vrr/prima.cpp
2
0
Fork 0
mirror of https://github.com/Lizonghang/prima.cpp.git synced 2025-09-09 10:24:38 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

disable prefetch in standalone mode

Browse source
This commit is contained in:
Lizonghang 2025-02-12 00:17:33 +04:00
parent 6a50d494d2
commit b163918b46
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
src/llama.cpp
View file

@ -18108,9 +18108,9 @@ static int llama_decode_internal(
timer(manage_graph_tensors); timer(manage_graph_tensors);
int next_gf_id = (i + 1) % gf.size(); int next_gf_id = (i + 1) % gf.size();
manage_graph_tensors(gf[next_gf_id], POSIX_MADV_WILLNEED, true); manage_graph_tensors(gf[next_gf_id], POSIX_MADV_WILLNEED, n_world > 1);
if (my_rank == 0 && (is_last_l || (next_gf_id == (int)gf.size() - 1))) { if (my_rank == 0 && (is_last_l || (next_gf_id == (int)gf.size() - 1))) {
manage_graph_tensors(gf[0], POSIX_MADV_WILLNEED, true); manage_graph_tensors(gf[0], POSIX_MADV_WILLNEED, n_world > 1);
} }
if (cparams.unload && n_world > 1) { if (cparams.unload && n_world > 1) {