sched : offload_op also requires supports_op (#7977)

diff --git a/ggml-backend.c b/ggml-backend.c
index 2bec7bea38a85e10059551098a4779596c1fe0bb..26dce7f72421396b433f5aca4c078dc1d5f1ac7c 100644 (file)
--- a/ggml-backend.c
+++ b/ggml-backend.c
@@ -1172,7 +1172,7 @@ static int ggml_backend_sched_backend_id_from_cur(ggml_backend_sched_t sched, st
             // check if a backend with higher prio wants to offload the op
             if (src_backend_id == sched->n_backends - 1) {
                 for (int b = 0; b < src_backend_id; b++) {
-                    if (ggml_backend_offload_op(sched->backends[b], tensor)) {
+                    if (ggml_backend_supports_op(sched->backends[b], tensor) && ggml_backend_offload_op(sched->backends[b], tensor)) {
                         SET_CAUSE(tensor, "1.off");
                         return b;
                     }