Unable to generate an ONNX that works in ONNXRuntime

[edmosRovi]

Hi,

I tried to generate the ONNX model as in https://github.com/andresprados/SPIGA/issues/3 but when using it with ONNXRuntime I can't get it to work, because I get this failure:

"Fail: [ONNXRuntimeError] : 1 : FAIL : Load model from model.onnx failed:Node (/Squeeze) Op (Squeeze) [ShapeInferenceError] Dimension of input 1 must be 1 instead of 2"

Any idea what is going on?

Thanks.

[wangwei-630]

hello, did you solve this problem?

[edmosRovi]

Hello. No, not yet resolved.

[zhenyi1998]

Anyone managed to solve this issue? Really need to speed up the spiga using onxx :(

[naonao-beibei]

Hello. No, not yet resolved.

Please，Now are you solved？

[Mohammadakhavan75]

Hello,

I've been having trouble converting this model to ONNX, but I think I've narrowed down the problem to line 189 of the spiga/model/spiga.py file. In this line, the tensor is reshaped, which causes problems during the conversion to ONNX. This is because ONNX needs to know the exact shape of each variable (tensor) in order to create a graph. The reshape operation prevents ONNX from knowing the exact shape of the tensor, which causes the conversion to fail.