Parameters
Parameter
The learned numbers inside a model. Their count is the model size.
Parameters are the numbers a model learns during training, and their count is what people mean by model size: billions or trillions of them. They are not readable as rules. One parameter might encode a rough fact about the world, another a statistical tendency in language, and nobody can point to a single number and say what it stands for. What the model does comes from the pattern across all of them, not from any individual value.
Related terms
Model
The trained file of learned parameters that turns input into output.
Training
The expensive process that sets a model's parameters by exposing it to data.
Quantization
Storing weights in fewer bits so a model runs with less memory and cost.
Fine-tuning
Further training of an existing model on your own data to shape its behaviour.
Want this working inside your team?
It starts with one conversation about your workflow.