sagemaker-python-sdk/mxnet_gluon_mnist/mxnet_mnist_with_gluon_local_mode.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Local MNIST Training with MXNet and Gluon\n",
    "\n",
    "### Pre-requisites\n",
    "\n",
    "This notebook shows how to use the SageMaker Python SDK to run your code in a local container before deploying to SageMaker's managed training or hosting environments.  This can speed up iterative testing and debugging while using the same familiar Python SDK interface.  Just change your estimator's `instance_type` to `local`.  You could also use `local_gpu` if you're using an ml.p2 or ml.p3 notebook instance, but then you'll need to set `instance_count=1` since distributed, local, GPU training is not yet supported.\n",
    "\n",
    "In order to use this feature you'll need to install docker-compose (and nvidia-docker if training with a GPU).  Running the setup.sh script below will handle this for you.\n",
    "\n",
    "**Note, you can only run a single local notebook at one time.**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!/bin/bash ./setup.sh"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Overview\n",
    "\n",
    "MNIST is a widely used dataset for handwritten digit classification. It consists of 70,000 labeled 28x28 pixel grayscale images of hand-written digits. The dataset is split into 60,000 training images and 10,000 test images. There are 10 classes (one for each of the 10 digits). This tutorial will show how to train and test an MNIST model on SageMaker local mode using MXNet and the Gluon API."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import subprocess\n",
    "import boto3\n",
    "import sagemaker\n",
    "from sagemaker.mxnet import MXNet\n",
    "from mxnet import gluon\n",
    "from sagemaker import get_execution_role\n",
    "\n",
    "sagemaker_session = sagemaker.Session()\n",
    "\n",
    "instance_type = \"local\"\n",
    "\n",
    "if subprocess.call(\"nvidia-smi\") == 0:\n",
    "    ## Set type to GPU if one is present\n",
    "    instance_type = \"local_gpu\"\n",
    "\n",
    "print(\"Instance type = \" + instance_type)\n",
    "\n",
    "role = get_execution_role()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Download training and test data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "\n",
    "for inner_dir in [\"train\", \"test\"]:\n",
    "    data_dir = \"./data/{}/\".format(inner_dir)\n",
    "    if not os.path.exists(data_dir):\n",
    "        os.makedirs(data_dir)\n",
    "\n",
    "s3 = boto3.client(\"s3\")\n",
    "s3.download_file(\n",
    "    \"sagemaker-sample-files\",\n",
    "    \"datasets/image/MNIST/train/train-images-idx3-ubyte.gz\",\n",
    "    \"./data/train/train-images-idx3-ubyte.gz\",\n",
    ")\n",
    "s3.download_file(\n",
    "    \"sagemaker-sample-files\",\n",
    "    \"datasets/image/MNIST/train/train-labels-idx1-ubyte.gz\",\n",
    "    \"./data/train/train-labels-idx1-ubyte.gz\",\n",
    ")\n",
    "s3.download_file(\n",
    "    \"sagemaker-sample-files\",\n",
    "    \"datasets/image/MNIST/test/t10k-images-idx3-ubyte.gz\",\n",
    "    \"./data/test/t10k-images-idx3-ubyte.gz\",\n",
    ")\n",
    "s3.download_file(\n",
    "    \"sagemaker-sample-files\",\n",
    "    \"datasets/image/MNIST/test/t10k-labels-idx1-ubyte.gz\",\n",
    "    \"./data/test/t10k-labels-idx1-ubyte.gz\",\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Uploading the data\n",
    "\n",
    "We use the `sagemaker.Session.upload_data` function to upload our datasets to an S3 location. The return value `inputs` identifies the location -- we will use this later when we start the training job."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "inputs = sagemaker_session.upload_data(path=\"data\", key_prefix=\"data/mnist\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Implement the training function\n",
    "\n",
    "We need to provide a training script that can run on the SageMaker platform. The training scripts are essentially the same as one you would write for local training, except that you need to provide a `train` function. The `train` function will check for the validation accuracy at the end of every epoch and checkpoints the best model so far, along with the optimizer state, in the folder `/opt/ml/checkpoints` if the folder path exists, else it will skip the checkpointing. When SageMaker calls your function, it will pass in arguments that describe the training environment. Check the script below to see how this works.\n",
