Realm

Machine learning basics

Source-owned programs from the current OA SDK. Rust and Python examples follow their respective public APIs.

Rust · Python

The former Fashion-MNIST classifier tutorial has not been admitted to the Rust SDK. This page now points to the current Python ML basics program; a paired Rust regression example is available in the SDK.

Example

This example is currently documented in Python only.

Tu_ml_basics.py

1import oa
2x = oa.Matrix.new_from_f32([5, 1], [-2.0, -1.0, 0.0, 1.0, 2.0])
3target = oa.Matrix.new_from_f32([5, 1], [-3.0, -1.0, 1.0, 3.0, 5.0])
4
5model = oa.ml.Linear(1, 1, seed=0)
6optimizer = oa.ml.Sgd(model.parameters(), learning_rate=0.05)
7
8initial_loss = 0.0
9for step in range(80):
10 optimizer.zero_grad()
11 with oa.ml.GradientTape() as tape:
12 prediction = model.forward(x)
13 loss = oa.ml.loss.mse(prediction, target)
14 tape.backward(loss)
15 optimizer.step()
16 if step == 0:
17 initial_loss = oa.ml.metric.scalar_loss(loss)
18
19final_loss = oa.ml.metric.scalar_loss(loss)
20prediction = model.forward(x)
21values = prediction.read_f32()
22
23assert final_loss < initial_loss * 0.01
24assert max(abs(a - b) for a, b in zip(values, [-3, -1, 1, 3, 5])) < 0.1
25
26print(f"loss: {initial_loss:.6f} -> {final_loss:.6f}")
27print(values)
28

Run Rust tutorials from the OA checkout with cargo run --release --example <name>. Run Python with the environment containing the OA native extension. Consult each program's arguments and capability checks before running it.