一个大语言模型微调工具箱,由上海人工智能实验室开发。支持的开源LLM (2023.11.01),详情见哔哩哔哩公众号
在算力平台上进行搭建(InternStudio,该平台是活动期间免费提供使用)
# 如果你是在 InternStudio 平台,则从本地 clone 一个已有 pytorch 2.0.1 的环境:
/root/share/install_conda_env_internlm_base.sh xtuner0.1.9
# 如果你是在其他平台:
conda create --name xtuner0.1.9 python=3.10 -y
# 激活环境
conda activate xtuner0.1.9
# 进入家目录 (~的意思是 “当前用户的home路径”)
cd ~
# 创建版本文件夹并进入,以跟随本教程
mkdir xtuner019 && cd xtuner019
# 拉取 0.1.9 的版本源码
git clone -b v0.1.9 https://github.com/InternLM/xtuner
# 无法访问github的用户请从 gitee 拉取:
# git clone -b v0.1.9 https://gitee.com/Internlm/xtuner
# 进入源码目录
cd xtuner
# 从源码安装 XTuner
pip install -e '.[all]'
安装完成后, 创建一个微调 oasst1 数据集的工作路径,进入
mkdir ~/ft-oasst1 && cd ~/ft-oasst1
XTuner 提供多个开箱即用的配置文件,用户可以通过下列命令查看:
假如显示bash: xtuner: command not found的话可以考虑在终端输入
export PATH=$PATH:'/root/.local/bin'
显示的配置文件
拷贝一个配置文件到当前目录:?# xtuner copy-cfg ${CONFIG_NAME} ${SAVE_PATH}
在本案例中即:(注意最后有个英文句号,代表复制到当前路径)
xtuner copy-cfg internlm_chat_7b_qlora_oasst1_e3 .
由于下载模型很慢,在教学平台可以直接复制模型
ln -s /share/temp/model_repos/internlm-chat-7b ~/ft-oasst1/
由于 huggingface 网络问题,平台已经复制好,复制到正确位置即可:
cd ~/ft-oasst1
# ...-guanaco 后面有个空格和英文句号啊
cp -r /root/share/temp/datasets/openassistant-guanaco .
修改其中的模型和数据集为 本地路径
vim internlm_chat_7b_qlora_oasst1_e3_copy.py
减号代表要删除的行,加号代表要增加的行
# 修改模型为本地路径
- pretrained_model_name_or_path = 'internlm/internlm-chat-7b'
+ pretrained_model_name_or_path = './internlm-chat-7b'
# 修改训练数据集为本地路径
- data_path = 'timdettmers/openassistant-guanaco'
+ data_path = './openassistant-guanaco'
训练:
xtuner train ${CONFIG_NAME_OR_PATH}
也可以增加 deepspeed 进行训练加速:
xtuner train ${CONFIG_NAME_OR_PATH} --deepspeed deepspeed_zero2
例如,我们可以利用 QLoRA 算法在 oasst1 数据集上微调 InternLM-7B:
# 单卡
## 用刚才改好的config文件训练
xtuner train ./internlm_chat_7b_qlora_oasst1_e3_copy.py
# 多卡
NPROC_PER_NODE=${GPU_NUM} xtuner train ./internlm_chat_7b_qlora_oasst1_e3_copy.py
# 若要开启 deepspeed 加速,增加 --deepspeed deepspeed_zero2 即可
微调得到的 PTH 模型文件和其他杂七杂八的文件都默认在当前的?./work_dirs
?中,跑完训练后,当前路径是这样:
xtuner convert pth_to_hf ${CONFIG_NAME_OR_PATH} ${PTH_file_dir} ${SAVE_PATH}
在本示例中,为:
mkdir hf
export MKL_SERVICE_FORCE_INTEL=1
xtuner convert pth_to_hf ./internlm_chat_7b_qlora_oasst1_e3_copy.py ./work_dirs/internlm_chat_7b_qlora_oasst1_e3_copy/epoch_1.pth ./hf
此时,hf 文件夹即为我们平时所理解的所谓 “LoRA 模型文件”
2.4 部署与测试
基本语法
xtuner convert merge ./internlm-chat-7b ./hf ./merged --max-shard-size 2GB
# xtuner convert merge \
# ${NAME_OR_PATH_TO_LLM} \
# ${NAME_OR_PATH_TO_ADAPTER} \
# ${SAVE_PATH} \
# --max-shard-size 2GB
# 加载 Adapter 模型对话(Float 16)
xtuner chat ./merged --prompt-template internlm_chat
# 4 bit 量化加载
# xtuner chat ./merged --bits 4 --prompt-template internlm_chat
2.4.3 Demo
在/ft-oasst1/目录下建立cli_demo.py文件,代码如下:
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
model_name_or_path = "/root/model/Shanghai_AI_Laboratory/internlm-chat-7b"
tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(model_name_or_path, trust_remote_code=True, torch_dtype=torch.bfloat16, device_map='auto')
model = model.eval()
system_prompt = """You are an AI assistant whose name is InternLM (书生·浦语).
- InternLM (书生·浦语) is a conversational language model that is developed by Shanghai AI Laboratory (上海人工智能实验室). It is designed to be helpful, honest, and harmless.
- InternLM (书生·浦语) can understand and communicate fluently in the language chosen by the user such as English and 中文.
"""
messages = [(system_prompt, '')]
print("=============Welcome to InternLM chatbot, type 'exit' to exit.=============")
while True:
input_text = input("User >>> ")
input_text.replace(' ', '')
if input_text == "exit":
break
response, history = model.chat(tokenizer, input_text, history=messages)
messages.append((input_text, response))
print(f"robot >>> {response}")
修改代码
将model_name_or_path = "/root/model/Shanghai_AI_Laboratory/internlm-chat-7b"
修改为model_name_or_path = "merged"
???????