diff --git a/tensorflow/lite/g3doc/_book.yaml b/tensorflow/lite/g3doc/_book.yaml index c74057e193c..b9c4fea74e9 100644 --- a/tensorflow/lite/g3doc/_book.yaml +++ b/tensorflow/lite/g3doc/_book.yaml @@ -76,13 +76,12 @@ upper_tabs: path: /lite/performance/model_optimization - title: "Post-training quantization" path: /lite/performance/post_training_quantization - - title: "Post-training quantization example" - path: https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/g3doc/performance/post_training_quant.ipynb - - title: "Post-training integer quantization example" - path: https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/g3doc/performance/post_training_integer_quant.ipynb - - title: "Post-training float16 quantization example" - path: https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/g3doc/performance/post_training_float16_quant.ipynb - status: external + - title: "Post-training weight quantization" + path: /lite/performance/post_training_quant + - title: "Post-training integer quantization" + path: /lite/performance/post_training_integer_quant + - title: "Post-training float16 quantization" + path: /lite/performance/post_training_float16_quant - title: "Delegates" path: /lite/performance/delegates - title: "GPU delegate" diff --git a/tensorflow/lite/g3doc/performance/post_training_float16_quant.ipynb b/tensorflow/lite/g3doc/performance/post_training_float16_quant.ipynb index 22246f2edee..87f508165b8 100644 --- a/tensorflow/lite/g3doc/performance/post_training_float16_quant.ipynb +++ b/tensorflow/lite/g3doc/performance/post_training_float16_quant.ipynb @@ -3,7 +3,7 @@ "nbformat_minor": 0, "metadata": { "colab": { - "name": "post-training-fp16-quant.ipynb", + "name": "post_training-float16-quant.ipynb", "version": "0.3.2", "provenance": [], "private_outputs": true, @@ -11,11 +11,45 @@ "toc_visible": true }, "kernelspec": { - "display_name": "Python 2", - "name": "python2" + "name": "python3", + "display_name": "Python 3" } }, "cells": [ + { + "cell_type": "markdown", + "metadata": { + "id": "c8Cx-rUMVX25", + "colab_type": "text" + }, + "source": [ + "##### Copyright 2019 The TensorFlow Authors." + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "I9sUhVL_VZNO", + "colab_type": "code", + "colab": {}, + "cellView": "form" + }, + "source": [ + "#@title Licensed under the Apache License, Version 2.0 (the \"License\");\n", + "# you may not use this file except in compliance with the License.\n", + "# You may obtain a copy of the License at\n", + "#\n", + "# https://www.apache.org/licenses/LICENSE-2.0\n", + "#\n", + "# Unless required by applicable law or agreed to in writing, software\n", + "# distributed under the License is distributed on an \"AS IS\" BASIS,\n", + "# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n", + "# See the License for the specific language governing permissions and\n", + "# limitations under the License." + ], + "execution_count": 0, + "outputs": [] + }, { "cell_type": "markdown", "metadata": { @@ -23,22 +57,25 @@ "id": "6Y8E0lw5eYWm" }, "source": [ - "# Post Training FP16 Quantization" + "# Post-training float16 quantization" ] }, { "cell_type": "markdown", "metadata": { - "colab_type": "text", - "id": "CIGrZZPTZVeO" + "id": "CGuqeuPSVNo-", + "colab_type": "text" }, "source": [ "\n", " \n", " \n", + " \n", "
\n", - " Run in Google Colab\n", + " View on TensorFlow.org\n", " \n", - " View source on GitHub\n", + " Run in Google Colab\n", + " \n", + " View source on GitHub\n", "
