nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 28 Packages Projects Releases Wiki Activity

server : fix crash when seq_rm fails for hybrid/recurrent models (#18391)

Browse Source 
* server : fix crash when seq_rm fails for hybrid/recurrent models

* server : add allow_processing param to clear_slot
This commit is contained in:
o7si
2025-12-26 16:35:29 +01:00
committed by GitHub
parent af3be131c0
commit 4893cc07bb
1 changed files with 5 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/server/server-context.cpp
+4 -2
View File
@@ -1007,8 +1007,10 @@ private:
return ret; return ret;
} }
void clear_slot(server_slot & slot) const { void clear_slot(server_slot & slot, bool allow_processing = false) const {
if (!allow_processing) {
GGML_ASSERT(!slot.is_processing()); GGML_ASSERT(!slot.is_processing());
}
SLT_WRN(slot, "clearing slot with %zu tokens\n", slot.prompt.tokens.size()); SLT_WRN(slot, "clearing slot with %zu tokens\n", slot.prompt.tokens.size());
@@ -2336,7 +2338,7 @@ private:
if (!llama_memory_seq_rm(llama_get_memory(ctx), slot.id, p0, -1)) { if (!llama_memory_seq_rm(llama_get_memory(ctx), slot.id, p0, -1)) {
SLT_WRN(slot, "failed to truncate tokens with position >= %d - clearing the memory\n", p0); SLT_WRN(slot, "failed to truncate tokens with position >= %d - clearing the memory\n", p0);
clear_slot(slot); clear_slot(slot, /*allow_processing=*/true);
// there is no common part left // there is no common part left
slot.n_prompt_tokens_cache = 0; slot.n_prompt_tokens_cache = 0;