model training

The process that sets a model's weights by adjusting them, over many examples, until its predictions fit the data.

In more detail

A language model is pre-trained to predict the next token on a large text corpus, then post-trained, for example with human feedback, to follow instructions. Training happens once at the provider; an AI engineer mostly works with the finished model at inference time, steering it with context rather than retraining it.

Related terms

Taught in

References

Certification courses