" ] @@ -55,11 +92,10 @@ "[TensorFlow Lite](https://www.tensorflow.org/lite/) now supports\n", "converting weights to 16-bit floating point values during model conversion from TensorFlow to TensorFlow Lite's flat buffer format. This results in a 2x reduction in model size. Some harware, like GPUs, can compute natively in this reduced precision arithmetic, realizing a speedup over traditional floating point execution. The Tensorflow Lite GPU delegate can be configured to run in this way. However, a model converted to float16 weights can still run on the CPU without additional modification: the float16 weights are upsampled to float32 prior to the first inference. This permits a significant reduction in model size in exchange for a minimal impacts to latency and accuracy.\n", "\n", - "In this tutorial, we train an MNIST model from scratch, check its accuracy in TensorFlow, and then convert the saved model into a Tensorflow Lite flatbuffer\n", - "with float16 quantization. We finally check the\n", - "accuracy of the converted model and compare it to the original saved model. We\n", - "run the training script [mnist.py](https://github.com/tensorflow/models/blob/master/official/mnist/mnist.py) from\n", - "[Tensorflow official MNIST tutorial](https://github.com/tensorflow/models/tree/master/official/mnist).\n" + "In this tutorial, you train an MNIST model from scratch, check its accuracy in TensorFlow, and then convert the saved model into a Tensorflow Lite flatbuffer\n", + "with float16 quantization. Finally, check the\n", + "accuracy of the converted model and compare it to the original saved model. The training script, `mnist.py`, is available from the\n", + "[TensorFlow official MNIST tutorial](https://github.com/tensorflow/models/tree/master/official/mnist).\n" ] }, { @@ -69,7 +105,7 @@ "id": "2XsEP17Zelz9" }, "source": [ - "## Building an MNIST model" + "## Build an MNIST model" ] }, { @@ -105,7 +141,11 @@ }, "source": [ "import tensorflow as tf\n", - "tf.enable_eager_execution()" + "tf.enable_eager_execution()\n", + "\n", + "import numpy as np\n", + "\n", + "tf.logging.set_verbosity(tf.logging.DEBUG)" ], "execution_count": 0, "outputs": [] @@ -203,8 +243,7 @@ "id": "5NMaNZQCkW9X" }, "source": [ - "For the example, we only trained the model for a single epoch, so it only trains to ~96% accuracy.\n", - "\n" + "For the example, you trained the model for just a single epoch, so it only trains to ~96% accuracy." ] }, { @@ -253,10 +292,6 @@ "colab": {} }, "source": [ - "import tensorflow as tf\n", - "tf.enable_eager_execution()\n", - "tf.logging.set_verbosity(tf.logging.DEBUG)\n", - "\n", "converter = tf.lite.TFLiteConverter.from_saved_model(saved_model_dir)\n", "tflite_model = converter.convert()" ], @@ -391,8 +426,7 @@ "id": "-5l6-ciItvX6" }, "source": [ - "We can run the TensorFlow Lite model using the Python TensorFlow Lite\n", - "Interpreter. \n", + "Run the TensorFlow Lite model using the Python TensorFlow Lite Interpreter. \n", "\n", "### Load the test data\n", "\n", @@ -407,7 +441,6 @@ "colab": {} }, "source": [ - "import numpy as np\n", "_, mnist_test = tf.keras.datasets.mnist.load_data()\n", "images, labels = tf.cast(mnist_test[0], tf.float32)/255.0, mnist_test[1]\n", "\n", @@ -585,7 +618,10 @@ "colab": {} }, "source": [ - "print(eval_model(interpreter, mnist_ds))" + "# Create smaller dataset for demonstration purposes\n", + "mnist_ds_demo = mnist_ds.take(2000)\n", + "\n", + "print(eval_model(interpreter, mnist_ds_demo))" ], "execution_count": 0, "outputs": [] @@ -597,7 +633,7 @@ "id": "Km3cY9ry8ZlG" }, "source": [ - "We can repeat the evaluation on the float16 quantized model to obtain:\n" + "Repeat the evaluation on the float16 quantized model to obtain:" ] }, { @@ -612,7 +648,7 @@ "# doesn't have super optimized server CPU kernels. For this reason this may be\n", "# slower than the above float interpreter. But for mobile CPUs, considerable\n", "# speedup can be observed.\n", - "print(eval_model(interpreter_fp16, mnist_ds))\n" + "print(eval_model(interpreter_fp16, mnist_ds_demo))" ], "execution_count": 0, "outputs": [] @@ -624,7 +660,7 @@ "id": "L7lfxkor8pgv" }, "source": [ - "In this example, we have quantized a model to float16 with no difference in the accuracy.