From df6b5bb6e8077af7b62e93752115c4ba1573c3d0 Mon Sep 17 00:00:00 2001 From: Frank Date: Tue, 6 Oct 2026 04:18:53 -0700 Subject: [PATCH] MAX_SEQS 2->3: decode 33.06 t/s (+2.0% over depth-2 32.40; +4.2% over baseline). Bigger batch headroom co-schedules concurrent decode streams (decode is weight-streaming bound, so a fuller batch amortizes the weight read). KV pool 600k tokens (3x200k), above the 400k floor. Draft acceptance unchanged (0.755). Result: {"status":"keep","decode_tps":33.0578,"draft_accept":0.7549,"prefill_tps":452.9596,"ttft_ms":96315.4258} --- scripts/frank-primary.env | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/scripts/frank-primary.env b/scripts/frank-primary.env index 97dab1a..298e625 100644 --- a/scripts/frank-primary.env +++ b/scripts/frank-primary.env @@ -91,7 +91,7 @@ FRANK_PRIMARY_TP_WIRE=wht6 # start"); raise only after a real-traffic look at slot occupancy and the # KV pool's headroom line. The recurrent-state cost scales with it (54 # MiB/copy/card, 2 copies with speculation). -FRANK_PRIMARY_MAX_SEQS=2 +FRANK_PRIMARY_MAX_SEQS=3 # Served window. Native is 262,144; 200,000 is the publisher's production # value (compose) and what models.yml's contextWindow must quote — a longer # prompt is refused by the engine, never truncated.