serez-ai

Neural networks, autodiff, and training loops — with a Keras-like API. Build, train, and run models in a few lines of Serez Code.

Install

sz install serez-ai

Quick start

import "serez-ai"

Random.seed(42)

// XOR dataset
let X = [[0.0, 0.0], [0.0, 1.0], [1.0, 0.0], [1.0, 1.0]]
let y = [[0.0], [1.0], [1.0], [0.0]]

// Build the model
let model = new Sequential()
model.add(new Dense(2, 16, "relu"))
model.add(new Dense(16, 1, "sigmoid"))

// Train
let opt  = new Adam(0.01, 0.9, 0.999)
let loss = model.fit_opt(X, y, new BCE(), 1000, opt)

out "Final loss: {loss}"

// Predict
let pred = model.forward([[1.0, 0.0]])
out "XOR(1, 0) ≈ {pred[0][0]}"   // → ~0.97

Sequential model

Sequential stacks layers one after the other. Add layers with .add(), then train with .fit_opt():

let model = new Sequential()
model.add(new Dense(input_size, output_size, activation))

// fit_opt(X, y, loss_fn, epochs, optimizer) → final loss
let loss = model.fit_opt(X, y, new MSE(), 500, new SGD(0.01))

// Forward pass — returns array of predictions
let predictions = model.forward(X)

Layers

// Image classifier example
let model = new Sequential()
model.add(new Conv2D(32, 3, 1, 1))
model.add(new BatchNorm(32))
model.add(new Dense(32, 64, "relu"))
model.add(new Dropout(0.2))
model.add(new Dense(64, 10, "sigmoid"))

Optimizers

new Adam(lr, beta1, beta2)       // e.g. new Adam(0.001, 0.9, 0.999)
new SGD(lr)                      // e.g. new SGD(0.01)
new Momentum(lr, momentum)       // e.g. new Momentum(0.01, 0.9)

Loss functions

new MSE()   // Mean Squared Error — regression
new BCE()   // Binary Cross-Entropy — binary classification

Tensors

Create tensors directly and operate on them. Gradients are tracked automatically:

import "serez-ai"

let a = new Tensor([[1.0, 2.0], [3.0, 4.0]])
let b = new Tensor([[2.0, 0.0], [1.0, 3.0]])

// Operations — all differentiable
let c = Autodiff.add(a, b)
let d = Autodiff.mul(a, b)
let e = Autodiff.relu(c)

// Compute gradients
Autodiff.backward(e)

out a.grad   // gradient of the loss w.r.t. a
out b.grad   // gradient of the loss w.r.t. b

Autodiff operations

Random

Reproducible random numbers for dataset shuffling and weight initialization:

Random.seed(42)          // set seed for reproducibility
let n = Random.float()   // decimal in [0, 1)
let i = Random.int(10)   // int in [0, 10)