You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The format is a binary descriptor of the transforms and learner for the trained pipeline; create one and unzip it. As @GalOshri says, the ONNX output format is on the roadmap. We're always interested in what folks are looking for; any particular interest in ONNX? Or mainly just want to explore the output formats?
Adding to what @justinormont mentioned: the WriteAsync method saves the model as a zip file. You can then use ReadAsync to use the model in a different project.
Activity
GalOshri commentedon May 11, 2018
Not yet, but exporting ML.NET models to ONNX is on the roadmap. Is there a particular model/pipeline that you need to export to ONNX?
ChangweiZhang commentedon May 11, 2018
According to the API reference, the class PredictionModel have a WriteAsync method and I can write model to file in binary format. What is the format?
justinormont commentedon May 11, 2018
The format is a binary descriptor of the transforms and learner for the trained pipeline; create one and unzip it. As @GalOshri says, the ONNX output format is on the roadmap. We're always interested in what folks are looking for; any particular interest in ONNX? Or mainly just want to explore the output formats?
GalOshri commentedon May 11, 2018
Adding to what @justinormont mentioned: the WriteAsync method saves the model as a zip file. You can then use ReadAsync to use the model in a different project.
turowicz commentedon May 11, 2018
Interoperability with models trained in on different stack?
GalOshri commentedon May 12, 2018
@turowicz which stack do you have in mind?
enghch commentedon May 15, 2018
I'd like onnx to make it easy to consume my model with Windows ML to get the automatic hardware acceleration detection.
Fail gracefully if unable to instantiate data view with swept paramet…
code migration from machinelearning-automl to machinelearning repo fe…
Fail gracefully if unable to instantiate data view with swept paramet…
Added onnx export functionality for LpNormNormalizingTransformer (#4161)
Added Onnx export functionality to PCATransformer (#4188)