diff options
Diffstat (limited to 'docs/release/release_note_1.1.0.md')
| -rw-r--r-- | docs/release/release_note_1.1.0.md | 40 |
1 files changed, 40 insertions, 0 deletions
diff --git a/docs/release/release_note_1.1.0.md b/docs/release/release_note_1.1.0.md new file mode 100644 index 000000000..f267c4cbe --- /dev/null +++ b/docs/release/release_note_1.1.0.md @@ -0,0 +1,40 @@ +# NNAS 1.1.0 Release Note + +## Feature Highlights + +- `nncc` + - Available for Tizen Studio in Windows + - Available for Visual Studio in Windows +- `nnfw` + - Interpreter supports more operations + - CPU Arithmetic kernels support broadcasing + - Fully Connected Operation supports hybrid quantization + + +## nncc + +### Available for Tizen Studio in Windows +We now support `nncc` in Tizen Studio as plugin. For detailed information and simple tutorial, please refer to [NNCC Installation Guide for Tizen Studio](../nncc/v1.1.0/nncc_in_tizen_studio). + +#### Known Issues +- Output directory of nnpackage is fixed to `res/shared`. + +### Available for Visual Studio in Windows +We now support `nncc` in Visual Studio as Tizen extension program. For detailed information and simple tutorial, please refer to [NNCC Installation Guide for Visual Studio](../nncc/v1.1.0/nncc_in_visual_studio). + +#### Known Issues +- `nncc` in Visual Studio extension program only accepts `model.pb` and `model.info` in `model` folder. + - If user want to create nnpackage using `model2.pb` and `model2.info`, user should change the names as `model.pb` and `model.info` first. +- Output directory of nnpackage is fixed to `res/shared/model`. + +## nnfw + + ### The following operations are supported on Interpreter : + - Activation : Relu, Relu1, Relu6, Tanh + - Logistics + - Gather + - Instance Normalization + - Transpose Convolution + ### CPU Arithmetic kernels support broadcasing + ### Fully Connected Operation supports hybrid quantization + Note that this support is only for acl_neon backend. See [hybrid quatization document](https://www.tensorflow.org/lite/performance/post_training_quantization#weight_quantization) for more about hybrid quantization.
\ No newline at end of file |