From 0c098812a2347e728821a2a03416ac5d96f33b0a Mon Sep 17 00:00:00 2001 From: akudo7 Date: Thu, 24 Sep 2026 22:02:07 +0900 Subject: [PATCH] common : do not propagate n_seq_decision to speculative (MTP) context Decision sequences live on the main context only; propagating n_seq_max = n_parallel + n_seq_decision to the MTP draft context violates the n_outputs_max assert (n_outputs_max <= cparams.n_outputs_max). Verified: MTP + /v1/decision coexist after this fix (no assert crash). --- common/speculative.cpp | 1 + 1 file changed, 1 insertion(+) diff --git a/common/speculative.cpp b/common/speculative.cpp index 77dfe9535f3f..009ede3ad55c 100644 --- a/common/speculative.cpp +++ b/common/speculative.cpp @@ -2497,6 +2497,7 @@ common_params common_base_params_to_speculative(const common_params & params) { result.cache_type_k = params_spec.cache_type_k; result.cache_type_v = params_spec.cache_type_v; + result.n_seq_decision = 0; // decision sequences live on the main context only result.n_outputs_max = params.n_parallel; result.n_outputs_max_per_seq = 1;