From 8669c3db2b98f05775292778dd05f424ee0cd250 Mon Sep 17 00:00:00 2001
From: Kawrakow <iwankawrakow@gmail.com>
Date: Mon, 12 May 2025 07:47:46 +0300
Subject: GPU offload policy (#405)

* Adding GPU offload policy

* Minor

---------

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
---
 common/common.h | 1 +
 1 file changed, 1 insertion(+)

(limited to 'common/common.h')

diff --git a/common/common.h b/common/common.h
index b4f75236..fd83c9d3 100644
--- a/common/common.h
+++ b/common/common.h
@@ -143,6 +143,7 @@ struct gpt_params {
     std::vector<std::string> antiprompt; // strings upon which more user input is prompted (a.k.a. reverse prompts)
     std::vector<llama_model_kv_override> kv_overrides;
     std::vector<llama_model_tensor_buft_override> tensor_buft_overrides;
+    std::vector<std::pair<int,int>> offload_policy;
 
     bool lora_init_without_apply = false; // only load lora to memory, but do not apply it to ctx (user can manually apply lora later using llama_lora_adapter_apply)
     std::vector<llama_lora_adapter_info> lora_adapters; // lora adapter path with user defined scale
-- 
cgit v1.2.3