vertex Model Registry and Deployer

vertex-registry-and-deployer/.copier-answers.yml

@@ -0,0 +1,8 @@
+# Changes here will be overwritten by Copier
+_commit: 2024.09.24
+_src_path: gh:zenml-io/template-starter
+email: [email protected]
+full_name: ZenML GmbH
+open_source_license: apache
+project_name: ZenML Starter
+version: 0.1.0

vertex-registry-and-deployer/.dockerignore

@@ -0,0 +1,2 @@
+.venv*
+.requirements*

vertex-registry-and-deployer/LICENSE

@@ -0,0 +1,15 @@
+Apache Software License 2.0
+
+Copyright (c) ZenML GmbH 2024. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.

vertex-registry-and-deployer/README.md

@@ -0,0 +1,87 @@
+# 🚀 Deploying ML Models with ZenML on Vertex AI
+
+
+Welcome to your ZenML project for deploying ML models using Google Cloud's Vertex AI! This project provides a hands-on experience with MLOps pipelines using ZenML and Vertex AI. It contains a collection of ZenML steps, pipelines, and other artifacts to help you efficiently deploy your machine learning models.
+
+Using these pipelines, you can run data preparation, model training, registration, and deployment with a single command while using YAML files for [configuration](https://docs.zenml.io/user-guide/production-guide/configure-pipeline). ZenML takes care of tracking your metadata and [containerizing your pipelines](https://docs.zenml.io/how-to/customize-docker-builds).
+
+
+## 🏃 How to run
+
+In this project, we will train and deploy a classification model to [Vertex AI](https://cloud.google.com/vertex-ai). Before running any pipelines, set up your environment as follows, we need to set up our environment as follows:
+
+```bash
+# Set up a Python virtual environment, if you haven't already
+python3 -m venv .venv
+source .venv/bin/activate
+
+# Install requirements
+pip install -r requirements.txt
+```
+
+We will need to set up access to Google Cloud and Vertex AI. You can follow the instructions in the [ZenML documentation](https://docs.zenml.io/how-to/auth-management/gcp-service-connector)
+to register a service connector and set up your Google Cloud credentials.
+
+Once you have set up your Google Cloud credentials, we can create a stack and run the deployment pipeline:
+
+```bash
+# Register the artifact store
+zenml artifact-store register gs_store -f gcp --path=gs://bucket-name
+zenml artifact-store connect gs_store --connector gcp
+
+# Register the model registry
+zenml model-registry register vertex_registry --flavor=vertex --location=us-central1
+zenml model-registry connect vertex_registry --connector gcp
+
+# Register Model Deployer
+zenml model-deployer register vertex_deployer --flavor=vertex --location=us-central1
+zenml model-deployer connect vertex_deployer --connector vertex_deployer_connector
+
+# Register the stack
+zenml stack register vertex_stack --orchestrator default --artifact-store gs_store --model-registry vertex_registry --model-deployer vertex_deployer
+```
+
+Now that we have set up our stack, we can run the training pipeline, which will train and register the model into the Vertex AI model registry and Deploys it into Vertex AI endpoint.
+
+```bash
+python run.py --training-pipeline
+```
+
+Once the pipeline has completed, you can check the status of the model in the Vertex AI model registry and the deployed model in the Vertex AI endpoint.
+
+```bash
+# List models in the model registry
+zenml model-registry models list
+
+# List deployed models
+zenml model-deployer models list
+```
+
+You can also run the deployment pipeline separately:
+
+```bash
+python run.py --inference-pipeline
+```
+
+
+## 📜 Project Structure
+
+The project loosely follows [the recommended ZenML project structure](https://docs.zenml.io/how-to/setting-up-a-project-repository/best-practices):
+
+```
+.
+├── configs                                     # Pipeline configuration files
+│   ├── training.yaml                           # Configuration for training pipeline
+│   ├── inference.yaml                          # Configuration for inference pipeline
+├── pipelines                                   # `zenml.pipeline` implementations
+│   ├── training.py                             # Training pipeline
+│   ├── inference.py                            # Inference pipeline
+├── steps                                       # `zenml.step` implementations
+│   ├── model_trainer.py                        # Model training step
+│   ├── model_register.py                       # Model registration step
+│   ├── model_promoter.py                       # Model promotion step
+│   ├── model_deployer.py                       # Model deployment step to Vertex AI
+├── README.md                                   # This file
+├── requirements.txt                            # Extra Python dependencies
+└── run.py                                      # CLI tool to run pipelines with ZenML                                    # CLI tool to run pipelines on ZenML Stack
+```

