Evaluation & Operations
Measure speed and quality honestly, pick the right model, and handle licenses and privacy.
0 of 4 complete
- ▶ Up next12 min+50 XPMeasuring speed correctlyDefinitions, a repeatable method with llama-bench and a client-side streaming script, how to report prefill and…
- Pro14 min+50 XPMeasuring quality: perplexity, KL divergence and task evalsHow to tell whether a quant, a model swap or a prompt change made things better or worse: perplexity and KL…
- 12 min+50 XPModel licenses, provenance and privacyWhat 'open weights' really means, the main licence families and what they restrict, how to keep a licence record, how…
- Pro13 min+50 XPA framework for choosing the right modelA repeatable process from requirements to shortlist to decision: size classes, memory fit, dense vs MoE vs reasoning…