[SYSTEMML-618] SystemML-NN: Updating the MNIST softmax classifier example to remove unnecessary dependencies, and to simplify the invocation.
Project: http://git-wip-us.apache.org/repos/asf/incubator-systemml/repo Commit: http://git-wip-us.apache.org/repos/asf/incubator-systemml/commit/d47fe8fc Tree: http://git-wip-us.apache.org/repos/asf/incubator-systemml/tree/d47fe8fc Diff: http://git-wip-us.apache.org/repos/asf/incubator-systemml/diff/d47fe8fc Branch: refs/heads/master Commit: d47fe8fc5ca39ce6c50944a2fbaaca37ba3ba134 Parents: 60f3fe6 Author: Mike Dusenberry <[email protected]> Authored: Fri Jun 24 17:28:57 2016 -0700 Committer: Mike Dusenberry <[email protected]> Committed: Fri Jun 24 17:28:57 2016 -0700 ---------------------------------------------------------------------- .../Example - MNIST Softmax Classifier.ipynb | 84 +++++++++----------- .../SystemML-NN/examples/get_mnist_data.sh | 28 +++++++ .../SystemML-NN/examples/mnist_softmax.dml | 58 +++++++++----- 3 files changed, 107 insertions(+), 63 deletions(-) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/incubator-systemml/blob/d47fe8fc/scripts/staging/SystemML-NN/Example - MNIST Softmax Classifier.ipynb ---------------------------------------------------------------------- diff --git a/scripts/staging/SystemML-NN/Example - MNIST Softmax Classifier.ipynb b/scripts/staging/SystemML-NN/Example - MNIST Softmax Classifier.ipynb index 5c1de15..d7da0d2 100644 --- a/scripts/staging/SystemML-NN/Example - MNIST Softmax Classifier.ipynb +++ b/scripts/staging/SystemML-NN/Example - MNIST Softmax Classifier.ipynb @@ -15,8 +15,6 @@ }, "outputs": [], "source": [ - "import numpy as np\n", - "\n", "# Add SystemML PySpark API file.\n", "sc.addPyFile(\"https://raw.githubusercontent.com/apache/incubator-systemml/3d5f9b11741f6d6ecc6af7cbaa1069cde32be838/src/main/java/org/apache/sysml/api/python/SystemML.py\";)\n", "\n", @@ -36,43 +34,22 @@ "cell_type": "markdown", "metadata": {}, "source": [ - "The MNIST dataset contains labeled images of handwritten digits, where each example is a 28x28 pixel image of grayscale values scaled to [0,1] and stretched out as 784 pixels, and each label is a one-hot encoding over 10 possible digits. Here, we use TensorFlow's API for accessing the data, and retrieve 50,000 training examples, 5,000 validation examples, and 10,000 test examples. [Note: TensorFlow can easily be installed via [these instructions](https://www.tensorflow.org/versions/r0.9/get_started/os_setup.html#pip-installation).]" + "The MNIST dataset contains labeled images of handwritten digits, where each example is a 28x28 pixel image of grayscale values in the range [0,255] stretched out as 784 pixels, and each label is one of 10 possible digits in [0,9]. Here, we download 60,000 training examples, and 10,000 test examples, where the format is \"label, pixel_1, pixel_2, ..., pixel_n\"." ] }, { "cell_type": "code", "execution_count": null, "metadata": { - "collapsed": true + "collapsed": false }, "outputs": [], "source": [ "%%sh\n", - "mkdir -p examples/data/mnist/" - ] - }, - { - "cell_type": "code", - "execution_count": null, - "metadata": { - "collapsed": false, - "scrolled": false - }, - "outputs": [], - "source": [ - "import tensorflow as tf\n", - "from tensorflow.examples.tutorials.mnist import input_data\n", - "\n", - "# Get MNIST data\n", - "mnist = input_data.read_data_sets('MNIST_data', one_hot=True)\n", - "\n", - "# Save to