Remove per-channel flag from QuantizationConfig in MobileNetv2 Keras …

…notebook (sony#915) Co-authored-by: Ofir Gordon <[email protected]>
elad-c · Jan 3, 2024 · 0782d2e · 0782d2e
1 parent bca5634
commit 0782d2e
Showing 1 changed file with 0 additions and 2 deletions.
diff --git a/tutorials/notebooks/example_keras_imagenet.ipynb b/tutorials/notebooks/example_keras_imagenet.ipynb
@@ -337,13 +337,11 @@
     "\n",
     "# Set the following quantization configurations:\n",
     "# Choose the desired QuantizationErrorMethod for the quantization parameters search.\n",
-    "# Enable per-channel weights quantization.\n",
     "# Enable weights bias correction induced by quantization.\n",
     "# Enable shift negative corrections for improving 'signed' non-linear functions quantization (such as swish, prelu, etc.) \n",
     "# Set the threshold to filter outliers with z_score of 16. \n",
     "q_config = mct.QuantizationConfig(activation_error_method=QuantizationErrorMethod.MSE,\n",
     "                                  weights_error_method=QuantizationErrorMethod.MSE,\n",
-    "                                  weights_per_channel_threshold=True,\n",
     "                                  weights_bias_correction=True,\n",
     "                                  shift_negative_activation_correction=True,\n",
     "                                  z_threshold=16)\n",