diff --git a/src/llama.cpp b/src/llama.cpp index ca943eea..53dba75a 100644 --- a/src/llama.cpp +++ b/src/llama.cpp @@ -4771,7 +4771,7 @@ struct llama_context * llama_new_context_with_model( LLAMA_LOG_INFO("XXXXXXXXXXXXXXXXXXXXX Setting only active experts offload\n"); ggml_backend_sched_set_only_active_experts(ctx->sched, true); } - if (model->split_mode == LLAMA_SPLIT_MODE_GRAPH) { + if (model->split_mode == LLAMA_SPLIT_MODE_GRAPH && !model->has_tensor_overrides()) { ggml_backend_sched_set_split_mode_graph(ctx->sched, true); }