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

Make some "experimental" feature flag private

Browse Source 
And also removes all the related api documentation.

PiperOrigin-RevId: 303006102
Change-Id: Id135e802e6007e98f30e48a8232a51f7150c7aaa
This commit is contained in:
Feng Liu 2020-03-25 17:21:29 -07:00 committed by Feng Liu
parent b6016957b3
commit e6895b3648
2 changed files with 5 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
tensorflow/lite/python/lite.py
View File

@ -174,8 +174,8 @@ class TFLiteConverterBase(object):
self.optimizations = [] self.optimizations = []
self.representative_dataset = None self.representative_dataset = None
self.experimental_new_converter = _USE_EXPERIMENTAL_NEW_CONVERTER self.experimental_new_converter = _USE_EXPERIMENTAL_NEW_CONVERTER
self.experimental_new_quantizer = False self._experimental_new_quantizer = False
self.experimental_calibrate_only = False self._experimental_calibrate_only = False
# The 'GraphDebugInfo' contains the stack traces of all the original nodes # The 'GraphDebugInfo' contains the stack traces of all the original nodes
# in the `GraphDef` to the converter. # in the `GraphDef` to the converter.
self._debug_info = None self._debug_info = None
@ -259,7 +259,7 @@ class TFLiteConverterBase(object):
inference_output_type): inference_output_type):
allow_float = not self._is_int8_target_required() allow_float = not self._is_int8_target_required()
calibrate_quantize = _calibrator.Calibrator(result) calibrate_quantize = _calibrator.Calibrator(result)
if self.experimental_calibrate_only: if self._experimental_calibrate_only:
return calibrate_quantize.calibrate(self.representative_dataset.input_gen) return calibrate_quantize.calibrate(self.representative_dataset.input_gen)
else: else:
return calibrate_quantize.calibrate_and_quantize( return calibrate_quantize.calibrate_and_quantize(
@ -317,11 +317,6 @@ class TFLiteConverterV2(TFLiteConverterBase):
dataset to evaluate different optimizations. dataset to evaluate different optimizations.
experimental_new_converter: Experimental flag, subject to change. experimental_new_converter: Experimental flag, subject to change.
Enables MLIR-based conversion instead of TOCO conversion. Enables MLIR-based conversion instead of TOCO conversion.
experimental_new_quantizer: Experimental flag, subject to change.
Enables MLIR-based post-training quantization.
experimental_calibrate_only: Experimental flag, subject to change.
Calibrates the converted model with representative dataset, but not
quantize it.
Example usage: Example usage:
```python ```python
@ -609,11 +604,6 @@ class TFLiteConverter(TFLiteConverterBase):
the dataset to evaluate different optimizations. the dataset to evaluate different optimizations.
experimental_new_converter: Experimental flag, subject to change. experimental_new_converter: Experimental flag, subject to change.
Enables MLIR-based conversion instead of TOCO conversion. Enables MLIR-based conversion instead of TOCO conversion.
experimental_new_quantizer: Experimental flag, subject to change.
Enables MLIR-based post-training quantization.
experimental_calibrate_only: Experimental flag, subject to change.
Calibrates the converted model with representative dataset, but not
quantize it.
Example usage: Example usage:
```python ```python

4
tensorflow/lite/python/lite_v2_test.py
View File

@ -296,11 +296,11 @@ class FromConcreteFunctionTest(TestModels):
quantized_converter.representative_dataset = calibration_gen quantized_converter.representative_dataset = calibration_gen
# default quantizer # default quantizer
quantized_converter.experimental_new_quantizer = False quantized_converter._experimental_new_quantizer = False
old_tflite = quantized_converter.convert() old_tflite = quantized_converter.convert()
# new quantizer # new quantizer
quantized_converter.experimental_new_quantizer = True quantized_converter._experimental_new_quantizer = True
new_tflite = quantized_converter.convert() new_tflite = quantized_converter.convert()
for _ in range(5): for _ in range(5):