mnist-gpu

Overview ¶

MNIST GPU Inference Benchmark

This example demonstrates CPU vs WebGPU performance comparison for neural network inference. It creates a simple MLP model and measures forward pass times on both backends.

Usage:

go run ./examples/mnist-gpu -batch 256 -iterations 100

Note: This benchmark focuses on the compute-intensive operations (MatMul, ReLU). WebGPU shines for large matrix operations, while CPU may be faster for small batches.