tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-06-29 04:35:05 +00:00
Code Issues Packages Projects Releases Wiki Activity

kv-cache : fix unified::seq_rm to work with seq_id < 0 (#13985)

Browse Source 
ggml-ci
This commit is contained in:
Georgi Gerganov
2025-06-04 09:50:32 +03:00
committed by GitHub
parent 7e00e60ef8
commit e0e806f52e
1 changed files with 20 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
src/llama-kv-cache-unified.cpp
View File

@ -149,6 +149,7 @@ bool llama_kv_cache_unified::seq_rm(llama_seq_id seq_id, llama_pos p0, llama_pos
p1 = std::numeric_limits<llama_pos>::max();
}
if (seq_id >= 0) {
for (uint32_t i = 0; i < cells.size(); ++i) {
if (!cells.pos_in(i, p0, p1)) {
continue;
@ -160,6 +161,20 @@ bool llama_kv_cache_unified::seq_rm(llama_seq_id seq_id, llama_pos p0, llama_pos
}
}
}
} else {
// match any sequence
for (uint32_t i = 0; i < cells.size(); ++i) {
if (!cells.pos_in(i, p0, p1)) {
continue;
}
cells.rm(i);
if (new_head == cells.size()) {
new_head = i;
}
}
}
// If we freed up a slot, set head to it so searching can start there.
if (new_head != cells.size() && new_head < head) {