    "\n",
    "The script here is an adaptation of the [Gluon MNIST example](https://github.com/apache/incubator-mxnet/blob/master/example/gluon/mnist.py) provided by the [Apache MXNet](https://mxnet.incubator.apache.org/) project. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "!cat 'mnist.py'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Run the training script on SageMaker\n",
    "\n",
    "The ```MXNet``` class allows us to run our training function on SageMaker local mode. We need to configure it with our training script, an IAM role, the number of training instances, and the training instance type.  This is the only difference from [mxnet_mnist_with_gluon.ipynb](./mxnet_mnist_with_gluon.ipynb).  Instead of ``instance_type='ml.c4.xlarge'``, we set it to ``instance_type='local'``.  For local training with GPU, we could set this to \"local_gpu\".  In this case, `instance_type` was set above based on your whether you're running a GPU instance."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "m = MXNet(\n",
    "    \"mnist.py\",\n",
    "    role=role,\n",
    "    instance_count=1,\n",
    "    instance_type=instance_type,\n",
    "    framework_version=\"1.6.0\",\n",
    "    py_version=\"py3\",\n",
    "    hyperparameters={\n",
    "        \"batch-size\": 100,\n",
    "        \"epochs\": 20,\n",
    "        \"learning-rate\": 0.1,\n",
    "        \"momentum\": 0.9,\n",
    "        \"log-interval\": 100,\n",
    "    },\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "After we've constructed our `MXNet` object, we fit it using the data we uploaded to S3. Even though we're in local mode, using S3 as our data source makes sense because it maintains consistency with how SageMaker's distributed, managed training ingests data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "m.fit(inputs)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "After training, we use the MXNet object to deploy an MXNetPredictor object. This creates a SageMaker endpoint locally that we can use to perform inference. \n",
    "\n",
    "This allows us to perform inference on json encoded multi-dimensional arrays. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "predictor = m.deploy(initial_instance_count=1, instance_type=instance_type)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can now use this predictor to classify hand-written digits. Manually drawing into the image box loads the pixel data into a 'data' variable in this notebook, which we can then pass to the MXNet predictor."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from IPython.display import HTML\n",
    "\n",
    "HTML(open(\"input.html\").read())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Fetch the first image from the test dataset and display it."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import gzip\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "f = gzip.open(\"data/train/train-images-idx3-ubyte.gz\", \"r\")\n",
    "\n",
    "image_size = 28\n",
    "\n",
    "f.read(16)\n",
    "buf = f.read(image_size * image_size)\n",
    "data = np.frombuffer(buf, dtype=np.uint8).astype(np.float32)\n",
    "data = data.reshape(1, image_size, image_size, 1)\n",
    "\n",
    "image = np.asarray(data).squeeze()\n",
    "plt.imshow(image)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The predictor runs inference on our input data and returns the predicted digit (as a float value, so we convert to int for display)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "response = predictor.predict(data)\n",
    "print(int(response))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Clean-up\n",
    "\n",
    "Deleting the local endpoint when you're finished is important since you can only run one local endpoint at a time."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "predictor.delete_endpoint()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "conda_mxnet_p36",
   "language": "python",
   "name": "conda_mxnet_p36"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.13"
  },
  "notice": "Copyright 2017 Amazon.com, Inc. or its affiliates. All Rights Reserved.  Licensed under the Apache License, Version 2.0 (the \"License\"). You may not use this file except in compliance with the License. A copy of the License is located at http://aws.amazon.com/apache2.0/ or in the \"license\" file accompanying this file. This file 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."
 },
 "nbformat": 4,
 "nbformat_minor": 2
}