Clarify that currently 8-bit quantization is the only supported deployment path.

PiperOrigin-RevId: 378332718

Author: fredrec

tensorflow_model_optimization/g3doc/guide/quantization/training.md

@@ -37,8 +37,10 @@ cases. The code currently supports a
 ### Experiment with quantization and associated hardware
 
 Users can configure the quantization parameters (e.g. number of bits) and to
-some degree, the underlying algorithms. With these changes from the API
-defaults, there is no supported path to deployment.
+some degree, the underlying algorithms. Note that with these changes from the
+API defaults, there is currently no supported path for deployment to a backend.
+For instance, TFLite conversion and kernel implementations only support 8-bit
+quantization.
 
 APIs specific to this configuration are experimental and not subject to backward
 compatibility.

tensorflow_model_optimization/g3doc/guide/quantization/training_comprehensive_guide.ipynb

@@ -496,8 +496,9 @@
       },
       "source": [
         "**Your use case**: using the following APIs means that there is no\n",
-        "supported path to deployment. The features are also experimental and not\n",
-        "subject to backward compatibility.\n",
+        "supported path to deployment. For instance, TFLite conversion\n",
+        "and kernel implementations only support 8-bit quantization.\n",
+        "The features are also experimental and not subject to backward compatibility.\n",
         "  * `tfmot.quantization.keras.QuantizeConfig`\n",
         "  * `tfmot.quantization.keras.quantizers.Quantizer`\n",
         "  * `tfmot.quantization.keras.quantizers.LastValueQuantizer`\n",