Has anyone tried benchmarking Burn against ONNXRuntime?

[benjamintli]

I'm curious how Burn compares to something like ONNXRuntime. I see the ease of importing ONNX based models to be virtually the same between Burn and ORT, so I'm curious if anyone has tried doing a benchmark.

[laggui]

We don't have any benchmarks against ONNX runtimes. There are many EPs so it depends which targets you are interested in. We're really improving the performance for our cubecl based backends (CUDA, wgpu + spirv) so the next release should be much more meaningful :)