Validate usage of dim_value (values should be > 0) and dim_param (all values with the same string should equate to the same size) in shapes in the model.

[joao-alves97]

I tried to replicate fastT5 for a different transformers model (EncoderDecoderModel) but I'm always facing this error:

onnxruntime.capi.onnxruntime_pybind11_state.RuntimeException: [ONNXRuntimeError] : 6 : RUNTIME_EXCEPTION : Non-zero status code returned while running MatMul node. Name:'MatMul_155' Status Message: /onnxruntime_src/onnxruntime/core/framework/op_kernel.cc:44 OrtValue* onnxruntime::OpKernelContext::OutputMLValue(int, const onnxruntime::TensorShape&, size_t) status.IsOK() was false. Shape mismatch attempting to re-use buffer. {1,1,768} != {1,133,768}. Validate usage of dim_value (values should be > 0) and dim_param (all values with the same string should equate to the same size) in shapes in the model.

environment: huggingface-hub 0.0.8 importlib-metadata 4.10.0 numpy 1.21.5 onnx 1.10.2 onnxruntime 1.4.0 packaging 21.3 pip 19.2.3 progress 1.6 protobuf 3.19.1 psutil 5.8.0 pyparsing 3.0.6 regex 2021.11.10 requests 2.26.0 sacremoses 0.0.46 sentencepiece 0.1.96 setuptools 41.2.0 six 1.16.0 tokenizers 0.10.3 torch 1.10.1 tqdm 4.62.3 transformers 4.4.2

Anyone having a similar error? Any idea on how to solve it? Thanks

[yuslepukhin]

Do you have a repro code and would you be able to share your model?

[yuslepukhin]

Also would you be able to try it out with the latest release?

[stale[bot]]

This issue has been automatically marked as stale due to inactivity and will be closed in 7 days if no further activity occurs. If further support is needed, please provide an update and/or more details.