Change layer.get_weights()[0] to layer.embeddings in the warmstart_embedding_matrix tutorial.

This prevents an error when the vocab is large and the embedding tensor gets sharded.

PiperOrigin-RevId: 542046545

Author: tensorflower-gardener

site/en/tutorials/text/warmstart_embedding_matrix.ipynb

@@ -525,7 +525,7 @@
       "outputs": [],
       "source": [
         "embedding_weights_base = (\n",
-        "    model.get_layer(\"text_input\").get_layer(\"embedding\").get_weights()[0]\n",
+        "    model.get_layer(\"text_input\").get_layer(\"embedding\").embeddings\n",
         ")\n",
         "vocab_base = vectorize_layer.get_vocabulary()"
       ]
@@ -681,7 +681,7 @@
         "\n",
         "# Verify the shape of updated weights\n",
         "# The new weights shape should reflect the new vocabulary size\n",
-        "text_input_new.get_layer(\"embedding\").get_weights()[0].shape"
+        "text_input_new.get_layer(\"embedding\").embeddings.shape"
       ]
     },
     {