onnx / onnx-r Goto Github PK

From Susan:

I see using protobufs to interface with ONNX http://onnx.ai/onnx-r/articles/protobufs.html- Thinking aloud - I wonder if ONNX will also support the format that Tensorflow Lite supports (Flatbuffers) instead of protobufs - and therefore will ONNX-R eventually support Flatbuffers.

From Chin:

As far as the FlatBuffer format, I don't see ONNX APIs directly support it in near future. But I did hear a suggestion to build a TF Liter converter in one of my ONNX Converters SIG meetings. Currently we are still waiting for takers to looking into the details. Meanwhile, one could convert between ONNX and FlatBuffer via ONNX <->Tensorflow converters.

There is a 404 error on the website documentation.

Hi, are there any plans to support saving models in ONNX format, similar to the PMML library? Something like:

library(onnx)

model <- lm(Species ~ ., iris)
onnx(model)

Is there a way you are supposed to set the path of python configuration for R?

library(onnx)
node_def <- make_node("Relu", list("X"), list("Y"))
Error: Python module onnx was not found.

Detected Python configuration:

python: /usr/bin/python
libpython: /usr/lib/python2.7/config-x86_64-linux-gnu/libpython2.7.so
pythonhome: /usr:/usr
version: 2.7.12 (default, Dec 4 2017, 14:50:18) [GCC 5.4.0 20160609]
numpy: /usr/lib/python2.7/dist-packages/numpy
numpy_version: 1.11.0
onnx: [NOT FOUND]

python versions found:
/usr/bin/python
/usr/bin/python3
/home/jichiang/anaconda3/bin/python

I think we can support training as well (especially for fine-tuning a pre-trained model). TensorFlow R API might be a good first candidate to target.

Hello ONNX partner,

Based on best practices at most of the ONNX member companies, the ONNX community is recommending that we rename the master branch in our repos as main.

GitHub has simplified the process as described at https://github.com/github/renaming. Please switch over once all dependencies are aware of the change.

Thanks!

Is there any plans to support conversion of r packages to ONNX format?