6 ms·

Pretty nice to get downscaled mobile models with will be huge in 2024 when the compilation platforms like onnx coreml webgpu get smooth enough