Update examples

Author: ospillinger

examples/pytorch/image-classifier/README.md

@@ -37,11 +37,7 @@ The `predict()` function will be triggered once per request. The AlexNet model r
 # predictor.py
 
 def predict(sample, metadata):
-    if "url" in sample:
-        image = requests.get(sample["url"]).content
-    elif "base64" in sample:
-        image = base64.b64decode(sample["base64"])
-
+    image = requests.get(sample["url"]).content
     img_pil = Image.open(BytesIO(image))
     img_tensor = preprocess(img_pil)
     img_tensor.unsqueeze_(0)

examples/pytorch/image-classifier/predictor.py

@@ -1,5 +1,4 @@
 import requests
-import base64
 from PIL import Image
 from io import BytesIO
 from torchvision import transforms
@@ -21,11 +20,7 @@
 
 
 def predict(sample, metadata):
-    if "url" in sample:
-        image = requests.get(sample["url"]).content
-    elif "base64" in sample:
-        image = base64.b64decode(sample["base64"])
-
+    image = requests.get(sample["url"]).content
     img_pil = Image.open(BytesIO(image))
     img_tensor = preprocess(img_pil)
     img_tensor.unsqueeze_(0)

examples/pytorch/iris-classifier/README.md

@@ -25,9 +25,9 @@ def init(metadata):
     # download the model from S3 (location specified in the metadata field of our api configuration)
     s3 = boto3.client("s3")
     bucket, key = re.match(r"s3:\/\/(.+?)\/(.+)", metadata["model"]).groups()
-    s3.download_file(bucket, key, "iris_model.pth")
+    s3.download_file(bucket, key, "weights.pth")
 
-    model.load_state_dict(torch.load("iris_model.pth"))
+    model.load_state_dict(torch.load("weights.pth"))
     model.eval()
 ```
 
@@ -72,7 +72,7 @@ A `deployment` specifies a set of resources that are deployed together. An `api`
     path: src/predictor.py
     python_path: src/
     metadata:
-      model: s3://cortex-examples/pytorch/iris-classifier/nn.pth
+      model: s3://cortex-examples/pytorch/iris-classifier/weights.pth
   tracker:
     model_type: classification
 ```

examples/pytorch/iris-classifier/cortex.yaml

@@ -7,6 +7,6 @@
     path: src/predictor.py
     python_path: src/
     metadata:
-      model: s3://cortex-examples/pytorch/iris-classifier/nn.pth
+      model: s3://cortex-examples/pytorch/iris-classifier/weights.pth
   tracker:
     model_type: classification

examples/pytorch/iris-classifier/src/my_model.py renamed to examples/pytorch/iris-classifier/src/model.py

@@ -1,9 +1,9 @@
-from sklearn.datasets import load_iris
-from sklearn.model_selection import train_test_split
 import torch
 import torch.nn as nn
 import torch.nn.functional as F
 from torch.autograd import Variable
+from sklearn.datasets import load_iris
+from sklearn.model_selection import train_test_split
 from sklearn.metrics import accuracy_score
 
 
@@ -54,4 +54,4 @@ def forward(self, X):
 
     print("prediction accuracy {}".format(accuracy_score(test_y.data, predict_y.data)))
 
-    torch.save(model.state_dict(), "iris_model.pth")
+    torch.save(model.state_dict(), "weights.pth")

examples/pytorch/iris-classifier/src/predictor.py

@@ -1,7 +1,7 @@
-import torch
-from my_model import IrisNet
 import boto3
 import re
+import torch
+from model import IrisNet
 
 labels = ["iris-setosa", "iris-versicolor", "iris-virginica"]
 
@@ -11,8 +11,8 @@
 def init(metadata):
     s3 = boto3.client("s3")
     bucket, key = re.match(r"s3:\/\/(.+?)\/(.+)", metadata["model"]).groups()
-    s3.download_file(bucket, key, "iris_model.pth")
-    model.load_state_dict(torch.load("iris_model.pth"))
+    s3.download_file(bucket, key, "weights.pth")
+    model.load_state_dict(torch.load("weights.pth"))
     model.eval()
 
 

examples/sklearn/mpg-estimation/README.md

@@ -1,4 +1,4 @@
-# Deploy a Sklearn Linear Regression model
+# Deploy a scikit-learn linear regression model
 
 This example shows how to deploy a sklearn linear regression model trained on the MPG dataset.
 
@@ -18,12 +18,11 @@ model = None
 def init(metadata):
     global model
 
-    # download the model from S3 (location specified in the metadata field of our api configuration)
+    # download the model from S3 (path specified in the metadata field of our api configuration)
     s3 = boto3.client("s3")
     bucket, key = re.match(r"s3:\/\/(.+?)\/(.+)", metadata["model"]).groups()
-    s3.download_file(bucket, key, "mpg.joblib")
-
-    model = load("mpg.joblib")
+    s3.download_file(bucket, key, "linreg.joblib")
+    model = load("linreg.joblib")
 ```
 
 ### Predict
@@ -34,14 +33,15 @@ The `predict()` function will be triggered once per request. We extract the feat
 # predictor.py
 
 def predict(sample, metadata):
-    arr = [
+    input_array = [
         sample["cylinders"],
         sample["displacement"],
         sample["horsepower"],
         sample["weight"],
         sample["acceleration"],
     ]
-    result = model.predict([arr])
+
+    result = model.predict([input_array])
     return np.asscalar(result)
 ```
 
@@ -60,7 +60,7 @@ A `deployment` specifies a set of resources that are deployed together. An `api`
 - kind: api
   name: mpg
   predictor:
-    path: src/predictor.py
+    path: predictor.py
     metadata:
       model: s3://cortex-examples/sklearn/mpg-estimation/linreg.joblib
   tracker:

examples/sklearn/mpg-estimation/cortex.yaml

@@ -4,7 +4,7 @@
 - kind: api
   name: mpg
   predictor:
-    path: src/predictor.py
+    path: predictor.py
     metadata:
       model: s3://cortex-examples/sklearn/mpg-estimation/linreg.joblib
   tracker:

examples/sklearn/mpg-estimation/src/model.py renamed to examples/sklearn/mpg-estimation/model.py

@@ -3,7 +3,7 @@
 from sklearn.linear_model import LinearRegression
 from sklearn.model_selection import train_test_split
 from joblib import dump
-import boto3
+
 
 if __name__ == "__main__":
     df = pd.read_csv(
@@ -21,4 +21,4 @@
 
     regression_model = LinearRegression()
     regression_model.fit(X_train, y_train)
-    dump(regression_model, "mpg.joblib")
+    dump(regression_model, "linreg.joblib")

examples/sklearn/mpg-estimation/src/predictor.py renamed to examples/sklearn/mpg-estimation/predictor.py

@@ -1,7 +1,7 @@
-from joblib import load
 import boto3
-import numpy as np
 import re
+import numpy as np
+from joblib import load
 
 
 model = None
@@ -11,17 +11,18 @@ def init(metadata):
     global model
     s3 = boto3.client("s3")
     bucket, key = re.match(r"s3:\/\/(.+?)\/(.+)", metadata["model"]).groups()
-    s3.download_file(bucket, key, "mpg.joblib")
-    model = load("mpg.joblib")
+    s3.download_file(bucket, key, "linreg.joblib")
+    model = load("linreg.joblib")
 
 
 def predict(sample, metadata):
-    arr = [
+    input_array = [
         sample["cylinders"],
         sample["displacement"],
         sample["horsepower"],
         sample["weight"],
         sample["acceleration"],
     ]
-    result = model.predict([arr])
+
+    result = model.predict([input_array])
     return np.asscalar(result)