Nengo-edge

===================

======================

**Added**

- Updated documentation with new content (additional tutorials, CLI documentation,
FAQ section). (`9`_)
- Added support for input string processing and NLP model inference in
``SavedModelRunner``. (`17`_)
- Added ``NetworkTokenizer`` class to perform remote calls to a device
CLI that supports sentencepiece tokenization. (`17`_)
- Added stdio-based CLI runner for ``np_mfcc``. (`19`_)

**Changed**

- ``SavedModelRunner`` tokenizer now uses ``SentencepieceTokenizer`` instead of
``FastSentencepieceTokenizer`` to ensure compatibility with the core sentencepiece
library. (`17`_)
- Moved ``device_modules/network_tokenizer.py`` to ``network_runner.py``. (`19`_)

**Fixed**

- Fixed model output decoding for ASR. ``SavedModelRunner`` now removes blank
tokens and merges repeating tokens before detokenization. (`17`_)

**Removed**

- Removed support for streaming in ``SavedModelRunner``. (`19`_)

.. _9: https://github.com/nengo/nengo-edge/pull/9
.. _17: https://github.com/nengo/nengo-edge/pull/17
.. _19: https://github.com/nengo/nengo-edge/pull/19

============================

**Added**

- Added warning when a downloaded nengo-edge model artifacts' version does not
match local environment nengo-edge version. (`6`_)
- Added ``nengo-edge package-dataset`` CLI command, which can be used to validate
and package KWS and ASR datasets. (`10`_)
- ``SavedModelRunner.run`` now automatically decodes ASR model outputs via the exported
sentencepiece tokenizer. (`15`_)

**Changed**

- ``SavedModelRunner`` now uses Keras' SavedModel format, instead of the raw
TensorFlow SavedModel format. (`8`_)
- ``SavedModelRunner`` can now take ragged ``object``-arrays as input. (`8`_)
- TensorFlow is now an optional dependency installed with
``pip install nengo-edge[optional]``. (`13`_)

.. _6: https://github.com/nengo/nengo-edge/pull/6
.. _8: https://github.com/nengo/nengo-edge/pull/8
.. _10: https://github.com/nengo/nengo-edge/pull/10
.. _13: https://github.com/nengo/nengo-edge/pull/13
.. _15: https://github.com/nengo/nengo-edge/pull/15
.. _17: https://github.com/nengo/nengo-edge/pull/17

=======================

**Added**

- Added ``CoralRunner`` for running models exported for the Coral board. (`4`_)
- Added ``DiscoRunner`` for running models exported for the Disco board. (`4`_)
- Added ``NordicRunner`` for running models exported for the Nordic board. (`4`_)
- Added on-device MFCC extraction code
(``device_modules.np_mfcc.LogMelFeatureExtractor``). (`4`_)
- Added two new examples demonstrating how to run models exported for the
Coral/Disco/Nordic devices. (`4`_)

**Changed**

- Renamed ``tflite_runner.Runner`` to ``TFLiteRunner``. (`4`_)
- Renamed ``saved_model_runner.Runner`` to ``SavedModelRunner``. (`4`_)
- ``TFLiteRunner.reset_state`` now takes a ``batch_size`` argument, which can be used
to prepare the model to run with different batch sizes. (`5`_)

.. _4: https://github.com/nengo/nengo-edge/pull/4
.. _5: https://github.com/nengo/nengo-edge/pull/5

===========================

**Fixed**

- Fixed an issue causing pip to refuse to install ``nengo-edge``. (`3`_)

.. _3: https://github.com/nengo/nengo-edge/pull/3

==========================

Initial release