From 0cc2e49fd96eca29761b323a16f41683fe112fc6 Mon Sep 17 00:00:00 2001 From: RJ Ascani Date: Tue, 2 Jul 2024 10:27:06 -0700 Subject: [PATCH] Disable per-channel quant on requantize test (#2616) BUG=cleanup --- tensorflow/lite/micro/tools/requantize_flatbuffer_test.py | 1 + 1 file changed, 1 insertion(+) diff --git a/tensorflow/lite/micro/tools/requantize_flatbuffer_test.py b/tensorflow/lite/micro/tools/requantize_flatbuffer_test.py index 342f893bf7a..885af89c7f4 100644 --- a/tensorflow/lite/micro/tools/requantize_flatbuffer_test.py +++ b/tensorflow/lite/micro/tools/requantize_flatbuffer_test.py @@ -62,6 +62,7 @@ def convert_tfl_converter(keras_model, converter.representative_dataset = representative_dataset_gen # TODO(b/324385802): Support per-channel quantization for FullyConnected. converter._experimental_disable_per_channel_quantization_for_dense_layers = True + converter._experimental_disable_per_channel = True return converter.convert()