\n", + "In this example, you have quantized a model to float16 with no difference in the accuracy.\n", "\n", "It's also possible to evaluate the fp16 quantized model on the GPU. To perform all arithmetic with the reduced precision values, be sure to create the `TfLiteGPUDelegateOptions` struct in your app and set `precision_loss_allowed` to `1`, like this:\n", "\n", diff --git a/tensorflow/lite/g3doc/performance/post_training_integer_quant.ipynb b/tensorflow/lite/g3doc/performance/post_training_integer_quant.ipynb index 1629b4c9af3..c7150114e01 100644 --- a/tensorflow/lite/g3doc/performance/post_training_integer_quant.ipynb +++ b/tensorflow/lite/g3doc/performance/post_training_integer_quant.ipynb @@ -3,7 +3,7 @@ "nbformat_minor": 0, "metadata": { "colab": { - "name": "post-training--integer-quant.ipynb", + "name": "post_training_integer_quant.ipynb", "version": "0.3.2", "provenance": [], "private_outputs": true, @@ -11,11 +11,45 @@ "toc_visible": true }, "kernelspec": { - "display_name": "Python 2", - "name": "python2" + "name": "python3", + "display_name": "Python 3" } }, "cells": [ + { + "cell_type": "markdown", + "metadata": { + "id": "_DDaAex5Q7u-", + "colab_type": "text" + }, + "source": [ + "##### Copyright 2019 The TensorFlow Authors." + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "W1dWWdNHQ9L0", + "colab_type": "code", + "colab": {}, + "cellView": "form" + }, + "source": [ + "#@title Licensed under the Apache License, Version 2.0 (the \"License\");\n", + "# you may not use this file except in compliance with the License.\n", + "# You may obtain a copy of the License at\n", + "#\n", + "# https://www.apache.org/licenses/LICENSE-2.0\n", + "#\n", + "# Unless required by applicable law or agreed to in writing, software\n", + "# distributed under the License is distributed on an \"AS IS\" BASIS,\n", + "# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n", + "# See the License for the specific language governing permissions and\n", + "# limitations under the License." + ], + "execution_count": 0, + "outputs": [] + }, { "cell_type": "markdown", "metadata": { @@ -23,7 +57,7 @@ "id": "6Y8E0lw5eYWm" }, "source": [ - "# Post Training Integer Quantization" + "# Post-training integer quantization" ] }, { @@ -35,10 +69,13 @@ "source": [ "\n", " \n", " \n", + " \n", "
\n", - " Run in Google Colab\n", + " View on TensorFlow.org\n", " \n", - " View source on GitHub\n", + " Run in Google Colab\n", + " \n", + " View source on GitHub\n", "
" ] @@ -58,11 +95,10 @@ "In contrast to [post-training \"on-the-fly\" quantization](https://colab.sandbox.google.com/github/tensorflow/tensorflow/blob/master/tensorflow/lite/tutorials/post_training_quant.ipynb)\n", ", which only stores weights as 8-bit ints, in this technique all weights *and* activations are quantized statically during model conversion.\n", "\n", - "In this tutorial, we train an MNIST model from scratch, check its accuracy in TensorFlow, and then convert the saved model into a Tensorflow Lite flatbuffer\n", - "with full quantization. We finally check the\n", - "accuracy of the converted model and compare it to the original saved model. We\n", - "run the training script [mnist.py](https://github.com/tensorflow/models/blob/master/official/mnist/mnist.py) from\n", - "[Tensorflow official MNIST tutorial](https://github.com/tensorflow/models/tree/master/official/mnist).\n" + "In this tutorial, you train an MNIST model from scratch, check its accuracy in TensorFlow, and then convert the saved model into a Tensorflow Lite flatbuffer\n", + "with full quantization. Finally, check the\n", + "accuracy of the converted model and compare it to the original saved model. The training script, `mnist.py`, is available from the\n", + "[TensorFlow official MNIST tutorial](https://github.com/tensorflow/models/tree/master/official/mnist).