ONNX: difference in MAX cpu <-> gpu execution

Hi Laurens,

Thanks for raising this issue. It definitely looks like a bug, but we unfortunately won’t be able to address it.

We’ve recently shifted our focus toward emerging GenAI and LLM use cases, which means we’re investing more in model and kernel development with MAX and Mojo. For more details, you can check out the MAX Graph API in our docs.

As we prioritize these workloads, we’re gradually phasing out support for the legacy ONNX format in future releases. That said, we’re here to ensure you still have everything you need for building models in MAX. If you’d like any advice or help with our APIs, please don’t hesitate to reach out.

Thanks again for your support and understanding.

2 Likes