Tf-lite conversion is fixed.

PiperOrigin-RevId: 352546075

examples/colab/tf2_image_retraining.ipynb

@@ -394,8 +394,6 @@
       "outputs": [],
       "source": [
         "#@title Optimization settings\n",
-        "# docs_infra: no_execute\n",
-        "# TODO(b/156102192)\n",
         "optimize_lite_model = False  #@param {type:\"boolean\"}\n",
         "#@markdown Setting a value greater than zero enables quantization of neural network activations. A few dozen is already a useful amount.\n",
         "num_calibration_examples = 60  #@param {type:\"slider\", min:0, max:1000, step:1}\n",
@@ -428,7 +426,6 @@
       },
       "outputs": [],
       "source": [
-        "# docs_infra: no_execute\n",
         "interpreter = tf.lite.Interpreter(model_content=lite_model_content)\n",
         "# This little helper wraps the TF Lite interpreter as a numpy-to-numpy function.\n",
         "def lite_model(images):\n",
@@ -447,7 +444,6 @@
       "outputs": [],
       "source": [
         "#@markdown For rapid experimentation, start with a moderate number of examples.\n",
-        "# docs_infra: no_execute\n",
         "num_eval_examples = 50  #@param {type:\"slider\", min:0, max:700}\n",
         "eval_dataset = ((image, label)  # TFLite expects batch size 1.\n",
         "                for batch in train_generator\n",