Xavier Dupré

I started to look into that here: [Use function when converting into ONNX](http://www.xavierdupre.fr/app/mlprodict/helpsphinx/notebooks/onnx_sklearn_functions.html). Every piece of a pipeline is exported as a function. It stores the hyperparameters as well. It...

I was trying to replicate in #572 but in my tiny example, no rows are lost. Do you know which rows are missing? I would also need the version of...

There was very limited maintenance on sparkml converters. Unit tests were disabled some time ago. Fixing this issue may require quite some work.

This issue is connected to #446. There is no easy way to fix them right now. This is a work in progress without any fixed deadline.

I wrote a PR to remove unnecessary Identity node in a graph. https://github.com/microsoft/onnxconverter-common/pull/21, we may modify the convert function to use it. It does not solve the inner issue but...

https://github.com/onnx/tensorflow-onnx should be used to convert tensorflow models. The documentation is better there.

HashingVectorizer requires the support of sparse feature. It is part of ONNX spec but not yet part of onnxruntime. We can work on a converter but not easily test it...

This task requires a couple of items: add a new operator in ONNX (we can skip that for demo purpose), implement a new operator in onnxruntime, add support for sparse...

Most of the model are supported. The whole list is available in the documentation: http://onnx.ai/sklearn-onnx/supported.html. Only sparse is not supported yet.

You should use [tensorflow-onnx](https://github.com/onnx/tensorflow-onnx).