Most machine learning (ML) and deep learning (DL) libraries are dominated by the Python ecosystem. But with the rapid growth of Flutter and Dart, there’s a clear need for native Dart ML tools.
That’s why I built ml_transformer — a lightweight, modular, and extensible machine learning & neural network library written entirely in Dart.
🔑 Key Features
Transformer Architecture → Pre-LN block, Multi-Head Attention, FeedForward, LayerNorm, Rotary Positional Encoding (RoPE).
Core Layers → Embedding, Linear, LayerNorm, Attention, TransformerBlock.
Mini Training Engine → Autograd + AdamW optimizer for tiny training loops.
Quantization Helpers → int8, int4, int2, fp16, bf16.
ONNX Bridge → JSON-based import/export (protobuf support planned).
RNN Family → RNN, LSTM, GRU with batched forward.
Dataset & Batch API → Easy dataset loading & batching.
Callbacks & Monitoring → EarlyStopping, History.
FFI Acceleration → Native performance examples with NumPy references.
Unit Testing → Verified with NumPy integration and strong test coverage.
MIT Licensed → Fully open-source and free to use.
🎯 Why Dart for ML?
Flutter-first integration → Run ML pipelines directly in your apps.
Lightweight → Smaller footprint compared to heavy Python-based stacks.
Educational → Great for learning Transformer and RNN internals.
Prototyping → Build and test small-scale models quickly.
🌍 Who Can Use It?
🧑🎓 Students → Explore neural networks in a new language.
👨💻 Developers → Add ML pipelines directly into Flutter projects.
🔬 Researchers → Use as a minimalist sandbox for experiments.
🚀 Entrepreneurs → Quickly validate ideas with ML-enabled prototypes.
📦 Installation
Add it to your pubspec.yaml:
dependencies:
ml_transformer: ^1.0.0
📊 Resources
📦 Pub.dev package
📊 Score & Metrics
💻 GitHub Repository
💬 Closing Thoughts
This is just the beginning. ml_transformer is not intended to replace big ML frameworks — but to open the door for Dart/Flutter developers to experiment with ML natively.
I’d love to hear your feedback, contributions, and ideas to make it even better 🙌
👉 Try it out, fork it, and let’s push ML in Dart forward together!
Top comments (0)