Merge pull request #1 from heltonricardo/config-train

Arquivos de configuração, treinamento e teste

.github/workflows/pipeline.yml

@@ -0,0 +1,91 @@
+name: Train Pipeline
+
+on:
+  push:
+  pull_request:
+    branches:
+      - layers
+      - config-train
+  workflow_dispatch:
+  # Allows external webhook trigger
+  repository_dispatch:
+    types:
+      - webhook
+jobs:
+  test_train:
+    name: Test train
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v1
+        with:
+          fetch-depth: 0
+
+      - name: Set up Python 3.9
+        uses: actions/setup-python@v1
+        with:
+          python-version: 3.9
+
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          if [ -f requirements.txt ]; then pip install -r requirements.txt; fi
+
+      - name: Test files
+        run: pytest test_train.py
+
+  check_python_code_style:
+    name: Check Python code style
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v1
+        with:
+          fetch-depth: 0
+
+      - name: Set up Python 3.9
+        uses: actions/setup-python@v1
+        with:
+          python-version: 3.9
+
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          if [ -f requirements.txt ]; then pip install -r requirements.txt; fi
+
+      - name: Validate Python Code Style
+        run: pycodestyle .
+
+  train_pipeline:
+    name: Train pipeline
+    runs-on: ubuntu-latest
+    needs: [test_train]
+    steps:
+      - uses: actions/checkout@v1
+        with:
+          fetch-depth: 0
+
+      - name: Set up Python 3.9
+        uses: actions/setup-python@v1
+        with:
+          python-version: 3.9
+
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          if [ -f requirements.txt ]; then pip install -r requirements.txt; fi
+
+      - name: Train Model
+        run: python train.py
+
+  build_image:
+    name: Build image
+    runs-on: ubuntu-latest
+    needs: [train_pipeline]
+    steps:
+      - uses: actions/checkout@v2
+      - name: Docker Login
+        run: docker login -u ${{secrets.DOCKER_USER}} -p ${{secrets.DOCKER_PASSWORD}}
+      - name: Build the Docker image
+        run: docker build . --file Dockerfile --tag ${{secrets.DOCKER_USER}}/${{secrets.DOCKER_IMAGE_NAME}}
+
+      - name: Docker Push
+        run: docker push ${{secrets.DOCKER_USER}}/${{secrets.DOCKER_IMAGE_NAME}}

Dockerfile

@@ -0,0 +1,7 @@
+FROM python:3.9
+WORKDIR /code
+COPY ./requirements.txt /code/requirements.txt
+RUN pip install --no-cache-dir --upgrade -r /code/requirements.txt
+COPY ./app /code/app
+
+CMD ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "80"]

environment.yml

@@ -0,0 +1,6 @@
+name: mlops-env
+dependencies:
+  - python=3.9
+  - pip
+  - pip:
+      - -r requirements.txt

main.py

@@ -0,0 +1,117 @@
+import os
+import mlflow
+import numpy as np
+from pydantic import BaseModel
+from fastapi import FastAPI
+
+
+class FetalHealthData(BaseModel):
+    accelerations: float
+    fetal_movement: float
+    uterine_contractions: float
+    severe_decelerations: float
+
+
+app = FastAPI(
+    title="Fetal Health API",
+    openapi_tags=[
+        {"name": "Health", "description": "Get api health"},
+        {"name": "Prediction", "description": "Model prediction"},
+    ],
+)
+
+
+def load_model():
+    """
+    Loads a pre-trained model from an MLflow server.
+
+    This function connects to an MLflow server using the provided tracking URI,
+    username, and password.
+    It retrieves the latest version of the 'fetal_health' model registered on
+    the server.
+    The function then loads the model using the specified run ID and returns
+    the loaded model.
+
+    Returns:
+        loaded_model: The loaded pre-trained model.
+
+    Raises:
+        None
+    """
+    print("reading model...")
+    MLFLOW_TRACKING_URI = \
+        "https://dagshub.com/renansantosmendes/mlops-ead.mlflow"
+    MLFLOW_TRACKING_USERNAME = "renansantosmendes"
+    MLFLOW_TRACKING_PASSWORD = "b63baf8c662a23fa00deb74ba86600278769e5dd"
+    os.environ["MLFLOW_TRACKING_USERNAME"] = MLFLOW_TRACKING_USERNAME
+    os.environ["MLFLOW_TRACKING_PASSWORD"] = MLFLOW_TRACKING_PASSWORD
+    print("setting mlflow...")
+    mlflow.set_tracking_uri(MLFLOW_TRACKING_URI)
+    print("creating client..")
+    client = mlflow.MlflowClient(tracking_uri=MLFLOW_TRACKING_URI)
+    print("getting registered model...")
+    registered_model = client.get_registered_model("fetal_health")
+    print("read model...")
+    run_id = registered_model.latest_versions[-1].run_id
+    logged_model = f"runs:/{run_id}/model"
+    loaded_model = mlflow.pyfunc.load_model(logged_model)
+    print(loaded_model)
+    return loaded_model
+
+
+@app.on_event(event_type="startup")
+def startup_event():
+    """
+    A function that is called when the application starts up. It loads a model
+    into the global variable `loaded_model`.
+
+    Parameters:
+        None
+
+    Returns:
+        None
+    """
+    global loaded_model
+    loaded_model = load_model()
+
+
+@app.get(path="/", tags=["Health"])
+def api_health():
+    """
+    A function that represents the health endpoint of the API.
+
+    Returns:
+        dict: A dictionary containing the status of the API, with the key
+        "status" and the value "healthy".
+    """
+    return {"status": "healthy"}
+
+
+@app.post(path="/predict", tags=["Prediction"])
+def predict(request: FetalHealthData):
+    """
+    Predicts the fetal health based on the given request data.
+
+    Args:
+        request (FetalHealthData): The request data containing the fetal
+        health parameters.
+
+    Returns:
+        dict: A dictionary containing the prediction of the fetal health.
+
+    Raises:
+        None
+    """
+    global loaded_model
+    received_data = np.array(
+        [
+            request.accelerations,
+            request.fetal_movement,
+            request.uterine_contractions,
+            request.severe_decelerations,
+        ]
+    ).reshape(1, -1)
+    print(received_data)
+    prediction = loaded_model.predict(received_data)
+    print(prediction)
+    return {"prediction": str(np.argmax(prediction[0]))}