example app

Author: alvinwan

examples/app/README.md

@@ -0,0 +1,3 @@
+super simple flask app for serving NBDT predictions
+
+deployed to repl.it (link tbd)

examples/app/app.py

@@ -0,0 +1,41 @@
+from flask import Flask
+from nbdt.model import SoftNBDT
+from nbdt.models import wrn28_10_cifar10
+from torchvision import transforms
+from nbdt.utils import DATASET_TO_CLASSES, load_image_from_path, maybe_install_wordnet
+
+app = Flask(__name__)
+
+
+@app.route('/')
+def home():
+    # load pretrained NBDT
+    model = wrn28_10_cifar10()
+    model = SoftNBDT(
+      pretrained=True,
+      dataset='CIFAR10',
+      arch='wrn28_10_cifar10',
+      hierarchy='wordnet',
+      model=model)
+
+    # load + transform image
+    im = load_image_from_path("https://images.pexels.com/photos/1170986/pexels-photo-1170986.jpeg?auto=compress&cs=tinysrgb&dpr=2&w=32")
+    transform = transforms.Compose([
+      transforms.Resize(32),
+      transforms.CenterCrop(32),
+      transforms.ToTensor(),
+      transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)),
+    ])
+    x = transform(im)[None]
+
+    # run inference
+    outputs, decisions = model.forward_with_decisions(x)  # use `model(x)` to obtain just logits
+    _, predicted = outputs.max(1)
+    return {
+        'predicted': [DATASET_TO_CLASSES['CIFAR10'][pred] for pred in predicted],
+        'decisions': [[info['name'] for info in decision] for decision in decisions]
+    }
+
+
+if __name__ == '__main__':
+    app.run(host='0.0.0.0', port=8080)

examples/app/pyproject.toml

@@ -0,0 +1,10 @@
+[tool]
+[tool.poetry]
+name = "nbdt-demo"
+version = "0.0.1"
+description = ""
+authors = ["Alvin Wan <hi@alvinwan.com>"]
+[tool.poetry.dependencies]
+python = "^3.7"
+nbdt = "^0.0.1"
+flask = "^1.1.1"

examples/app/requirements.txt

@@ -0,0 +1,2 @@
+nbdt==0.0.1
+flask==1.1.1

nbdt/bin/nbdt

@@ -22,13 +22,13 @@ model = SoftNBDT(
 
 # load + transform image
 im = load_image_from_path(sys.argv[1])
-transforms = transforms.Compose([
+transform = transforms.Compose([
   transforms.Resize(32),
   transforms.CenterCrop(32),
   transforms.ToTensor(),
   transforms.Normalize((0.4914, 0.4822, 0.4465), (0.2023, 0.1994, 0.2010)),
 ])
-x = transforms(im)[None]
+x = transform(im)[None]
 
 # run inference
 outputs, decisions = model.forward_with_decisions(x)  # use `model(x)` to obtain just logits