Add markdown explanation for TorchVision Neo NB (#3295)

* Add markdown explanation for TorchVision Neo NB

* Use present tense

Author: jkroll-aws

sagemaker_neo_compilation_jobs/pytorch_torchvision/pytorch_torchvision_neo.ipynb

@@ -11,7 +11,19 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "Amazon SageMaker Neo is an API to compile machine learning models to optimize them for our choice of hardware targets. Currently, Neo supports pre-trained PyTorch models from [TorchVision](https://pytorch.org/docs/stable/torchvision/models.html). General support for other PyTorch models is forthcoming."
+    "Amazon SageMaker Neo is an API to compile machine learning models to optimize them for our choice of hardware targets. Currently, Neo supports pre-trained PyTorch models from [TorchVision](https://pytorch.org/docs/stable/torchvision/models.html). General support for other PyTorch models is forthcoming.\n",
+    "\n",
+    "### Runtime\n",
+    "\n",
+    "This notebook takes approximately 8 minutes to run.\n",
+    "\n",
+    "### Contents\n",
+    "\n",
+    "1. [Import ResNet18 from TorchVision](#Import-ResNet18-from-TorchVision)\n",
+    "1. [Invoke Neo Compilation API](#Invoke-Neo-Compilation-API)\n",
+    "1. [Deploy the model](#Deploy-the-model)\n",
+    "1. [Send requests](#Send-requests)\n",
+    "1. [Delete the Endpoint](#Delete-the-Endpoint)"
    ]
   },
   {
@@ -25,7 +37,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "We'll import [ResNet18](https://arxiv.org/abs/1512.03385) model from TorchVision and create a model artifact `model.tar.gz`."
+    "We import the [ResNet18](https://arxiv.org/abs/1512.03385) model from TorchVision and create a model artifact `model.tar.gz`."
    ]
   },
   {
@@ -41,6 +53,13 @@
     "!{sys.executable} -m pip install --upgrade sagemaker"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Specify the input data shape. For more information, see [Prepare Model for Compilation](https://docs.aws.amazon.com/sagemaker/latest/dg/neo-compilation-preparing-model.html)."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -68,6 +87,13 @@
     "### Upload the model archive to S3"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Specify parameters for the compilation job and upload the `model.tar.gz` archive file."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -110,6 +136,13 @@
     "### Create a PyTorch SageMaker model"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Use the `PyTorchModel` and define parameters including the path to the model, the `entry_point` script that is used to perform inference, and other version and environment variables."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -139,6 +172,13 @@
     "### Use Neo compiler to compile the model"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Run the compilation job, which is saved in S3 at the specified `compiled_model_path` location."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -163,6 +203,13 @@
     "## Deploy the model"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Deploy the compiled model to an endpoint so it can be used for inference."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -183,11 +230,18 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "Let's try to send a cat picture.\n",
+    "Let's send a picture to the endpoint to predict the image subject.\n",
     "\n",
     "![title](cat.jpg)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Open the image and pass the payload as a bytearray to the predictor, receiving a response."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -206,6 +260,13 @@
     "print(\"Most likely class: {}\".format(np.argmax(result)))"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Use the ImageNet class ID response to look up which subject the image contains, and with what probability."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -217,12 +278,12 @@
     "with open(\"imagenet1000_clsidx_to_labels.txt\", \"r\") as f:\n",
     "    for line in f:\n",
     "        key, val = line.strip().split(\":\")\n",
-    "        object_categories[key] = val\n",
+    "        object_categories[key] = val.strip(\" \").strip(\",\")\n",
     "print(\n",
-    "    \"Result: label - \"\n",
-    "    + object_categories[str(np.argmax(result))]\n",
-    "    + \" probability - \"\n",
-    "    + str(np.amax(result))\n",
+    "    \"The label is\",\n",
+    "    object_categories[str(np.argmax(result))],\n",
+    "    \"with probability\",\n",
+    "    str(np.amax(result))[:5],\n",
     ")"
    ]
   },
@@ -231,7 +292,7 @@
    "metadata": {},
    "source": [
     "## Delete the Endpoint\n",
-    "Having an endpoint running will incur some costs. Therefore as a clean-up job, we should delete the endpoint."
+    "Delete the endpoint to avoid incurring costs now that it is no longer needed."
    ]
   },
   {
@@ -261,7 +322,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.6.10"
+   "version": "3.6.13"
   }
  },
  "nbformat": 4,