\n" ] }, { @@ -72,7 +108,7 @@ "id": "2XsEP17Zelz9" }, "source": [ - "## Building an MNIST model" + "## Build an MNIST model" ] }, { @@ -194,8 +230,7 @@ "id": "5NMaNZQCkW9X" }, "source": [ - "For the example, we only trained the model for a single epoch, so it only trains to ~96% accuracy.\n", - "\n" + "For the example, you train the model for just a single epoch, so it only trains to ~96% accuracy." ] }, { @@ -353,7 +388,7 @@ "colab_type": "text" }, "source": [ - "Finally, convert the model like usual. Note, by default the converted model will still use float input and outputs for invocation convenience." + "Finally, convert the model like usual. By default, the converted model will still use float input and outputs for invocation convenience." ] }, { @@ -411,7 +446,7 @@ "id": "-5l6-ciItvX6" }, "source": [ - "We can run the TensorFlow Lite model using the Python TensorFlow Lite\n", + "Run the TensorFlow Lite model using the Python TensorFlow Lite\n", "Interpreter. \n", "\n", "### Load the test data\n", @@ -580,6 +615,7 @@ "\n", " input_index = interpreter.get_input_details()[0][\"index\"]\n", " output_index = interpreter.get_output_details()[0][\"index\"]\n", + "\n", " for img, label in mnist_ds:\n", " total_seen += 1\n", " interpreter.set_tensor(input_index, img)\n", @@ -605,7 +641,10 @@ "colab": {} }, "source": [ - "print(eval_model(interpreter, mnist_ds))" + "# Create smaller dataset for demonstration purposes\n", + "mnist_ds_demo = mnist_ds.take(2000)\n", + "\n", + "print(eval_model(interpreter, mnist_ds_demo))" ], "execution_count": 0, "outputs": [] @@ -617,7 +656,7 @@ "id": "Km3cY9ry8ZlG" }, "source": [ - "We can repeat the evaluation on the fully quantized model to obtain:\n" + "Repeat the evaluation on the fully quantized model to obtain:" ] }, { @@ -632,7 +671,8 @@ "# doesn't have super optimized server CPU kernels. For this reason this may be\n", "# slower than the above float interpreter. But for mobile CPUs, considerable\n", "# speedup can be observed.\n", - "print(eval_model(interpreter_quant, mnist_ds))\n" + "# Only use 2000 for demonstration purposes\n", + "print(eval_model(interpreter_quant, mnist_ds_demo))" ], "execution_count": 0, "outputs": [] @@ -644,8 +684,8 @@ "id": "L7lfxkor8pgv" }, "source": [ - "In this example, we have fully quantized a model with no difference in the accuracy." + "In this example, you have fully quantized a model with no difference in the accuracy." ] } ] -} +} \ No newline at end of file diff --git a/tensorflow/lite/g3doc/performance/post_training_quant.ipynb b/tensorflow/lite/g3doc/performance/post_training_quant.ipynb index 8bc02eedf68..89b2c2bc842 100644 --- a/tensorflow/lite/g3doc/performance/post_training_quant.ipynb +++ b/tensorflow/lite/g3doc/performance/post_training_quant.ipynb @@ -1,5 +1,55 @@ { + "nbformat": 4, + "nbformat_minor": 0, + "metadata": { + "colab": { + "name": "post_training_quant.ipynb", + "version": "0.3.2", + "provenance": [], + "private_outputs": true, + "collapsed_sections": [], + "toc_visible": true + }, + "kernelspec": { + "name": "python3", + "display_name": "Python 3" + } + }, "cells": [ + { + "cell_type": "markdown", + "metadata": { + "id": "_-GR0EDHM1SO", + "colab_type": "text" + }, + "source": [ + "##### Copyright 2019 The TensorFlow Authors." + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "R3yYtBPkM2qZ", + "colab_type": "code", + "colab": {}, + "cellView": "form" + }, + "source": [ + "#@title Licensed under the Apache License, Version 2.0 (the \"License\");\n", + "# you may not use this file except in compliance with the License.