Neograd

A minimal deep learning framework written in Go. It includes: • A basic autograd engine • Feedforward neural network (MLP) implementation • Tokenizer with BPE-like encoding • Gradient descent

Running an Example

Train a simple MLP using random data:

go run cmd/examples/neuralnet/main.go

It will initialize a network, run forward/backward passes, and print the training loss/accuracy.