experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
rmothukuru-patch-1
STT-tensorflow/tensorflow/lite/testing/model_coverage
History
Feng Liu 201ac54c8d Make the experimental_new_quantizer flag to be public 
PiperOrigin-RevId: 351421198
Change-Id: I90a1df2a7d6ad074cc7f0d1d04a005a16580fb4c
2021-01-12 12:35:36 -08:00
..
testdata Quantization tests: compare TFLite model output agains pre-computed golden values. 2020-12-16 23:26:25 -08:00
BUILD Remove keras dependency from TFLite tests and BUILD files 2020-10-31 10:02:05 -07:00
model_coverage_lib_test.py Remove keras dependency from TFLite tests and BUILD files 2020-10-31 10:02:05 -07:00
model_coverage_lib.py Make the experimental_new_quantizer flag to be public 2021-01-12 12:35:36 -08:00
README.md Quantization tests: compare TFLite model output agains pre-computed golden values. 2020-12-16 23:26:25 -08:00

README.md

TensorFlow Lite model coverage tests

Various conversion tests on popular mobile models.

Golden values

Some tests rely on pre-computed golden values. The main goal is to detect changes affecting unintended parts of TFLite.

Should a golden value test fail after an intended change, the golden values can be updated with the following command:

bazel run //third_party/tensorflow/lite/testing/model_coverage:<target> --test_output=all -- --update_goldens

Notice bazel run instead of bazel test and the addition of the --update_golden flag.

The updated golden data files must then be included in the change list.