Streaming likelihood + memory-aware auto (draft; Tasks 1-3 + v3 plan)#23

Draft

edeno wants to merge 15 commits into

mainfrom

streaming-likelihood-hmm

Commits on Apr 20, 2026

Audit streaming-likelihood plan for current main

edeno
and
claude
committed
Rewrite streaming plan around memory, not compute redundancy

edeno
and
claude
committed
Task 1: _resolve_n_chunks memory-aware heuristic

edeno
and
claude
committed
Task 2: Flip n_chunks default to 'auto' on all predict signatures

edeno
and
claude
committed
Task 3: guard return_outputs='log_likelihood' when streaming

edeno
and
claude
committed
Rewrite streaming plan to v3: multi-knob memory-aware auto

edeno
and
claude
committed
Extend streaming plan v3 to cover fit + estimate_parameters

edeno
and
claude
committed

Commits on Apr 23, 2026

Commits on Apr 24, 2026