RustGPT: A pure-Rust transformer LLM built from scratch

A complete Large Language Model implementation in pure Rust with no external ML frameworks. Built from the ground up using only ndarray for matrix operations.

This project demonstrates how to build a transformer-based language model from scratch in Rust, including:

• Pre-training on factual text completion
• Instruction tuning for conversational AI
• Interactive chat mode for testing
• Full backpropagation with gradient clipping
• Modular architecture with clean separation of concerns

Start with these two core files to understand the implementation:

• src/main.rs - Training pipeline, data preparation, and interactive mode
• src/llm.rs - Core LLM implementation with forward/backward passes and training logic

The model uses a transformer-based architecture with the following components:

Input Text → Tokenization → Embeddings → Transformer Blocks → Output Projection → Predictions

src/
├── main.rs              # 🎯 Training pipeline and interactive mode
├── llm.rs               # 🧠 Core LLM implementation and training logic
├── lib.rs               # 📚 Library exports and constants
├── transformer.rs       # 🔄 Transformer block (attention + feed-forward)
├── self_attention.rs    # 👀 Multi-head self-attention mechanism  
├── feed_forward.rs      # ⚡ Position-wise feed-forward networks
├── embeddings.rs        # 📊 Token embedding layer
├── output_projection.rs # 🎰 Final linear layer for vocabulary predictions
├── vocab.rs            # 📝 Vocabulary management and tokenization
├── layer_norm.rs       # 🧮 Layer normalization
└── adam.rs             # 🏃 Adam optimizer implementation

tests/
├── llm_test.rs         # Tests for core LLM functionality
├── transformer_test.rs # Tests for transformer blocks
├── self_attention_test.rs # Tests for attention mechanisms
├── feed_forward_test.rs # Tests for feed-forward layers
├── embeddings_test.rs  # Tests for embedding layers
├── vocab_test.rs       # Tests for vocabulary handling
├── adam_test.rs        # Tests for optimizer
└── output_projection_test.rs # Tests for output layer

The implementation includes two training phases:

1. Pre-training: Learns basic world knowledge from factual statements

   • "The sun rises in the east and sets in the west"
   • "Water flows downhill due to gravity"
   • "Mountains are tall and rocky formations"

2. Instruction Tuning: Learns conversational patterns

   • "User: How do mountains form? Assistant: Mountains are formed through tectonic forces..."
   • Handles greetings, explanations, and follow-up questions

# Clone and run
git clone https://github.com/tekaratzas/RustGPT.git 
cd RustGPT
cargo run

# The model will:
# 1. Build vocabulary from training data
# 2. Pre-train on factual statements (100 epochs)  
# 3. Instruction-tune on conversational data (100 epochs)
# 4. Enter interactive mode for testing

After training, test the model interactively:

Enter prompt: How do mountains form?
Model output: Mountains are formed through tectonic forces or volcanism over long geological time periods

Enter prompt: What causes rain?
Model output: Rain is caused by water vapor in clouds condensing into droplets that become too heavy to remain airborne

• Vocabulary Size: Dynamic (built from training data)
• Embedding Dimension: 128
• Hidden Dimension: 256
• Max Sequence Length: 80 tokens
• Architecture: 3 Transformer blocks + embeddings + output projection

• Optimizer: Adam with gradient clipping
• Pre-training LR: 0.0005 (100 epochs)
• Instruction Tuning LR: 0.0001 (100 epochs)
• Loss Function: Cross-entropy loss
• Gradient Clipping: L2 norm capped at 5.0

• Custom tokenization with punctuation handling
• Greedy decoding for text generation
• Gradient clipping for training stability
• Modular layer system with clean interfaces
• Comprehensive test coverage for all components

# Run all tests
cargo test

# Test specific components
cargo test --test llm_test
cargo test --test transformer_test
cargo test --test self_attention_test

# Build optimized version
cargo build --release

# Run with verbose output
cargo test -- --nocapture

This implementation demonstrates key ML concepts:

• Transformer architecture (attention, feed-forward, layer norm)
• Backpropagation through neural networks
• Language model training (pre-training + fine-tuning)
• Tokenization and vocabulary management
• Gradient-based optimization with Adam

Perfect for understanding how modern LLMs work under the hood!

• ndarray - N-dimensional arrays for matrix operations
• rand + rand_distr - Random number generation for initialization

No PyTorch, TensorFlow, or Candle - just pure Rust and linear algebra!

Contributions are welcome! This project is perfect for learning and experimentation.

• 🏪 Model Persistence - Save/load trained parameters to disk (currently all in-memory)
• ⚡ Performance optimizations - SIMD, parallel training, memory efficiency
• 🎯 Better sampling - Beam search, top-k/top-p, temperature scaling
• 📊 Evaluation metrics - Perplexity, benchmarks, training visualizations

• Advanced architectures (multi-head attention, positional encoding, RoPE)
• Training improvements (different optimizers, learning rate schedules, regularization)
• Data handling (larger datasets, tokenizer improvements, streaming)
• Model analysis (attention visualization, gradient analysis, interpretability)

1. Fork the repository
2. Create a feature branch: git checkout -b feature/model-persistence
3. Make your changes and add tests
4. Run the test suite: cargo test
5. Submit a pull request with a clear description

• Follow standard Rust conventions (cargo fmt)
• Add comprehensive tests for new features
• Update documentation and README as needed
• Keep the "from scratch" philosophy - avoid heavy ML dependencies

• 🚀 Beginner: Model save/load, more training data, config files
• 🔥 Intermediate: Beam search, positional encodings, training checkpoints
• ⚡ Advanced: Multi-head attention, layer parallelization, custom optimizations

Questions? Open an issue or start a discussion!

No PyTorch, TensorFlow, or Candle - just pure Rust and linear algebra!