5.17. Model Runtime

3.2.0

New features

  • Support for dynamic batching in ModelRunner. Adds the ability to support dynamic batching with any batch size input data.

  • Support for conditional model execution in compiled graphs that contain more than one model. See Conditional execution in the Model Runtime User Guide.

Bug Fixes

  • Some errors could cause Model Runtime to hang instead of raising an error. This has been fixed.

Other improvements

None

Known issues

None

Compatibility changes

None

3.1.0

New features

  • Initial publication of API documentation and user guide for the Model Runtime library used by the Poplar Triton Backend and Graphcore’s version of TensorFlow Serving.

    Enables running of compiled PopEF models using a C++ or Python API and the native IPU runtime.

Bug Fixes

None

Other improvements

None

Known issues

None

Compatibility changes

None