Shumai (Meta) is a fast differentiable tensor library in JavaScript and TypeScript that allows for efficient computation on tensors and supports automatic differentiation.
It is built with Bun and Flashlight and is designed for software engineers and researchers working with JavaScript and TypeScript.
Features
- Efficient computation on tensors
- Support for automatic differentiation
Use Cases
- Creating datasets
- Training small models
- Advanced/fine-grained training/inference logic
- Building applications
Suited For
- Software engineers
- Researchers
FAQ
Shumai (Meta) is built for JavaScript and TypeScript.
Yes, Shumai (Meta) provides efficient computation on tensors.
Yes, Shumai (Meta) supports automatic differentiation.