\n", + "# You may obtain a copy of the License at\n", + "#\n", + "# https://www.apache.org/licenses/LICENSE-2.0\n", + "#\n", + "# Unless required by applicable law or agreed to in writing, software\n", + "# distributed under the License is distributed on an \"AS IS\" BASIS,\n", + "# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n", + "# See the License for the specific language governing permissions and\n", + "# limitations under the License." + ], + "execution_count": 0, + "outputs": [] + }, { "cell_type": "markdown", "metadata": { @@ -7,7 +57,7 @@ "id": "6Y8E0lw5eYWm" }, "source": [ - "# Post Training Quantization" + "# Post-training weight quantization" ] }, { @@ -17,14 +67,17 @@ "id": "CIGrZZPTZVeO" }, "source": [ - "\u003ctable class=\"tfo-notebook-buttons\" align=\"left\"\u003e\n", - " \u003ctd\u003e\n", - " \u003ca target=\"_blank\" href=\"https://colab.research.google.com/github/tensorflow/tensorflow/blob/master/tensorflow/lite/tutorials/post_training_quant.ipynb\"\u003e\u003cimg src=\"https://www.tensorflow.org/images/colab_logo_32px.png\" /\u003eRun in Google Colab\u003c/a\u003e\n", - " \u003c/td\u003e\n", - " \u003ctd\u003e\n", - " \u003ca target=\"_blank\" href=\"https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/tutorials/post_training_quant.ipynb\"\u003e\u003cimg src=\"https://www.tensorflow.org/images/GitHub-Mark-32px.png\" /\u003eView source on GitHub\u003c/a\u003e\n", - " \u003c/td\u003e\n", - "\u003c/table\u003e" + "\n", + " \n", + " \n", + " \n", + "
\n", + " View on TensorFlow.org\n", + " \n", + " Run in Google Colab\n", + " \n", + " View source on GitHub\n", + "
" ] }, { @@ -38,34 +91,32 @@ "\n", "[TensorFlow Lite](https://www.tensorflow.org/lite/) now supports\n", "converting weights to 8 bit precision as part of model conversion from\n", - "tensorflow graphdefs to TFLite's flat buffer format. Weight quantization\n", + "tensorflow graphdefs to TensorFlow Lite's flat buffer format. Weight quantization\n", "achieves a 4x reduction in the model size. In addition, TFLite supports on the\n", "fly quantization and dequantization of activations to allow for:\n", "\n", "1. Using quantized kernels for faster implementation when available.\n", - "\n", "2. Mixing of floating-point kernels with quantized kernels for different parts\n", " of the graph.\n", "\n", - "Note that the activations are always stored in floating point. For ops that\n", + "The activations are always stored in floating point. For ops that\n", "support quantized kernels, the activations are quantized to 8 bits of precision\n", "dynamically prior to processing and are de-quantized to float precision after\n", "processing. Depending on the model being converted, this can give a speedup over\n", "pure floating point computation.\n", "\n", "In contrast to\n", - "[quantization aware training](https://github.com/tensorflow/tensorflow/tree/master/tensorflow/contrib/quantize)\n", + "[quantization aware training](https://github.com/tensorflow/tensorflow/tree/r1.14/tensorflow/contrib/quantize)\n", ", the weights are quantized post training and the activations are quantized dynamically \n", "at inference in this method.\n", "Therefore, the model weights are not retrained to compensate for quantization\n", "induced errors. It is important to check the accuracy of the quantized model to\n", "ensure that the degradation is acceptable.\n", "\n", - "In this tutorial, we train an MNIST model from scratch, check its accuracy in\n", - "tensorflow and then convert the saved model into a Tensorflow Lite flatbuffer\n", - "with weight quantization. We finally check the\n", - "accuracy of the converted model and compare it to the original saved model. We\n", - "run the training script mnist.py from\n", + "This tutorial trains an MNIST model from scratch, checks its accuracy in\n", + "TensorFlow, and then converts the saved model into a Tensorflow Lite flatbuffer\n", + "with weight quantization. Finally, it checks the\n", + "accuracy of the converted model and compare it to the original saved model. The training script, `mnist.py`, is from\n", "[Tensorflow official mnist tutorial](https://github.com/tensorflow/models/tree/master/official/mnist).