Tommy van der Vorst

> 2. WONNX does not currently support parameterized dimensions, which would be required to implement the text encoder. The shape inference engine in WONNX now supports this (it allows you...

> I looked into this a few weeks ago - it is a significant chunk of work for 2 reasons: > > 1. The ops to implement are complicated (i.e...

You can now also use `nnx prepare model.onnx model-prepared.onnx --set some_dim=123` (if you have a fully inferred model that is)

Yeah found that solution too, unfortunately that tool is not available for brew-installed Python..?

A temporary fix that works for me is to download the list of trusted roots from Mozilla [here](https://wiki.mozilla.org/CA/Included_Certificates) (pick [PEM of Root Certificates in Mozilla's Root Store with the Websites...

Can you share the specific onnx file you are using? The error in general means that the output is missing somehow. If it is in the ONNX file and properly...

@mayjs backend test appears to have passed? Re the UNet sample: how large would the download size be for running the example/test? I might be able to host the files...

Hm, the file linked does not have all its shapes inferred (`nnx prepare` is unable to infer all shapes, but that is expected as shape inference for Conv is not...

> Probably the easiest way to fix this would be to run the tests locally... I can try later if necessary.

Good news and bad news: The above does seem to be a bug in the optimizer, it appears to attempt constant folding on the missing node. I just committed https://github.com/webonnx/wonnx/commit/5d20e966473ad71fcdafba0bf5664a34f07f8a95...