AI Foundations
Batch
A small group of training examples processed together in one training step.
Example
Instead of processing a million examples at once, training handles groups of 256.
Why it matters
It helps builders use computing equipment efficiently while keeping learning manageable.
You'll hear it
“This group is too large to fit in the available memory.”
Can you control it?
Set by the team
The training team chooses batch size based on the model, hardware, data, and training behavior. Users of the completed model do not control it.
Common questions
- Must every batch contain exactly the same number of examples?
- Not always. A final group may be smaller, depending on how the teaching material is divided.
- Why not train on the whole dataset at once?
- Smaller batches use less memory and let the model update repeatedly as it moves through the data.
- Does batch size affect model quality?
- Yes. Group size can change how steadily learning progresses, as well as its speed and memory needs.