\n" ] }, @@ -76,7 +127,7 @@ "id": "2XsEP17Zelz9" }, "source": [ - "## Building an MNIST model" + "## Build an MNIST model" ] }, { @@ -91,59 +142,57 @@ }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "gyqAw1M9lyab" + "id": "gyqAw1M9lyab", + "colab": {} }, - "outputs": [], "source": [ "! pip uninstall -y tensorflow\n", "! pip install -U tf-nightly" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "WsN6s5L1ieNl" + "id": "WsN6s5L1ieNl", + "colab": {} }, - "outputs": [], "source": [ "import tensorflow as tf\n", "tf.enable_eager_execution()" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "00U0taBoe-w7" + "id": "00U0taBoe-w7", + "colab": {} }, - "outputs": [], "source": [ "! git clone --depth 1 https://github.com/tensorflow/models" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "4XZPtSh-fUOc" + "id": "4XZPtSh-fUOc", + "colab": {} }, - "outputs": [], "source": [ "import sys\n", "import os\n", "\n", - "if sys.version_info.major \u003e= 3:\n", + "if sys.version_info.major >= 3:\n", " import pathlib\n", "else:\n", " import pathlib2 as pathlib\n", @@ -151,7 +200,9 @@ "# Add `models` to the python path.\n", "models_path = os.path.join(os.getcwd(), \"models\")\n", "sys.path.append(models_path)" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "markdown", @@ -165,31 +216,31 @@ }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "eMsw_6HujaqM" + "id": "eMsw_6HujaqM", + "colab": {} }, - "outputs": [], "source": [ "saved_models_root = \"/tmp/mnist_saved_model\"" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "hWSAjQWagIHl" + "id": "hWSAjQWagIHl", + "colab": {} }, - "outputs": [], "source": [ "# The above path addition is not visible to subprocesses, add the path for the subprocess as well.\n", "# Note: channels_last is required here or the conversion may fail. \n", "!PYTHONPATH={models_path} python models/official/mnist/mnist.py --train_epochs=1 --export_dir {saved_models_root} --data_format=channels_last" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "markdown", @@ -198,7 +249,7 @@ "id": "5NMaNZQCkW9X" }, "source": [ - "For the example, we only trained the model for a single epoch, so it only trains to ~96% accuracy.\n", + "For the example, since you trained the model for just a single epoch, so it only trains to ~96% accuracy.\n", "\n" ] }, @@ -216,17 +267,17 @@ }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "Xp5oClaZkbtn" + "id": "Xp5oClaZkbtn", + "colab": {} }, - "outputs": [], "source": [ "saved_model_dir = str(sorted(pathlib.Path(saved_models_root).glob(\"*\"))[-1])\n", "saved_model_dir" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "markdown", @@ -242,19 +293,19 @@ }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "_i8B2nDZmAgQ" + "id": "_i8B2nDZmAgQ", + "colab": {} }, - "outputs": [], "source": [ "import tensorflow as tf\n", "tf.enable_eager_execution()\n", "converter = tf.lite.TFLiteConverter.from_saved_model(saved_model_dir)\n", "tflite_model = converter.convert()" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "markdown", @@ -268,31 +319,31 @@ }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "vptWZq2xnclo" + "id": "vptWZq2xnclo", + "colab": {} }, - "outputs": [], "source": [ "tflite_models_dir = pathlib.Path(\"/tmp/mnist_tflite_models/\")\n", "tflite_models_dir.mkdir(exist_ok=True, parents=True)" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "Ie9pQaQrn5ue" + "id": "Ie9pQaQrn5ue", + "colab": {} }, - "outputs": [], "source": [ "tflite_model_file = tflite_models_dir/\"mnist_model.tflite\"\n", "tflite_model_file.write_bytes(tflite_model)" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "markdown", @@ -306,14 +357,12 @@ }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "g8PUvLWDlmmz" + "id": "g8PUvLWDlmmz", + "colab": {} }, - "outputs": [], - "source": [ + "source": [ "# Note: If you don't have a recent tf-nightly installed, the\n", "# \"optimizations\" line will have no effect.