Skip to content

Latest commit

 

History

History
34 lines (24 loc) · 775 Bytes

README.md

File metadata and controls

34 lines (24 loc) · 775 Bytes
Raw

minigrad

A neural network library in golang on an autograd engine. Faster and easy to use like PyTorch. ( This is Andrej Karpathy's Micrograd implementation in Go)

Usage

package main

import (
  mg "github.com/0verread/minigrad"
)

func main() {
  a := mg.NewValue(-4.0, nil, "")
  b := mg.NewValue(2.0, nil, "")
  c := a.Add(b)
  d := a.Mul(b).Add(b.Pow(3))
  c = c.Add(c.Add(mg.NewValue(1.0, nil, "")))
  c = c.Add(mg.NewValue(1.0, nil, "").Add(c))
  d = d.Add(d.Mul(2.0).Add(b.Add(a).Relu()))

  d.Backward()

}

Contribution

Feel free to raise a issue or PR to make enhancements or report a bug. I would love to see more test cases.

License

minigrad is under MIT.