update gluon examples (aws#11)

Author: jesterhazy

sagemaker-python-sdk/mxnet_gluon_cifar10/cifar10.ipynb

@@ -97,21 +97,19 @@
    "source": [
     "## Run the training script on SageMaker\n",
     "\n",
-    "The ```MXNet``` class allows us to run our training function as a distributed training job on SageMaker infrastructure. We need to configure it with our training script, an IAM role, the number of training instances, and the training instance type. In this case we will run our training job on four p2.xlarge instances. "
+    "The ```MXNet``` class allows us to run our training function as a distributed training job on SageMaker infrastructure. We need to configure it with our training script, an IAM role, the number of training instances, and the training instance type. In this case we will run our training job on four `ml.p2.xlarge` instances. "
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
-   "metadata": {
-    "collapsed": true
-   },
+   "metadata": {},
    "outputs": [],
    "source": [
     "m = MXNet(\"cifar10.py\", \n",
     "          role=role, \n",
     "          train_instance_count=4, \n",
-    "          train_instance_type=\"p2.xlarge\",\n",
+    "          train_instance_type=\"ml.p2.xlarge\",\n",
     "          hyperparameters={'batch_size': 128, \n",
     "                           'epochs': 50, \n",
     "                           'learning_rate': 0.1, \n",
@@ -142,7 +140,9 @@
    "source": [
     "## Prediction\n",
     "\n",
-    "After training, we use the MXNet object to create and deploy a hosted prediction endpoint. We can use the object returned by `deploy` to call the endpoint and perform inference on our sample image."
+    "After training, we use the MXNet estimator object to create and deploy a hosted prediction endpoint. We can use a CPU-based instance for inference (in this case an `ml.c4.xlarge`), even though we trained on GPU instances.\n",
+    "\n",
+    "The predictor object returned by `deploy` lets us call the new endpoint and perform inference on our sample images. "
    ]
   },
   {
@@ -151,7 +151,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "predictor = m.deploy(min_instances=1, max_instances=1, instance_type='c4.xlarge')"
+    "predictor = m.deploy(initial_instance_count=1, instance_type='ml.c4.xlarge')"
    ]
   },
   {
@@ -178,9 +178,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "metadata": {
-    "collapsed": true
-   },
+   "metadata": {},
    "outputs": [],
    "source": [
     "# load the CIFAR10 samples, and convert them into format we can use with the prediction endpoint\n",
@@ -232,9 +230,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "metadata": {
-    "collapsed": true
-   },
+   "metadata": {},
    "outputs": [],
    "source": [
     "m.delete_endpoint()"

sagemaker-python-sdk/mxnet_gluon_cifar10/cifar10.py

@@ -33,7 +33,6 @@ def train(current_host, hosts, num_cpus, num_gpus, channel_input_dirs, model_dir
     ctx = [mx.gpu(i) for i in range(num_gpus)] if num_gpus > 0 else [mx.cpu()]
     net = models.get_model('resnet34_v2', ctx=ctx, pretrained=False, classes=10)
     batch_size *= max(1, len(ctx))
-    learning_rate *= max(1, len(ctx))
 
     # load training and validation data
     # we use the gluon.data.vision.CIFAR10 class because of its built in pre-processing logic,

sagemaker-python-sdk/mxnet_gluon_cifar10/credentials.py

@@ -13,23 +13,18 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "metadata": {
-    "collapsed": true
-   },
+   "metadata": {},
    "outputs": [],
    "source": [
-    "import credentials # put your credentials in credentials.py\n",
     "import os\n",
     "import boto3\n",
     "import sagemaker\n",
     "from sagemaker.mxnet import MXNet\n",
     "from mxnet import gluon\n",
     "\n",
-    "\n",
     "sagemaker_session = sagemaker.Session()\n",
     "\n",
-    "# Replace with a role that gives SageMaker access to s3 and cloudwatch\n",
-    "# see 1-Creating_a_role_allowing_SageMaker_to_access_S3_Cloudwatch_ECR.ipynb\n",
+    "# Replace with a role that gives SageMaker access to S3 and CloudWatch\n",
     "role='SageMakerRole'"
    ]
   },
@@ -43,9 +38,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "metadata": {
-    "collapsed": true
-   },
+   "metadata": {},
    "outputs": [],
    "source": [
     "gluon.data.vision.MNIST('./data/train', train=True)\n",
@@ -64,9 +57,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "metadata": {
-    "collapsed": true
-   },
+   "metadata": {},
    "outputs": [],
    "source": [
     "inputs = sagemaker_session.upload_data(path='data', key_prefix='data/mnist')"
@@ -86,9 +77,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "metadata": {
-    "collapsed": true
-   },
+   "metadata": {},
    "outputs": [],
    "source": [
     "!cat 'mnist.py'"
@@ -106,15 +95,13 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "metadata": {
-    "collapsed": true
-   },
+   "metadata": {},
    "outputs": [],
    "source": [
     "m = MXNet(\"mnist.py\", \n",
     "          role=role, \n",
     "          train_instance_count=1, \n",
-    "          train_instance_type=\"c4.xlarge\",\n",
+    "          train_instance_type=\"ml.c4.xlarge\",\n",
     "          enable_cloudwatch_metrics=True,\n",
     "          hyperparameters={'batch_size': 100, \n",
     "                         'epochs': 20, \n",
@@ -134,7 +121,6 @@
    "cell_type": "code",
    "execution_count": null,
    "metadata": {
-    "collapsed": true,
     "scrolled": true
    },
    "outputs": [],
@@ -155,12 +141,11 @@
    "cell_type": "code",
    "execution_count": null,
    "metadata": {
-    "collapsed": true,
     "scrolled": true
    },
    "outputs": [],
    "source": [
-    "predictor = m.deploy(instance_type = 'c4.xlarge', min_instances = 1, max_instances = 1)"
+    "predictor = m.deploy(initial_instance_count=1, instance_type='ml.c4.xlarge')"
    ]
   },
   {
@@ -173,9 +158,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "metadata": {
-    "collapsed": true
-   },
+   "metadata": {},
    "outputs": [],
    "source": [
     "from IPython.display import HTML\n",
@@ -193,7 +176,6 @@
    "cell_type": "code",
    "execution_count": null,
    "metadata": {
-    "collapsed": true,
     "scrolled": true
    },
    "outputs": [],
@@ -214,9 +196,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "metadata": {
-    "collapsed": true
-   },
+   "metadata": {},
    "outputs": [],
    "source": [
     "m.delete_endpoint()"