\n", "tf.logging.set_verbosity(tf.logging.INFO)\n", @@ -321,7 +370,9 @@ "tflite_quant_model = converter.convert()\n", "tflite_model_quant_file = tflite_models_dir/\"mnist_model_quant.tflite\"\n", "tflite_model_quant_file.write_bytes(tflite_quant_model)" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "markdown", @@ -329,22 +380,22 @@ "colab_type": "text", "id": "PhMmUTl4sbkz" }, - "source": [ + "source": [ "Note how the resulting file, is approximately `1/4` the size." ] }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "JExfcfLDscu4" + "id": "JExfcfLDscu4", + "colab": {} }, - "outputs": [], "source": [ "!ls -lh {tflite_models_dir}" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "markdown", @@ -363,7 +414,7 @@ "id": "-5l6-ciItvX6" }, "source": [ - "We can run the TensorFlow Lite model using the python TensorFlow Lite\n", + "Run the TensorFlow Lite model using the Python TensorFlow Lite\n", "Interpreter. \n", "\n", "### load the test data\n", @@ -373,13 +424,11 @@ }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "eTIuU07NuKFL" + "id": "eTIuU07NuKFL", + "colab": {} }, - "outputs": [], "source": [ "import numpy as np\n", "mnist_train, mnist_test = tf.keras.datasets.mnist.load_data()\n", @@ -389,7 +438,9 @@ "# `tf.lite.Interpreter.resize_tensor_input` to also change it for\n", "# the interpreter.\n", "mnist_ds = tf.data.Dataset.from_tensor_slices((images, labels)).batch(1)" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "markdown", @@ -403,48 +454,48 @@ }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "Jn16Rc23zTss" + "id": "Jn16Rc23zTss", + "colab": {} }, - "outputs": [], "source": [ "interpreter = tf.lite.Interpreter(model_path=str(tflite_model_file))\n", "interpreter.allocate_tensors()\n", "input_index = interpreter.get_input_details()[0][\"index\"]\n", "output_index = interpreter.get_output_details()[0][\"index\"]" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "J8Pztk1mvNVL" + "id": "J8Pztk1mvNVL", + "colab": {} }, - "outputs": [], "source": [ "tf.logging.set_verbosity(tf.logging.DEBUG)\n", "interpreter_quant = tf.lite.Interpreter(model_path=str(tflite_model_quant_file))" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "Afl6yGvWyqAr" + "id": "Afl6yGvWyqAr", + "colab": {} }, - "outputs": [], "source": [ "interpreter_quant.allocate_tensors()\n", "input_index = interpreter_quant.get_input_details()[0][\"index\"]\n", - "output_index = interpreter_quant.get_output_details()[0][\"index\"]\n" - ] + "output_index = interpreter_quant.get_output_details()[0][\"index\"]" + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "markdown", @@ -458,13 +509,11 @@ }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "AKslvo2kwWac" + "id": "AKslvo2kwWac", + "colab": {} }, - "outputs": [], "source": [ "for img, label in mnist_ds.take(1):\n", " break\n", @@ -472,17 +521,17 @@ "interpreter.set_tensor(input_index, img)\n", "interpreter.invoke()\n", "predictions = interpreter.get_tensor(output_index)" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "XZClM2vo3_bm" + "id": "XZClM2vo3_bm", + "colab": {} }, - "outputs": [], "source": [ "import matplotlib.pylab as plt\n", "\n", @@ -491,7 +540,9 @@ "_ = plt.title(template.format(true= str(label[0].numpy()),\n", " predict=str(predictions[0])))\n", "plt.grid(False)" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "markdown", @@ -505,13 +556,11 @@ }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "05aeAuWjvjPx" + "id": "05aeAuWjvjPx", + "colab": {} }, - "outputs": [], "source": [ "def eval_model(interpreter, mnist_ds):\n", " total_seen = 0\n", @@ -530,20 +579,22 @@ " (total_seen, float(num_correct) / float(total_seen)))\n", "\n", " return