Vertex AI
Unified ML platform for building ML solutions end-to-end
Provides Deep Learning Environments (DL VM, DL Containers)
Supports Vertex Notebooks for writing code (JupyterLab under the hood)
Provides powerful pretrained ML models trained using Google’s data
NOTE
- Vertex AI is the unification of AI Platform and the addition of AutoML
- Even though the name contains AI this service is used for ML and DL
AutoML
Allows to easily train high-quality, custom ML models
Upload data to AutoML tables and choose what needs to be predicted and the rest is handles on its own
AI Platform
This service is deprecated and Vertex AI is now recommended
Supports preparing of datasets for supervised training with Data Labeling
Notebooks to write and document building ML models
A Model registry to hold all your trained models
Pipelines for setting up automated CI/CD to rapidly deploy new changes (MLOps)
TensorFlow Enterprise
Low-level deep learning machine learning platform created by the Google Brain Team
Written in Python, C++ and CUDA, provides APIs to use other languages
A Tensor is a multi-dimensional Array (Similar to NumPy ndarray objects)
They can reside in accelerator memory (like a GPU)
It is an special data structure that is created for machine learning workloads
Google has their own hardware called Tensor Processing Units (TPU) which is specially optimized for tensor data structure
TPUs are ~ 50 times more powerful that traditional chips
Cloud GPUs
Add GPUs to your workloads for machine learning, scientific computing and 3D Visualization
These are referred to as fractional GPUs
The preconfigured ML/DL VMs and Containers that contain GPUs are extremely costly so in place of them Fractional GPUs can be used for light and short workloads