Rust ML Primitives 🦀🧠

A collection of foundational machine learning primitives implemented from scratch in Rust.

🎯 Philosophy

Each primitive:

• Does one thing well (UNIX philosophy)
• Can be used standalone or composed with others
• Is well-documented with examples
• Focuses on educational clarity while being performant

📦 Primitives

Foundational (Level 1)

Crate	Status	Description
tensor-core	🔨 In Progress	Basic tensor operations
autodiff-core	📝 Planned	Automatic differentiation
text-primitives	📝 Planned	Tokenizers, vocabularies
embedding-core	📝 Planned	Embedding layers

Neural Components (Level 2)

Crate	Status	Description
rnn-primitives	📝 Planned	RNN cells and layers
lstm-primitives	📝 Planned	LSTM with gating
attention-primitives	📝 Planned	Attention mechanisms
linear-layers	📝 Planned	Dense/linear layers

Training (Level 3)

Crate	Status	Description
optimizer-core	📝 Planned	SGD, Adam, optimizers
nn-framework	📝 Planned	Neural network framework

🏗️ Example Frameworks

• mini-torch: PyTorch-like API using our primitives
• rusty-llm: Minimal LLM implementation

🚀 Getting Started

Using a Single Primitive

# In your Cargo.toml
[dependencies]
tensor-core = "0.1"

## Build and test

```sh
# Clone the repository
git clone https://github.com/rust-ml-learners/rust-ml-primitives
cd rust-ml-primitives

# Build everything
cargo build --workspace

# Build specific crate
cargo build -p tensor-core

# Run tests
cargo test --workspace

# Run benchmarks
cargo bench --workspace