Release note 1.28.0 (#13443)

This commit is release note for 1.28.0.

ONE-DCO-1.0-Signed-off-by: Hyeongseok Oh <hseok82.oh@samsung.com>
Co-authored-by: Jiyoung Giuliana Yun <jy910.yun@samsung.com>

docs/release/1.28/index.rst

@@ -0,0 +1,13 @@
+.. ONE documentation master file, created by
+   sphinx-quickstart on Wed Sep 06 12:18:12 2022.
+   You can adapt this file completely to your liking, but it should at least
+   contain the root `toctree` directive.
+
+1.28
+====
+
+.. toctree::
+   :maxdepth: 2
+   :caption: Contents:
+
+  ./release-note-1.28.0.md

docs/release/1.28/release-note_1.28.0.md

@@ -0,0 +1,43 @@
+# Release Note 1.28.0
+
+## ONE Runtime
+
+### Python API
+
+- Support experimental python API
+  - Refer howto document for more details
+
+### On-device Training
+
+- Support on-device training with circle model
+  - Training parameter can be passed to _onert_ via _onert_`s experimental API or loading new model format including training information: _circle_plus_
+  - Trained model can be exported to _circle_ model via experimental API `nnfw_train_export_circle`
+  - Supporting Transfer learning from a pre-trained circle model
+- Introduce _circle_plus_gen_ tool
+  - Generates a _circle_plus_ model file with a given training hyperparameters
+  - Shows a _circle_plus model details
+
+### Runtime configuration API
+- _onert_ supports runtime configuration API for prepare and execution phase via experimental APIs
+  - `nnfw_set_prepare_config` sets configuration for prepare phase, and `nnfw_reset_prepare_config` resets it to default value
+  - `nnfw_set_execution_config` sets configuration for execution phase, and `nnfw_reset_execution_config` resets it to default value
+  - Supporting prepare phase configuration: prepare execution time profile
+  - Supporting execution phase configuration: dump minmax data, dump execution trace, dump execution time
+- Introduce new API to set _onert_ workspace directory: `nnfw_set_workspace`
+  - _onert_ workspace directory is used to store intermediate files during prepare and execution phase
+
+### Minmax Recorder
+- Now _onert_'s minmax recorder dumps raw file format instead of HDF5 format
+- _onert_ dumps minmax data into workspace directory
+
+### On-device Compilation
+- _onert_ supports full quantization of uint8/int16 type weight and activation.
+  - To quantize activation, _onert_ requires minmax data of activation.
+- _onert_ supports on-device code generation for special backend requiring special binary format such as DSP, NPU.
+  - Introduce new experimental API for code generation: `nnfw_codegen`
+
+### Type-aware model I/O usage
+- If loaded model is quantized model, _onert_ allows float type I/O buffer
+  - _onert_ converts float type input buffer to quantized type internally
+  - _onert_ fills float type output buffers by converting quantized type output data to float type internally
+- On multimodel package, _onert_ allows edges between quantized model and float type model