float(num_correct) / float(total_seen)" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "DqXBnDfJ7qxL" + "id": "DqXBnDfJ7qxL", + "colab": {} }, - "outputs": [], "source": [ "print(eval_model(interpreter, mnist_ds))" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "markdown", @@ -552,21 +603,21 @@ "id": "Km3cY9ry8ZlG" }, "source": [ - "We can repeat the evaluation on the weight quantized model to obtain:\n" + "Repeat the evaluation on the weight quantized model to obtain:\n" ] }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "-9cnwiPp6EGm" + "id": "-9cnwiPp6EGm", + "colab": {} }, - "outputs": [], "source": [ - "print(eval_model(interpreter_quant, mnist_ds))\n" - ] + "print(eval_model(interpreter_quant, mnist_ds))" + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "markdown", @@ -576,7 +627,7 @@ }, "source": [ "\n", - "In this example, we have compressed model with no difference in the accuracy." + "In this example, the compressed model has no difference in the accuracy." ] }, { @@ -586,31 +637,29 @@ "id": "M0o1FtmWeKZm" }, "source": [ - "\n", - "\n", "## Optimizing an existing model\n", "\n", - "We now consider another example. Resnets with pre-activation layers (Resnet-v2) are widely used for vision applications.\n", + "Resnets with pre-activation layers (Resnet-v2) are widely used for vision applications.\n", " Pre-trained frozen graph for resnet-v2-101 is available at the\n", " [Tensorflow Lite model repository](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/g3doc/models.md).\n", "\n", - "We can convert the frozen graph to a TFLite flatbuffer with quantization by:\n" + "You can convert the frozen graph to a TensorFLow Lite flatbuffer with quantization by:\n" ] }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "v5p5VcNPjILQ" + "id": "v5p5VcNPjILQ", + "colab": {} }, - "outputs": [], "source": [ "archive_path = tf.keras.utils.get_file(\"resnet_v2_101.tgz\", \"https://storage.googleapis.com/download.tensorflow.org/models/tflite_11_05_08/resnet_v2_101.tgz\", extract=True)\n", "archive_path = pathlib.Path(archive_path)\n", "archive_dir = str(archive_path.parent)" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "markdown", @@ -624,26 +673,24 @@ }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "g_Q_OMEJ4LIc" + "id": "g_Q_OMEJ4LIc", + "colab": {} }, - "outputs": [], "source": [ "! cat {archive_dir}/resnet_v2_101_299_info.txt" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "ujCAFhqm-C6H" + "id": "ujCAFhqm-C6H", + "colab": {} }, - "outputs": [], "source": [ "graph_def_file = pathlib.Path(archive_path).parent/\"resnet_v2_101_299_frozen.pb\"\n", "input_arrays = [\"input\"] \n", @@ -652,22 +699,23 @@ " str(graph_def_file), input_arrays, output_arrays, input_shapes={\"input\":[1,299,299,3]})\n", "converter.optimizations = [tf.lite.Optimize.OPTIMIZE_FOR_SIZE]\n", "resnet_tflite_file = graph_def_file.parent/\"resnet_v2_101_quantized.tflite\"\n", - "resnet_tflite_file.write_bytes(converter.convert())\n" - ] + "resnet_tflite_file.write_bytes(converter.convert())" + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "code", - "execution_count": 0, "metadata": { - "colab": {}, "colab_type": "code", - "id": "vhOjeg1x9Knp" + "id": "vhOjeg1x9Knp", + "colab": {} }, - "outputs": [], "source": [ - "\n", "!ls -lh {archive_dir}/*.tflite" - ] + ], + "execution_count": 0, + "outputs": [] }, { "cell_type": "markdown", @@ -683,21 +731,5 @@ "The optimized model top-1 accuracy is 76.8, the same as the floating point model." ] } - ], - "metadata": { - "colab": { - "collapsed_sections": [], - "name": "post-training-quant.ipynb", - "private_outputs": true, - "provenance": [], - "toc_visible": true, - "version": "0.3.2" - }, - "kernelspec": { - "display_name": "Python 2", - "name": "python2" - } - }, - "nbformat": 4, - "nbformat_minor": 0 -} + ] +} \ No newline at end of file