IrError(OutputNodeNotFound("/Cast_output_0"))

[BinaryWarlock]

Describe the bug
IrError(OutputNodeNotFound("/Cast_output_0"))

To Reproduce
Steps to reproduce the behavior:
Using onnx-clip model which you can grab at https://lakera-clip.s3.eu-west-1.amazonaws.com/clip_image_model_vitb32.onnx (about 336 MiB)

1. python -m onnxruntime.tools.make_dynamic_shape_fixed --dim_param image_batch_size --dim_value 1 clip_image_model_vitb32.onnx clip_image_model_vitb32_batchsize1.onnx
2. nnx prepare clip_image_model_vitb32_batchsize1.onnx clip_image_model_vitb32_batchsize1.prepared.onnx
3. from wonnx import Session; Session.from_path('clip_image_model_vitb32_batchsize1.prepared.onnx')
4. Receive IrError(OutputNodeNotFound("/Cast_output_0"))

The node does seem to exist when examined in https://netron.app/ so I don't know what's happening here.

[cchance27]

Running into a similar issue with a faceparser onnx... i end up with.... output node for output /cp/resnet/conv1/Conv_output_0 not found

but checking in netron... its available

[phhusson]

Using onnx-simplifier on my model seemed to bypass that error. I tried to understand why but I failed.

Edit: Nevermind the reason was that I had to do nnx prepare on the mode, (which actually doesn't work on my model but that's another story). The error OutputNodeFound("/Constant_output_0") here meant that it couldn't find the shape of /Constant_output_0 because I didn't launch shape inference first. onnx-simplify alongst the simplification does compute the shapes.