vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

llama : fix missing worst case flag during reserve

Browse source
This commit is contained in:
slaren 2024-10-31 17:49:15 +01:00
parent dec6ce2535
commit b135927ca4
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
src/llama.cpp
View file

@ -19694,7 +19694,7 @@ struct llama_context * llama_new_context_with_model(
int n_nodes_tg = ggml_graph_n_nodes(gf_tg); int n_nodes_tg = ggml_graph_n_nodes(gf_tg);
// reserve again with pp graph to avoid ggml-alloc reallocations during inference // reserve again with pp graph to avoid ggml-alloc reallocations during inference
gf_pp = llama_build_graph(*ctx, ubatch_pp, false); gf_pp = llama_build_graph(*ctx, ubatch_pp, true);
if (!ggml_backend_sched_reserve(ctx->sched, gf_pp)) { if (!ggml_backend_sched_reserve(ctx->sched, gf_pp)) {
LLAMA_LOG_ERROR("%s: failed to allocate compute buffers\n", __func__); LLAMA_LOG_ERROR("%s: failed to allocate compute buffers\n", __func__);
llama_free(ctx); llama_free(ctx);