Samplers and probes

Inferlets that exercise the forward-pass output surface — every sampler kind, every probe kind, scoring candidate strings, watermarking generated text.

Inferlet	What it shows
sampler-suite	End-to-end correctness check for every sampler kind (token / distribution / raw-logits / logprob / logprobs / entropy) attached to the same forward-pass slot.
raw-logits-demo	Microbench for Sampler::raw_logits(). Measures per-call overhead vs. greedy.
output-validation	Score candidate strings by their normalized log-probability (yes/no, A/B/C/D).
best-of-n	Best-of-N generation with diversity ranking.
watermarking	Text watermarking via green/red token partitioning.

Related guide

• Samplers and probabilities: every Sampler and Probe kind.
• The forward pass: the primitive these inferlets drive directly.
• Forking and saving: the fork mechanic best-of-n builds on.