Skip to content

Latest commit

 

History

History
28 lines (24 loc) · 1.46 KB

019-20210212.md

File metadata and controls

28 lines (24 loc) · 1.46 KB

Thursday Feb 12 2021 5-6pm PST

Agenda

  • Greetings!
  • Discussion topics
    • Converters project updates
    • Roadmap discussion

Attendees

  • Chin Huang (onnx-tf, IBM)
  • Guenther Schmuelling (tf2onnx, Microsoft)
  • Kevin Chen (Tensor-RT)
  • Ting Su (Matlab)

Notes

  • ONNX-TF fully supports opset 13, except squeeze and unsqueeze where axes becoming an input, meaning a tensor instead list of ints. The issue is the corresponding TF APIs support only primitive types for axis.
  • TF2ONNX released TFLite support, being able to convert TFLite directly to ONNX, making a pass at Hugging face models, added SentencePiece as custom op
  • Tensor-RT is officially on opset 11, planning on support opset 13 within next release, working on QDQ tool
  • Matlab keeps working on operators, import in two ways, 1. functional form, up to opset 13 2. layer type of model, still working on some constraints
  • Would like to have Pytorch to ONNX converter updates in our SIG meetings
  • Frontend roadmap items:
    • Convert from Tensorflow-js vs onnx-js? Convert from google/jax, like jax2onnx vs jax to tf + tf2onnx?
  • Backend roadmap items:
    • Add NHWC option in ONNX, at the model level? input data format? and/or certain ops? Would be nice if the user of the model can easily know the data format.
    • Leverage Onnx model zoo models as standard test and verification for all backends