From c431bee94184beb619530039e8dd508168cfaf25 Mon Sep 17 00:00:00 2001
From: hzg0601 <hzg0601@163.com>
Date: Thu, 27 Jul 2023 14:27:11 +0800
Subject: [PATCH] =?UTF-8?q?=E5=88=A0=E9=99=A4requirements=E7=9A=84?=
 =?UTF-8?q?=E4=B8=AD=E6=96=87?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 models/loader/loader.py | 7 +++++++
 requirements.txt        | 1 -
 2 files changed, 7 insertions(+), 1 deletion(-)

diff --git a/models/loader/loader.py b/models/loader/loader.py
index f43bb1d..c5c80fd 100644
--- a/models/loader/loader.py
+++ b/models/loader/loader.py
@@ -177,6 +177,13 @@ class LoaderCheckPoint:
                 )
 
         elif self.is_llamacpp:
+            
+            # 要调用llama-cpp模型，如vicuma-13b量化模型需要安装llama-cpp-python库
+            # but!!! 实测pip install 不好使，需要手动从ttps://github.com/abetlen/llama-cpp-python/releases/下载
+            # 而且注意不同时期的ggml格式并不！兼！容!!!因此需要安装的llama-cpp-python版本也不一致，需要手动测试才能确定
+            # 实测ggml-vicuna-13b-1.1在llama-cpp-python 0.1.63上可正常兼容
+            # 不过！！！本项目模型加载的方式控制的比较严格，与llama-cpp-python的兼容性较差，很多参数设定不能使用，
+            # 建议如非必要还是不要使用llama-cpp
 
             try:
                 from llama_cpp import Llama
diff --git a/requirements.txt b/requirements.txt
index 84adf2a..4c981b2 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -23,7 +23,6 @@ openai
 #accelerate~=0.18.0
 #peft~=0.3.0
 #bitsandbytes; platform_system != "Windows"
-
 torch~=2.0.0
 pydantic~=1.10.7
 starlette~=0.26.1