Skip to content

Commit 1a94186

Browse files
committed
metal : disable graph concurrency optimization due to bug (#2413)
1 parent b5472ea commit 1a94186

File tree

1 file changed

+4
-3
lines changed

1 file changed

+4
-3
lines changed

llama.cpp

Lines changed: 4 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1722,9 +1722,10 @@ static bool llama_eval_internal(
17221722

17231723
#ifdef GGML_USE_METAL
17241724
if (lctx.ctx_metal && N == 1) {
1725-
if (!ggml_metal_if_optimized(lctx.ctx_metal)) {
1726-
ggml_metal_graph_find_concurrency(lctx.ctx_metal, gf);
1727-
}
1725+
// TODO: disabled until #2413 is resolved
1726+
//if (!ggml_metal_if_optimized(lctx.ctx_metal)) {
1727+
// ggml_metal_graph_find_concurrency(lctx.ctx_metal, gf);
1728+
//}
17281729
ggml_metal_set_n_cb (lctx.ctx_metal, n_threads);
17291730
ggml_metal_graph_compute(lctx.ctx_metal, gf);
17301731
ggml_metal_get_tensor (lctx.ctx_metal, cur);

0 commit comments

Comments
 (0)