CI/CD for Machine Learning
| Course Details | ||||||||
|---|---|---|---|---|---|---|---|---|
|
||||||||
| Languages | ||||||||
|
||||||||
| Outcomes | ||||||||
|
||||||||
| Price & Terms | ||||||||
|
||||||||
| Technical | ||||||||
|
||||||||
| Specification | ||||||||
|
||||||||
Build GitHub Actions pipelines that train models, version datasets with DVC and automate hyperparameter tuning for machine learning projects.
Continuous integration and delivery bring automation, reliability and reproducibility to machine learning development. You write YAML workflows that run on GitHub Actions, train a classification model inside a pipeline, track datasets with Data Version Control, and compare model performance between branches before searching for better hyperparameters.
What You Will Achieve
- Write YAML workflows that run Python code whenever code is pushed or a pull request opens
- Pass credentials to a workflow through secrets, environment variables and the GITHUB_TOKEN
- Version datasets with DVC, store them on a remote, and train a classification model through a reproducible pipeline
- Compare metrics between Git branches, tune a model with GridSearchCV, and open a pull request with the best settings automatically
Program Overview
CI/CD Concepts and YAML Syntax
Sets out how continuous practices fit the software development life cycle and machine learning experiments, then introduces the YAML files that define pipelines.
- Build, test and deploy stages
- Continuous integration, delivery and deployment compared
- Machine learning workflow and experimentation
- YAML mappings and sequences
GitHub Actions Workflows
Breaks a workflow into its parts and builds basic CI pipelines that react to repository events.
- Events, actions, jobs, steps, runners and context
- Push and pull request triggers on feature branches
- Reading the workflow log
- Environment variables, secrets and GITHUB_TOKEN
Model Training, Data Versioning and Hyperparameter Tuning
Moves model work into the pipeline with Continuous Machine Learning (CML) and DVC, then compares training runs and automates the search for better settings.
- CML training jobs and markdown reports
- DVC tracking, .dvc files and remotes
- DVC pipelines defined in dvc.yaml
- Branch comparisons, GridSearchCV and automated pull requests
Certificate Awarded (Paid)
Job Outlook
These skills apply wherever machine learning models are trained and maintained as software, in teams that keep code, data and model results in shared Git repositories and automate their updates.
Good for: Practitioners who want model training, dataset tracking and tuning to run automatically each time their repository changes.
Keep in mind: The hands-on pipelines use GitHub Actions, DVC and CML. Continuous deployment is defined in the first chapter, while the later chapters center on integration, training and tuning.
Bottom line: A tool-specific course on wiring DVC and CML into GitHub Actions so training runs and tuning results are tracked and compared.
Frequently Asked Questions
Which CI/CD platform do the exercises use?
The workflows run on GitHub Actions. You define them in YAML, trigger them with push and pull request events, choose runner machines, and read the workflow log to see what each job did.
What role does DVC play in the pipelines?
Data Version Control tracks changes to datasets through .dvc files and stores the data on a remote. Its pipelines, configured in a dvc.yaml file, make model training repeatable and record metrics and plots you can compare across branches.
What is CML used for in this course?
The Continuous Machine Learning GitHub Action runs model training inside a workflow and produces a markdown report that holds the trained model's metrics and plots, so each run leaves a readable record.
How is hyperparameter tuning automated?
Tuning with scikit-learn's GridSearchCV is added to the DVC pipeline and run in its own GitHub Actions workflow. The workflow then opens a pull request that carries the best model configuration it found.
User Reviews
Write your review
You must be logged in to post a review.
Reviews can only be posted by registered learners. Log in or create a free account, then you will return here to post.
There are no reviews yet.