ONNX as Graph IR frontend for MLIR

[Zhipeng Huang]

Hi MLIR Team,

Is it possible to utilize ONNX as graph IR frontend for MLIR ? Or TF Graph would be the only assumed option ?

[Alex Zinenko]

Hi Zhipeng,

MLIR is designed to be an open infrastructure of dialects and operations.  As such, an ONNX graph dialect can be designed by any interested party.  Similarly to TensorFlow, it would make sense for it to live in the ONNX codebase.  If somebody is willing to implement it, we will be happy to discuss potential integrations.

Cheers,
Alex

On Wed, Apr 24, 2019 at 10:11 AM Zhipeng Huang <zhipe...@gmail.com> wrote:

Hi MLIR Team,

Is it possible to utilize ONNX as graph IR frontend for MLIR ? Or TF Graph would be the only assumed option ?

--
You received this message because you are subscribed to the Google Groups "MLIR" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mlir+uns...@tensorflow.org.
To view this discussion on the web visit https://groups.google.com/a/tensorflow.org/d/msgid/mlir/4605b1e6-0b1c-49b0-9cb0-c5809805d2ec%40tensorflow.org.

--

-- Alex

[Zhipeng Huang]

Thanks Alex,

Our team would be very interested in such integration. Part of the motivation is that we are currently designing a model zoo CI infra for ONNX and having MLIR could be very beneficial for better interoperability with backend accelerators.

Would be great to find a time to discuss in detail :)

On Wednesday, April 24, 2019 at 7:17:56 PM UTC+8, Alex Zinenko wrote:

Hi Zhipeng,

MLIR is designed to be an open infrastructure of dialects and operations.  As such, an ONNX graph dialect can be designed by any interested party.  Similarly to TensorFlow, it would make sense for it to live in the ONNX codebase.  If somebody is willing to implement it, we will be happy to discuss potential integrations.

Cheers,
Alex

On Wed, Apr 24, 2019 at 10:11 AM Zhipeng Huang <zhipe...@gmail.com> wrote:

Hi MLIR Team,

Is it possible to utilize ONNX as graph IR frontend for MLIR ? Or TF Graph would be the only assumed option ?

--
You received this message because you are subscribed to the Google Groups "MLIR" group.

To unsubscribe from this group and stop receiving emails from it, send an email to ml...@tensorflow.org.

To view this discussion on the web visit https://groups.google.com/a/tensorflow.org/d/msgid/mlir/4605b1e6-0b1c-49b0-9cb0-c5809805d2ec%40tensorflow.org.

--

-- Alex