dankclaw: production spec -> 3-layer chain draft-mtp,ngram-mod,ngram-map-k4v (user request; ~equal real-prose speed)

scripts/frank-default-large.env

@@ -12,23 +12,15 @@ PORT=7770
 # ~70GB headroom, so 2x262k quantized KV fits. Verify at startup.
 CTX_SIZE=524288
 PARALLEL=2
-# Spec: draft-mtp uses the MTP drafter BUNDLED inside the target GGUF.
+# Spec: 3-layer chain, chosen by the user 2026-08-14 (aesthetic/short-term
-# llama.cpp auto-detects it (b10411+).  DRAFT_FILE stays EMPTY.
+# experiment): draft-mtp first, then ngram-mod, then ngram-map-k4v as
-#
+# fallbacks.  On real prose this benches ~equal to draft-mtp alone
-# Winner = draft-mtp, n_max=5, p_min off.
+# (10.4 vs ~10 t/s @16k) - the ngram layers only pay off on self-similar text.
-# Realistic-prompt tuning (2026-08-14, temp 0.8, production settings):
+# Bundled MTP drafter; llama.cpp auto-detects it (b10411+).  DRAFT_FILE empty.
-#   n4 12.5 | n5 13.3 (PEAK) | n6 10.6 | n7 10.1 | n10 8.6 | baseline none 7.2.
+# n_max=5 is the realistic-text peak (n4 12.5, n6 10.6, n10 8.6).
-# The 54-config matrix (repetitive "fleet" prompt) reported n_max=10 ~30 t/s,
+# Occasional slow rounds possible when fallbacks misfire; revert to
-# but that prompt MTP drafts almost perfectly; long drafts collapse to ~18%
+# draft-mtp (alone) if the assistant reports stalls.
-# acceptance on real writing, so n_max=10 is actually ~8.6 t/s there.
+SPEC_TYPE=draft-mtp,ngram-mod,ngram-map-k4v
-# Layered draft-mtp,ngram-mod is faster on paper (matrix median 133 t/s) but
-# non-deterministic (10-148 t/s) and showed a degenerate instant-EOS -> the
-# same premature-stop class that retired DFlash. Rejected as production-unsafe.
-# p_min off: --spec-draft-p-min 0.8 was neutral to slightly worse for draft-mtp.
-# No premature-stop on realistic text at ubatch 256 (see batch notes below);
-# the earlier 'premature EOS' on the repetitive bench prompt was sampling
-# variance, not a corruption bug.
-SPEC_TYPE=draft-mtp
 DRAFT_FILE=
 SPEC_DRAFT_N_MAX=5
 SPEC_DRAFT_P_MIN=