Hi, onnx issue when running the inference part of examples/chinese_prosody_polyphone/run.sh

[dyustc]

Hi, I went to the frontend folder, it seems the g2p_prosody.py is a ready-to-go inference script, only with this missing files. So where am I suppose to find these files in this repo? Correct me if you would, thanks. usage: g2p_prosody.py [-h] --text TEXT --hanzi2pinyin_file HANZI2PINYIN_FILE --polyphone_phone_file POLYPHONE_PHONE_FILE --polyphone_prosody_model POLYPHONE_PROSODY_MODEL

[dyustc]

hi, when I am running examples/chinese_prosody_polyphone/run.sh, the last part is inference part, I got the following error [E:onnxruntime:, sequential_executor.cc:369 Execute] Non-zero status code returned while running Reshape node. Name:'/transform/self_attn/Reshape_4' Status Message: /onnxruntime_src/onnxruntime/core/providers/cpu/tensor/reshape_helper.h:40 onnxruntime::ReshapeHelper::ReshapeHelper(const onnxruntime::TensorShape&, onnxruntime::TensorShapeVector&, bool) gsl::narrow_cast<int64_t>(input_shape.Size()) == size was false. The input tensor cannot be reshaped to the requested shape. Input shape:{7,1,768}, requested shape:{10,8,96}

also before that, when I convert .pt file to .onnx file, it also hints me to use onnx version 14 instead of version 13, otherwise the conversion would fail

[mszpc]

I have also encountered the same problem. I'm not sure if you have found a solution to this problem

[wk448378469]

The pre trained model provided by the official uses the onnx file exported from Torch version 1.13.1. However, there is no limited version in requirements. txt, so the 2.0.1 torch will be downloaded.

Solution:

Option 1: requirements. txt torch==1.13.1 Option 2: Modify wetts/frontend/export_onnx. py dummy_input shape, such as 1 512, and then limit the input length to 1 512 when using your own ONNX model，and download torch nightly version。

If the specific reason is unclear, why is the dynamic in torch. onnx. export included in version 2.x The first axis of the input in axes has been set to "T", representing "dynamic", but it has not taken effect

[Jackiexiao]

pytorch 2.0+ bug: see: https://github.com/pytorch/pytorch/issues/99701