Update comments to suggest current qat support

PiperOrigin-RevId: 367074181

Author: nutsiepully

tensorflow_model_optimization/python/core/quantization/keras/default_8bit/default_8bit_quantize_registry.py

@@ -116,9 +116,9 @@ class Default8BitQuantizeRegistry(
       _no_quantize(layers.ZeroPadding1D),
       _no_quantize(layers.ZeroPadding2D),
       # _no_quantize(layers.ZeroPadding3D),
-      # Enable once verified.
-      # layers.SeparableConv1D,
-      # layers.SeparableConv2D,
+
+      # Supported via modifications in Transforms.
+      # layers.SeparableConv1D, layers.SeparableConv2D,
 
       # Core Layers
       _no_quantize(layers.ActivityRegularization),
@@ -156,7 +156,7 @@ class Default8BitQuantizeRegistry(
       # layers.Embedding: ['embeddings'],
 
       # BatchNormalization is handled elsewhere, in the cases
-      # where it's preceded by convolutional layers or a Dense layer.
+      # where it's preceded by convolutional layers.
       #   layers.BatchNormalization: [],
 
       # Merge layers to be added.