nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 25 Packages Projects Releases Wiki Activity

server : fix ctx checkpoint invalidation (#20671)

Browse Source
This commit is contained in:
Georgi Gerganov
2026-03-17 15:21:14 +02:00
committed by GitHub
parent 627670601a
commit 8cc2d81264
1 changed files with 3 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/server/server-context.cpp
+3 -3
View File
@@ -2402,11 +2402,11 @@ private:
} }
{ {
// erase any checkpoints with pos_min > pos_min_thold // erase any checkpoints with pos_max > pos_next
for (auto it = slot.prompt.checkpoints.begin(); it != slot.prompt.checkpoints.end();) { for (auto it = slot.prompt.checkpoints.begin(); it != slot.prompt.checkpoints.end();) {
const auto & cur = *it; const auto & cur = *it;
if (cur.pos_min > pos_min_thold) { if (cur.pos_max > pos_next) {
SLT_WRN(slot, "erased invalidated context checkpoint (pos_min = %d, pos_max = %d, n_tokens = %" PRId64 ", n_swa = %d, size = %.3f MiB)\n", cur.pos_min, cur.pos_max, cur.n_tokens, n_swa, (float) cur.data.size() / 1024 / 1024); SLT_WRN(slot, "erased invalidated context checkpoint (pos_min = %d, pos_max = %d, n_tokens = %" PRId64 ", n_swa = %d, pos_next = %d, size = %.3f MiB)\n", cur.pos_min, cur.pos_max, cur.n_tokens, n_swa, pos_next, (float) cur.data.size() / 1024 / 1024);
it = slot.prompt.checkpoints.erase(it); it = slot.prompt.checkpoints.erase(it);
} else { } else {
++it; ++it;