Floating-point converter for FP32, FP64, FP16, bfloat16, TensorFloat-32 and arbitrary IEEE 754-style floating-point types.
This is a website for hands-on exploration of floating-point types, and a convenience tool for sanity checks and low-level debugging.
- Decimal to float representation conversion, and vice versa
- Wide range of IEEE 754-style types, including that of arbitrary exponent and significand width
- Configurable rounding modes for different use cases (e.g. truncation for deep learning)
Please open an issue if you encounter a bug, have questions, or would like to make a suggestion. Pull requests are also welcome.
This project will require nodejs
and yarn
. The following yarn scripts should help you get started on local development.
Install the npm packages this project depends on.
Runs the app in the development mode. Open http://localhost:3000 to view it in the browser. The page will reload if you make edits and you will also see any lint errors in the console.
Launches the test runner in the interactive watch mode. See running tests for more information.