docs.rs failed to build burn-cubecl-0.22.0
Please check the build logs for more information.
See Builds for ideas on how to fix a failed build, or Metadata for how to configure docs.rs builds.
If you believe this is docs.rs' fault, open an issue.
Please check the build logs for more information.
See Builds for ideas on how to fix a failed build, or Metadata for how to configure docs.rs builds.
If you believe this is docs.rs' fault, open an issue.
Visit the last successful build:
burn-cubecl-0.22.0-pre.4
Burn CubeCL Backend
CubeBackend implements Burn's tensor operations as CubeCL kernels, compiled just in time for the
device they run on. CUDA, ROCm, Metal, Vulkan, WebGPU, wgpu and the CubeCL CPU runtime all share
this one backend type; a tensor's device says which runtime it uses.
Usage
Applications select a runtime with a Burn feature and a Device constructor:
| Runtime | Burn feature | Device | Crate |
|---|---|---|---|
| CUDA | cuda |
Device::cuda(0) |
burn-cuda |
| ROCm | rocm |
Device::rocm(0) |
burn-rocm |
| wgpu | wgpu |
Device::wgpu(Default::default()) |
burn-wgpu |
| Metal | metal |
Device::metal(Default::default()) |
burn-wgpu |
| Vulkan | vulkan |
Device::vulkan(Default::default()) |
burn-wgpu |
| WebGPU | webgpu |
Device::webgpu(Default::default()) |
burn-wgpu |
| CPU | cpu |
Device::cpu() |
burn-cpu |
Use this crate directly to write custom kernels: kernel and ops hold the building blocks, and
cubecl is re-exported so kernels use the same CubeCL version as the backend. See the
custom CubeCL kernel
chapter of the Burn Book.
Feature Flags
cuda,hip,wgpu,metal,vulkan,webgpu,cpu: compile in a CubeCL runtime.fusion(default): kernel fusion through burn-fusion.autotune(default): benchmark kernel variants at runtime and keep the fastest.fft: FFT kernels.template: launch hand-written, non-JIT kernels.tracing: instrument operations with thetracingcrate.
Part of the Burn deep learning framework. See the Burn Book and the API documentation.