experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Edit Hexagon documentation to reflect new supported models

Browse Source 
PiperOrigin-RevId: 312144610
Change-Id: I9c8b0d9ad6ea4b745b4bb985ca143cca660a5b14
This commit is contained in:
Sachin Joglekar 2020-05-18 13:36:18 -07:00 committed by TensorFlower Gardener
parent 869920697b
commit da67fcddef
1 changed files with 9 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
tensorflow/lite/g3doc/performance/hexagon_delegate.md
View File

@ -22,15 +22,15 @@ are supported, including:
**Supported models:** **Supported models:**
The Hexagon delegate currently supports quantized models generated using The Hexagon delegate supports all models that conform to our
[quantization-aware training](https://github.com/tensorflow/tensorflow/tree/r1.13/tensorflow/contrib/quantize), [8-bit symmetric quantization spec](https://www.tensorflow.org/lite/performance/quantization_spec),
e.g., including those generated using
[these quantized models](https://www.tensorflow.org/lite/guide/hosted_models#quantized_models) [post-training integer quantization](https://www.tensorflow.org/lite/performance/post_training_integer_quant).
hosted on the TensorFlow Lite repo. It does not (yet) support models with UInt8 models trained with the legacy
[8-bit symmetric quantization spec](https://www.tensorflow.org/lite/performance/quantization_spec). [quantization-aware training](https://github.com/tensorflow/tensorflow/tree/r1.13/tensorflow/contrib/quantize)
Sample models include path are also supported, for e.g.,
[MobileNet V1](https://storage.googleapis.com/download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_1.0_224_quant.tgz), [these quantized versions](https://www.tensorflow.org/lite/guide/hosted_models#quantized_models)
[SSD Mobilenet](https://storage.googleapis.com/download.tensorflow.org/models/tflite/coco_ssd_mobilenet_v1_1.0_quant_2018_06_29.zip). on our Hosted Models page.
## Hexagon Delegate Java API ## Hexagon Delegate Java API
@ -254,10 +254,6 @@ ro.board.platform`).
## FAQ ## FAQ
* Will the delegate support models created using
[post-training quantization](https://www.tensorflow.org/lite/performance/post_training_quantization)?
* This is tentatively planned for a future release, though there is no
concrete timeline.
* Which ops are supported by the delegate? * Which ops are supported by the delegate?
* See the current list of [supported ops and constraints](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/experimental/delegates/hexagon/README.md) * See the current list of [supported ops and constraints](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/experimental/delegates/hexagon/README.md)
* How can I tell that the model is using the DSP when I enable the delegate? * How can I tell that the model is using the DSP when I enable the delegate?