CSV\n", - "np.savetxt(\"examples/data/mnist/train_images.csv\", mnist.train.images, delimiter=\",\")\n", - "np.savetxt(\"examples/data/mnist/train_labels.csv\", mnist.train.labels, delimiter=\",\")\n", - "np.savetxt(\"examples/data/mnist/val_images.csv\", mnist.validation.images, delimiter=\",\")\n", - "np.savetxt(\"examples/data/mnist/val_labels.csv\", mnist.validation.labels, delimiter=\",\")\n", - "np.savetxt(\"examples/data/mnist/test_images.csv\", mnist.test.images, delimiter=\",\")\n", - "np.savetxt(\"examples/data/mnist/test_labels.csv\", mnist.test.labels, delimiter=\",\")" + "mkdir -p examples/data/mnist/\n", + "cd examples/data/mnist/\n", + "curl -O http://pjreddie.com/media/files/mnist_train.csv\n";, + "curl -O http://pjreddie.com/media/files/mnist_test.csv"; ] }, { @@ -100,17 +77,28 @@ "script = \"\"\"\n", "source(\"examples/mnist_softmax.dml\") as mnist_softmax\n", "\n", - "# Read data\n", - "X = read($X, format=\"csv\")\n", - "y = read($y, format=\"csv\")\n", + "# Read training data\n", + "data = read($data, format=\"csv\")\n", + "n = nrow(data)\n", + "\n", + "# Extract images and labels\n", + "images = data[,2:ncol(data)]\n", + "labels = data[,1]\n", "\n", - "X_val = read($X_val, format=\"csv\")\n", - "y_val = read($y_val, format=\"csv\")\n", + "# Scale images to [0,1], and one-hot encode the labels\n", + "images = images / 255.0\n", + "labels = table(seq(1, n), labels+1, n, 10)\n", + "\n", + "# Split into training (55,000 examples) and validation (5,000 examples)\n", + "X = images[5001:nrow(images),]\n", + "X_val = images[1:5000,]\n", + "y = labels[5001:nrow(images),]\n", + "y_val = labels[1:5000,]\n", "\n", "# Train\n", "[W, b] = mnist_softmax::train(X, y, X_val, y_val)\n", "\n", - "# Write model out\n", + "# Write model out (we will extract these back into PySpark)\n", "write(W, $Wout)\n", "write(b, $bout)\n", "\n", @@ -118,10 +106,7 @@ "print(\"\")\n", "\"\"\"\n", "ml.reset()\n", - "out = ml.executeScript(script, {\"X\": \"examples/data/mnist/train_images.csv\", \n", - " \"y\": \"examples/data/mnist/train_labels.csv\",\n", - " \"X_val\": \"examples/data/mnist/val_images.csv\", \n", - " \"y_val\": \"examples/data/mnist/val_labels.csv\"},\n", + "out = ml.executeScript(script, {\"data\": \"examples/data/mnist/mnist_train.csv\"},\n", " outputs=[\"W\", \"b\"])" ] }, @@ -162,23 +147,32 @@ "script = \"\"\"\n", "source(\"examples/mnist_softmax.dml\") as mnist_softmax\n", "\n", - "# Read data & coefficients\n", - "X_test = read($X_test, format=\"csv\")\n", - "y_test = read($y_test, format=\"csv\")\n", + "# Read test data\n", + "data = read($data, format=\"csv\")\n", + "n = nrow(data)\n", + "\n", + "# Extract images and labels\n", + "X_test = data[,2:ncol(data)]\n", + "y_test = data[,1]\n", + "\n", + "# Scale images to [0,1], and one-hot encode the labels\n", + "X_test = X_test / 255.0\n", + "y_test = table(seq(1, n), y_test+1, n, 10)\n", + "\n", + "# Read model coefficients\n", "W = read($W)\n", "b = read($b)\n", "\n", "# Eval on test set\n", "[loss, accuracy] = mnist_softmax::eval(X_test, y_test, W, b)\n", "\n", - "print(\"Accuracy: \" + accuracy)\n", + "print(\"Test Accuracy: \" + accuracy)\n", "\n", "print(\"\")\n", "print(\"\")\n", "\"\"\"\n", "ml.reset()\n", - "out = ml.executeScript(script, {\"X_test\": \"examples/data/mnist/test_images.csv\",\n", - " \"y_test\": \"examples/data/mnist/test_labels.csv\",\n", + "out = ml.executeScript(script, {\"data\": \"examples/data/mnist/mnist_test.csv\",\n", " \"W\": W, \"b\": b})" ] } http://git-wip-us.apache.org/repos/asf/incubator-systemml/blob/d47fe8fc/scripts/staging/SystemML-NN/examples/get_mnist_data.sh ---------------------------------------------------------------------- diff --git a/scripts/staging/SystemML-NN/examples/get_mnist_data.sh b/scripts/staging/SystemML-NN/examples/get_mnist_data.sh new file mode 100755 index 0000000..6fed70b --- /dev/null +++ b/scripts/staging/SystemML-NN/examples/get_mnist_data.sh @@ -0,0 +1,28 @@ +#!