[Discussion] How to get the onnx format of the model？

[mainwindow]

In order to convert the model to onnx format, I heard that saved_model format is required.
Execute in the command line: python -m spleeter separate -p spleeter:4stems -o output audio_example.mp3, and the resulting directory structure is as follows:
spleeter/pretrained_models/4stems/
2025/05/27 20:04 .probe
2019/10/24 17:36 ._checkpoint
2019/10/24 17:36 checkpoint
2019/10/24 17:35 model.data-00000-of-00001
2019/10/24 17:35 model.index
2019/10/24 17:35 model.meta

The saved_model format file was not found. Does converting a file to onnx format necessarily require a model in saved_model format?

There is another method to convert to onnx?

[csukuangfj]

Please have a look at
https://k2-fsa.github.io/sherpa/onnx/source-separation/models.html#spleeter

We have provided onnx models for 2-stem spleeter.

Also, there is C++ API for it.

@mainwindow