refactor(evaluate): test

.github/workflows/ci.yml

@@ -66,9 +66,8 @@ jobs:
       run: |
         pipenv install --dev
         cd src
-        # pipenv run coverage run -m unittest tests.inference_tests
-        # pipenv run coverage report
-        pipenv run python -m unittest tests.inference_tests
+        pipenv run coverage run -m unittest tests.inference_tests
+        pipenv run coverage report
   build_python_packages:
     runs-on: ubuntu-latest
     needs: packages_unit_tests

packages/inference/src/mlopspython_inference/inference_pillow.py

@@ -7,7 +7,7 @@
 
 
 # load and prepare the image
-def load_image(filename: str|BytesIO):
+def load_image(filename: str | BytesIO):
     # load the image
     img = load_img(filename, target_size=(224, 224))
     # convert to array
@@ -23,12 +23,25 @@ def load_image(filename: str|BytesIO):
 BASE_PATH = Path(__file__).resolve().parent
 
 
+class IModelLoader:
+    def load(self):
+        raise NotImplementedError()
+
+
+class ModelLoader(IModelLoader):
+    def __init__(self, model_path: str):
+        self.model_path = model_path
+
+    def load(self):
+        return load_model(self.model_path)
+
+
 class Inference:
-    def __init__(self, logging, model_path: str):
+    def __init__(self, logging, model_loader: IModelLoader):
         self.logger = logging.getLogger(__name__)
-        self.model = load_model(model_path)
+        self.model = model_loader.load()
 
-    def execute(self, filepath:str|BytesIO):
+    def execute(self, filepath: str | BytesIO):
         img = load_image(filepath)
         result = self.model.predict(img)
         values = [float(result[0][0]), float(result[0][1]), float(result[0][2])]

packages/inference/src/tests/inference_tests.py

@@ -1,8 +1,9 @@
 import unittest
+from dataclasses import dataclass
 from pathlib import Path
 import logging
-
-from mlopspython_inference.inference_pillow import Inference
+from unittest.mock import MagicMock
+from mlopspython_inference.inference_pillow import Inference, IModelLoader
 
 BASE_PATH = Path(__file__).resolve().parent
 output_directory = BASE_PATH / "output"
@@ -11,7 +12,14 @@
 class TestInference(unittest.TestCase):
 
     def test_inference(self):
-        inference = Inference(logging, str(input_directory / "model" / "final_model.h5"))
+        def predict(img):
+            return [[1.0, 2.370240289845506e-30, 0.0]]
+        model_mock = MagicMock()
+        model_mock.predict = predict
+
+        model_loader_mock = MagicMock(IModelLoader)
+        model_loader_mock.load = MagicMock(return_value=model_mock)
+        inference = Inference(logging, model_loader_mock)
         inference_result = inference.execute(str(input_directory / "images" / "cat.png"))
 
         expected_result = {'prediction': 'Cat', 'values': [1.0, 2.370240289845506e-30, 0.0]}

production/api/core/model/inference_pillow.py

@@ -1,14 +1,14 @@
 import io
 from pathlib import Path
-from mlopspython_inference.inference_pillow import Inference as InferencePillow
+from mlopspython_inference.inference_pillow import Inference as InferencePillow, ModelLoader
 
 BASE_PATH = Path(__file__).resolve().parent
 
 class Inference:
     def __init__(self, logging, app_settings):
         self.logger = logging.getLogger(__name__)
         model_path = (str(BASE_PATH / 'final_model.h5'))
-        self.inference = InferencePillow(logging, str(model_path))
+        self.inference = InferencePillow(logging, ModelLoader(str(model_path)))
 
     def execute(self, file, filename, settings=None):
         return self.inference.execute(io.BytesIO(file.read()))

production/api/tests/test_process.py

@@ -11,7 +11,7 @@ class TestProcess(unittest.IsolatedAsyncioTestCase):
     def test_process(self):
         model = Inference(logging, BaseAppSettings(logging))
         dir = "./tests/cats-dogs-others/f2b68110-3b7b-4cc1-98fc-caf451d71ff2/mlcli"
-        dest = "./tests/diff"
+        dest = "./tests/diff2"
         flist = [p for p in Path(dir).iterdir() if p.is_file()]
         for fp in flist:
             with open(fp, "rb") as pdf_stream:

train/evaluate/evaluate.py

@@ -1,13 +1,13 @@
 import json
 from pathlib import Path
 
-from mlopspython_inference.inference_pillow import Inference
+from mlopspython_inference.inference_pillow import Inference, ModelLoader
 
 
 def evaluate(logging, input_model_directory: Path, input_images_directory: Path, output_directory: Path,
              integration_output: Path):
     model_path = input_model_directory / "final_model.h5"
-    model = Inference(logging, str(model_path))
+    model = Inference(logging, ModelLoader(str(model_path)))
 
     statistics = {
                   "ok": 0,