/usr/bin/env bash +#------------------------------------------------------------- +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +# +#------------------------------------------------------------- + +DIR="$(cd "$(dirname "$0")" && pwd)" +mkdir -p $DIR/data/mnist/ +cd $DIR/data/mnist/ +curl -O http://pjreddie.com/media/files/mnist_train.csv +curl -O http://pjreddie.com/media/files/mnist_test.csv + http://git-wip-us.apache.org/repos/asf/incubator-systemml/blob/d47fe8fc/scripts/staging/SystemML-NN/examples/mnist_softmax.dml ---------------------------------------------------------------------- diff --git a/scripts/staging/SystemML-NN/examples/mnist_softmax.dml b/scripts/staging/SystemML-NN/examples/mnist_softmax.dml index f3c47e9..8ba24cb 100644 --- a/scripts/staging/SystemML-NN/examples/mnist_softmax.dml +++ b/scripts/staging/SystemML-NN/examples/mnist_softmax.dml @@ -159,30 +159,52 @@ generate_dummy_data = function() # # This runs if called as a script. # -# Ex: +# The MNIST dataset contains labeled images of handwritten digits, +# where each example is a 28x28 pixel image of grayscale values in +# the range [0,255] stretched out as 784 pixels, and each label is +# one of 10 possible digits in [0,9]. +# +# Here, we assume 60,000 training examples, and 10,000 test examples, +# where the format is "label, pixel_1, pixel_2, ..., pixel_n". +# +# 1. Download data +# ``` +# examples/get_mnist_data.sh +# ``` +# +# 2. Execute using Spark # ``` # $SPARK_HOME/bin/spark-submit --master local[*] --driver-memory 5G # --conf spark.driver.maxResultSize=0 --conf spark.akka.frameSize=128 # $SYSTEMML_HOME/target/SystemML.jar -f examples/mnist_softmax.dml -# -nvargs X=examples/data/mnist/train_images.csv y=examples/data/mnist/train_labels.csv -# X_val=examples/data/mnist/val_images.csv y_val=examples/data/mnist/val_labels.csv -# X_test=examples/data/mnist/test_images.csv y_test=examples/data/mnist/test_labels.csv -# out_dir=examples/model/mnist_softmax +# -nvargs train=examples/data/mnist/mnist_train.csv +# test=examples/data/mnist/mnist_test.csv out_dir=examples/model/mnist_softmax # ``` # -# The MNIST dataset contains labeled images of handwritten digits, -# where each example is a 28x28 pixel image of grayscale values -# scaled to [0,1] and stretched out as 784 pixels, and each label -# is a one-hot encoding over 10 possible digits. -# -# Read data -X = read($X, format="csv") -y = read($y, format="csv") -X_val = read($X_val, format="csv") -y_val = read($y_val, format="csv") -X_test = read($X_test, format="csv") -y_test = read($y_test, format="csv") +# Read training data +train = read($train, format="csv") +test = read($test, format="csv") + +# Extract images and labels +images = train[,2:ncol(train)] +labels = train[,1] +X_test = test[,2:ncol(test)] +y_test = test[,1] + +# Scale images to [0,1], and one-hot encode the labels +n = nrow(train) +n_test = nrow(test) +images = images / 255.0 +labels = table(seq(1, n), labels+1, n, 10) +X_test = X_test / 255.0 +y_test = table(seq(1, n_test), y_test+1, n_test, 10) + +# Split into training (55,000 examples) and validation (5,000 examples) +X = images[5001:nrow(images),] +X_val = images[1:5000,] +y = labels[5001:nrow(images),] +y_val = labels[1:5000,] # Train [W, b] = train(X, y, X_val, y_val) @@ -195,7 +217,7 @@ write(b, $out_dir+"/b") [loss, accuracy] = eval(X_test, y_test, W, b) # Output results -print("Accuracy: " + accuracy) +print("Test Accuracy: " + accuracy) write(accuracy, $out_dir+"/accuracy") print("")
