Topics
Topics
The six subjects the notes are grouped under, with the number of notes in each.
- Evaluation
Deciding what a comparison is actually measuring before any model is run.
2 notes - Benchmarks
Reading published scores, and understanding what a single number hides.
2 notes - Context
How input is segmented, counted and carried through a long request.
2 notes - Cost & Latency
The three costs of a request and why they rarely move together.
2 notes - Prompting
Wording, sampling settings and the variance they introduce into a test.
1 notes - Deployment
What changes about a comparison when the weights run on your own hardware.
1 notes