From eb449cdfa4319d8fd9066e4633b49f4c867dd11c Mon Sep 17 00:00:00 2001 From: Georgi Gerganov Date: Sun, 8 Feb 2026 09:40:04 +0200 Subject: [PATCH] server : improve context checkpoint logic (#19408) --- tools/server/server-context.cpp | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/tools/server/server-context.cpp b/tools/server/server-context.cpp index b71d496eeb..8ec8451339 100644 --- a/tools/server/server-context.cpp +++ b/tools/server/server-context.cpp @@ -2507,7 +2507,8 @@ private: slot.n_prompt_tokens_processed++; // process the last few tokens of the prompt separately in order to allow for a checkpoint to be created. - if (do_checkpoint && slot.task->n_tokens() - slot.prompt.n_tokens() == 64) { + const int n_last = std::min(n_batch, 512); + if (do_checkpoint && slot.task->n_tokens() == slot.prompt.n_tokens() + n_last) { break; } }