From 21ab727e83c550fdb777f386b417bbcb54f59da1 Mon Sep 17 00:00:00 2001 From: Concedo <39025047+LostRuins@users.noreply.github.com> Date: Tue, 30 Jan 2024 22:30:08 +0800 Subject: [PATCH] change split mode to rows --- gpttype_adapter.cpp | 1 + 1 file changed, 1 insertion(+) diff --git a/gpttype_adapter.cpp b/gpttype_adapter.cpp index 59819c57c..5c716908d 100644 --- a/gpttype_adapter.cpp +++ b/gpttype_adapter.cpp @@ -955,6 +955,7 @@ ModelLoadResult gpttype_load_model(const load_model_inputs inputs, FileFormat in } #endif model_params.main_gpu = cu_parseinfo_maindevice; + model_params.split_mode = llama_split_mode::LLAMA_SPLIT_ROW; llama_ctx_params.n_batch = blasbatchsize; llama_ctx_params.n_threads = n_threads;