burn-wgpu

name	burn-wgpu
description	Burn WGPU backend - Cross-platform GPU acceleration using WebGPU. Supports Vulkan, Metal, Direct3D, and WebGL.

burn-wgpu

Overview

burn-wgpu provides cross-platform GPU acceleration using WebGPU through the wgpu crate. It supports:

• Vulkan: Linux, Android
• Metal: macOS, iOS
• Direct3D 12: Windows
• WebGL: Web browsers

Key Features

• Cross-platform GPU: Works on all major platforms
• Web Support: Runs in browsers via WASM
• Automatic Shader Generation: No manual shader writing
• Memory Efficient: Smart memory management

Usage

Basic Setup

use burn::tensor::Tensor;
use burn_wgpu::WgpuDevice;

// Create GPU device
let device = WgpuDevice::default();

// Create tensor on GPU
let tensor = Tensor::from_data(&[1.0, 2.0, 3.0], &device);

// GPU-accelerated operations
let result = tensor.matmul(other_tensor);

Backend Type

use burn_wgpu::WgpuBackend;

type Backend = WgpuBackend;
let model: Model<Backend> = Model::default();

burn.toml Configuration

[model_exports]
default_backend = "wgpu"
default_device = "Gpu"

[wgpu]

Best Practices

1. Use for production GPU inference
2. Enable wgpu feature for WebAssembly
3. Profile memory usage for large models
4. Use async operations for better performance

Performance Considerations

• Significantly faster than CPU for large tensors
• Memory transfer overhead between CPU/GPU
• Best for batch processing and large models
• WebAssembly performance may vary by browser

When to Use

• Production GPU inference
• Cross-platform applications
• Web-based ML applications
• High-performance training

Related Crates

• burn: Core framework
• burn-onnx: ONNX model import with GPU acceleration