Add GCP custom predictor files and calls

model/api_calling.ipynb

@@ -0,0 +1,158 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "%load_ext autoreload\n",
+    "%autoreload 2"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import numpy as np"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "PROJECT_NAME = \"experiment-week\"\n",
+    "MODEL_NAME = \"garbage_model_h5_128_custom_prep\""
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Data prep"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "images_test = np.load('images.npy')\n",
+    "images_test.shape"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## Local API"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Make sure serialized version of your model with name `model.h5` is in the specified path: `models/{MODEL_NAME}`"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "from gcp_custom_predictor import MyPredictor"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "my_pred = MyPredictor.from_path(f'models/{MODEL_NAME}')"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## GCP API"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Calling `GCP Custom prediction routines` (includes preprocessing step):<br/>\n",
+    "https://cloud.google.com/ai-platform/prediction/docs/custom-prediction-routines "
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import gcp_api\n",
+    "from gcp_api import predict_json"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def predict_instances(instances, n=-1, return_imax=True):\n",
+    "    for row in instances[:n]:\n",
+    "        prediction = predict_json(PROJECT_NAME, MODEL_NAME, [row.tolist()])[0]\n",
+    "        yield np.argmax(prediction) if  return_imax else prediction"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "list(predict_instances(images_test, n=6))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "list(predict_instances(images_test, n=1, return_imax=False))"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.7.4"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}

model/gcp_api.py

@@ -0,0 +1,41 @@
+import googleapiclient.discovery
+from google.oauth2 import service_account
+
+# authentication
+SA_FILE = 'gcp_secret.json'
+CREDENTIALS = service_account.Credentials.from_service_account_file(SA_FILE)
+SERVICE = googleapiclient.discovery.build('ml', 'v1', credentials=CREDENTIALS)
+
+
+def predict_json(project, model, instances, service=SERVICE, version=None):
+    """Send json data to a deployed model for prediction.
+
+    Args:
+        project (str): project where the AI Platform Model is deployed.
+        model (str): model name.
+        instances ([Mapping[str: Any]]): Keys should be the names of Tensors
+            your deployed model expects as inputs. Values should be datatypes
+            convertible to Tensors, or (potentially nested) lists of datatypes
+            convertible to tensors.
+        service: authenticated service.
+        version: str, version of the model to target.
+    Returns:
+        Mapping[str: any]: dictionary of prediction results defined by the
+            model.
+    """
+    # Create the AI Platform service object.
+    # To authenticate set the environment variable
+    name = 'projects/{}/models/{}'.format(project, model)
+
+    if version is not None:
+        name += '/versions/{}'.format(version)
+
+    response = service.projects().predict(
+        name=name,
+        body={'instances': instances}
+    ).execute()
+
+    if 'error' in response:
+        raise RuntimeError(response['error'])
+
+    return response['predictions']

model/gcp_custom_predictor.py

@@ -0,0 +1,52 @@
+import os
+import numpy as np
+from tensorflow import keras
+
+
+class MyPredictor(object):
+    """An example Predictor for an AI Platform custom prediction routine."""
+
+    def __init__(self, model):
+        """Stores artifacts for prediction. Only initialized via `from_path`.
+        """
+        self._model = model
+
+    def predict(self, instances, **kwargs):
+        """Performs custom prediction.
+
+        Preprocesses inputs, then performs prediction using the trained Keras
+        model.
+
+        Args:
+            instances: A list of prediction input instances.
+            **kwargs: A dictionary of keyword args provided as additional
+                fields on the predict request body.
+
+        Returns:
+            A list of outputs containing the prediction results.
+        """
+        inputs = np.asarray(instances)
+        preprocessed_inputs = keras.applications.resnet50.preprocess_input(inputs)
+        outputs = self._model.predict(preprocessed_inputs)
+        return outputs.tolist()
+
+    @classmethod
+    def from_path(cls, model_dir):
+        """Creates an instance of MyPredictor using the given path.
+
+        This loads artifacts that have been copied from your model directory in
+        Cloud Storage. MyPredictor uses them during prediction.
+
+        Args:
+            model_dir: The local directory that contains the trained Keras
+                model and the pickled preprocessor instance. These are copied
+                from the Cloud Storage model directory you provide when you
+                deploy a version resource.
+
+        Returns:
+            An instance of `MyPredictor`.
+        """
+        model_path = os.path.join(model_dir, 'model.h5')
+        model = keras.models.load_model(model_path)
+
+        return cls(model)

model/setup.py

@@ -0,0 +1,6 @@
+from setuptools import setup
+
+setup(
+    name='gcp_custom_predictor',
+    version='0.1',
+    scripts=['gcp_custom_predictor.py'])