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

Fix codeblock in quantization docs.

Browse Source 
PiperOrigin-RevId: 236043945
This commit is contained in:
Suharsh Sivakumar 2019-02-27 20:08:21 -08:00 committed by TensorFlower Gardener
parent f15121396c
commit 4eec977db7
1 changed files with 5 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
tensorflow/lite/g3doc/convert/quantization.md
View File

@ -28,12 +28,11 @@ Currently, this requires training a model with
Convert the graph:
```
with tf.Session() as sess:
converter = tf.lite.TFLiteConverter.from_saved_model(saved_model_dir)
converter.inference_type = tf.lite.constants.QUANTIZED_UINT8
input_arrays = converter.get_input_arrays()
converter.quantized_input_stats = {input_arrays[0] : (0., 1.)} # mean, std_dev
tflite_model = converter.convert()
converter = tf.lite.TFLiteConverter.from_saved_model(saved_model_dir)
converter.inference_type = tf.lite.constants.QUANTIZED_UINT8
input_arrays = converter.get_input_arrays()
converter.quantized_input_stats = {input_arrays[0] : (0., 1.)} # mean, std_dev
tflite_model = converter.convert()
```
For fully integer models, the inputs are uint8. The `mean` and `std_dev values`