vertex-registry-and-deployer/configs/inference.yaml

@@ -0,0 +1,16 @@
+# environment configuration
+settings:
+  docker:
+    required_integrations:
+      - sklearn
+      - pandas
+    requirements:
+      - pyarrow
+
+# configuration of the Model Control Plane
+model:
+  name: "breast_cancer_classifier"
+  version: "production"
+  license: Apache 2.0
+  description: A breast cancer classifier
+  tags: ["breast_cancer", "classifier"]

vertex-registry-and-deployer/configs/training_sgd.yaml

@@ -0,0 +1,16 @@
+# environment configuration
+settings:
+  docker:
+    required_integrations:
+      - sklearn
+      - pandas
+    requirements:
+      - pyarrow
+
+# configuration of the Model Control Plane
+model:
+  name: breast_cancer_classifier
+  version: sgd
+  license: Apache 2.0
+  description: A breast cancer classifier
+  tags: ["breast_cancer", "classifier"]

vertex-registry-and-deployer/pipelines/__init__.py

@@ -0,0 +1,19 @@
+# Apache Software License 2.0
+#
+# Copyright (c) ZenML GmbH 2024. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+from .inference import inference
+from .training import training

vertex-registry-and-deployer/pipelines/inference.py

@@ -0,0 +1,45 @@
+# Apache Software License 2.0
+#
+# Copyright (c) ZenML GmbH 2024. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+from zenml import get_pipeline_context, pipeline
+from zenml.logger import get_logger
+
+logger = get_logger(__name__)
+
+
+@pipeline
+def inference(random_state: int, target: str):
+    """
+    Model inference pipeline.
+
+    This is a pipeline that loads the inference data, processes it with
+    the same preprocessing pipeline used in training, and runs inference
+    with the trained model.
+
+    Args:
+        random_state: Random state for reproducibility.
+        target: Name of target column in dataset.
+    """
+    # Get the production model artifact
+    model = get_pipeline_context().model.get_artifact("sklearn_classifier")
+
+    # Get the preprocess pipeline artifact associated with this version
+    preprocess_pipeline = get_pipeline_context().model.get_artifact(
+        "preprocess_pipeline"
+    )
+
+    # Link all the steps together by calling them and passing the output
+    #  of one step as the input of the next step.

vertex-registry-and-deployer/pipelines/training.py

@@ -0,0 +1,55 @@
+# Apache Software License 2.0
+#
+# Copyright (c) ZenML GmbH 2024. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+#
+
+from typing import Optional
+from uuid import UUID
+
+from steps import model_deployer, model_promoter, model_register, model_trainer
+from zenml import pipeline
+from zenml.client import Client
+from zenml.logger import get_logger
+
+logger = get_logger(__name__)
+
+
+@pipeline
+def training(
+    target: Optional[str] = "target",
+):
+    """Model training pipeline.
+
+    This is a pipeline that loads the data from a preprocessing pipeline,
+    trains a model on it and evaluates the model. If it is the first model
+    to be trained, it will be promoted to production. If not, it will be
+    promoted only if it has a higher accuracy than the current production
+    model version.
+
+    Args:
+        train_dataset_id: ID of the train dataset produced by feature engineering.
+        test_dataset_id: ID of the test dataset produced by feature engineering.
+        target: Name of target column in dataset.
+        model_type: The type of model to train.
+    """
+    # Link all the steps together by calling them and passing the output
+    # of one step as the input of the next step.
+
+    model, accuracy = model_trainer(target=target)
+    is_promoted = model_promoter(accuracy=accuracy)
+    if is_promoted:
+        model_registry_uri = model_register()
+        model_deployer(model_registry_uri=model_registry_uri)
+

vertex-registry-and-deployer/requirements.txt

@@ -0,0 +1,5 @@
+zenml[server]>=0.70.1
+notebook
+scikit-learn
+pyarrow
+pandas