to visualize the magic.

🎉 You're an AI teacher now!

You've successfully trained neural networks from scratch, understanding how they learn, and ready to try more demos.

Sign up to see the full architecture

4 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 rutorch?

Rutorch brings PyTorch-style neural networks to Rust, handling autograd on tensors with ops like matmul, ReLU, softmax, and stable log_softmax/nll_loss. It supports CPU and Metal backends for Apple Silicon GPUs, plus toy datasets for spiral/blob classification and char-level RNN/GRU/GPT models on math expressions. Run `cargo run --release -- xor` for instant XOR training or `cargo run --release -- char gpu rnn 80` to train language models, outputting CSVs for easy plotting.

Why is it gaining traction?

Unlike heavyweight Rust NN crates, rutorch feels like PyTorch—quick demos visualize decision boundaries, mimicking neural network playgrounds without Python. Metal accel speeds up RNN training on Macs, and gradient clipping keeps neural network architectures stable for sequences. It's a from-scratch framework for neural networks and deep learning basics, outputting data for tools like neural network visualization.

Who should use this?

Rust devs prototyping neural network architectures for toys like XOR or spirals, especially on macOS wanting GPU without CUDA hassle. Suited for educators building neural networks for babies-style intros or experimenting with RNNs akin to neural forecast or neural cellular automata. Skip if you need production-scale like github neural network c++ ports.

Verdict

Solid playground for Rust NN hobbyists—demos shine, Metal works—but 13 stars and 1.0% credibility mean it's immature: sparse docs, basic tests. Worth forking for personal neural network from scratch projects; not yet for serious use.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.