Custom training

Summary

Custom Training with Vertex AI Workbench allows developers to code their own machine learning models, controlling the entire ML development process from data preparation to deployment. It offers two environment options: pre-built containers with all necessary dependencies and libraries, or custom containers where developers define their tools. This flexibility caters to various ML training needs, making it ideal for those who require specific platforms like TensorFlow, PyTorch, or scikit-learn.

Takeaways

• 💡 Custom Training allows for full control over the ML development process from data preparation to deployment.
• 🛠️ You can code your own machine learning model using custom Training with Vertex AI Workbench.
• 💻 Vertex AI Workbench provides a single development environment for the entire data science workflow.
• 📈 It supports activities from exploring data to training and deploying machine learning models with code.
• 🏗️ Before coding, you must decide on the environment for your ML training code: pre-built or custom container.
• 🏠 A pre-built container is like a fully furnished kitchen with all necessary dependencies and libraries.
• 🔧 A custom container is like an empty room, allowing you to define the exact tools needed for the job.
• 🌐 If your ML training requires platforms like TensorFlow, PyTorch, Scikit-learn, or XGBoost, a pre-built container is recommended.
• 🛠️ Custom containers give you the flexibility to include only the necessary tools and dependencies for your specific ML tasks.
• 🔑 The choice between pre-built and custom containers depends on the specific needs and complexity of your ML project.

Q & A

What is custom training?

Custom training is the process of coding your own machine learning environment to control the entire machine learning development process, from data preparation to model deployment.

What is the purpose of using Vertex AI Workbench for custom training?

Vertex AI Workbench provides a single development environment for the entire data science workflow, from exploring data to training and deploying machine learning models with code.

What are the two options available for the environment in custom training?

The two options available for the environment in custom training are using a pre-built container or a custom container.

What is a pre-built container in the context of machine learning training?

A pre-built container is a fully furnished environment with all the necessary dependencies and libraries, like a kitchen with cabinets and appliances, suitable for ML training that requires platforms such as TensorFlow, PyTorch, Scikit-learn, or XGBoost.

What does a pre-built container include?

A pre-built container includes all the dependencies, libraries, and tools needed for machine learning training, such as cookware in a kitchen, representing the necessary tools to complete the job.

What is a custom container?

A custom container is an empty environment where you define and include only the exact tools and dependencies required for your specific machine learning training job.

Why would someone choose a custom container over a pre-built one?

A custom container would be chosen when the user has specific requirements that are not met by the pre-built container, allowing for a tailored environment to fit their unique needs.

How does the analogy of a kitchen relate to machine learning training environments?

The analogy of a kitchen is used to describe the completeness of a pre-built container, where everything is provided like in a fully furnished kitchen, whereas a custom container is likened to an empty room where you bring in only what you need.

What are some examples of platforms that might be included in a pre-built container?

Examples of platforms that might be included in a pre-built container are TensorFlow, PyTorch, Scikit-learn, and XGBoost.

What is the first step before coding begins in custom training?

The first step before coding begins in custom training is to determine the environment that the ML training code will use, choosing between a pre-built container or a custom container.

What is the role of Python code in custom training with a pre-built container?

Python code plays a crucial role in custom training with a pre-built container, as it is used to work with the included machine learning platforms and to develop the training scripts.