XTuner 微调 LLM:1.8B、多模态

一、多模态LLM原理简介

1.1. 给LLM装上电子眼:多模态LLM原理简介

1.1.1. 文本单模态

1.1.2. 文本+图像多模态

1.2. 什么型号的电子眼:LLaVA方案简介

        Haotian Liu等使用GPT-4V对图像数据生成描述,以此构建出大量<question text><image> -- <answer text>的数据对。利用这些数据对,配合文本单模态LLM,训练出一个Image Projector。

        所使用的文本单模型LLM和训练出来的Image Projector,统称为LLaVA模型

1.2.1. LLaVA训练阶段示意图

1.2.2. LLaVA测试阶段示意图

Image Projector的训练和测试,有点类似之前我们讲过的LoRA微调方案。

        二者都是在已有LLM的基础上,用新的数据训练一个新的小文件。

        只不过,LLM套上LoRA之后,有了新的灵魂(角色);而LLM套上Image Projector之后,才有了眼睛。

二、使用xtuner微调多模态LLM过程

2.1xtuner安装和配置

# 如果你是在 InternStudio 平台,则从本地 clone 一个已有 pytorch 的环境:
# pytorch    2.0.1   py3.10_cuda11.7_cudnn8.5.0_0

cd ~ && studio-conda xtuner0.1.17
# 如果你是在其他平台:
# conda create --name xtuner0.1.17 python=3.10 -y

# 激活环境
conda activate xtuner0.1.17
# 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 -p /root/xtuner0117 && cd /root/xtuner0117

# 拉取 0.1.17 的版本源码
git clone -b v0.1.17  https://github.com/InternLM/xtuner
# 无法访问github的用户请从 gitee 拉取:
# git clone -b v0.1.15 https://gitee.com/Internlm/xtuner

# 进入源码目录
cd /root/xtuner0117/xtuner

# 从源码安装 XTuner
pip install -e '.[all]' && cd ~

2.2概述

在本节中,我们将 自己构造 <question text><image>--<answer text> 数据对,基于InternLM2_Chat_1.8B这个文本单模态模型,使用LLaVA方案,训练一个给InternLM2_Chat_1.8B使用的Image Projector文件。

        LLaVA方案中,给LLM增加视觉能力的过程,即是训练Image Projector文件的过程。 该过程分为2个阶段:Pretrain和Finetune。

2.3Pretrain阶段

        在Pretrain阶段,我们会使用大量的图片+简单文本(caption, 即图片标题)数据对,使LLM理解图像中的普遍特征。即,对大量的图片进行粗看

        Pretrain阶段训练完成后,此时的模型已经有视觉能力了!但是由于训练数据中都是图片+图片标题,所以此时的模型虽然有视觉能力,但无论用户问它什么,它都只会回答输入图片的标题。即,此时的模型只会给输入图像“写标题”

Pretrain阶段相当于是开发LLM时预训练工作,对硬件要求非常高,有8卡的学有余力同学可以自行尝试。详见XTuner-LLaVALLaVA

        在本次实战营中,我们已经为大家提供了Pretrain阶段的产物——iter_2181.pth文件。它就是幼稚园阶段的Image Projector!大家带着iter_2181.pth文件继续进入下一阶段进行Finetune即可。

2.4Finetune阶段

        在Finetune阶段,我们会使用图片+复杂文本数据对,来对Pretrain得到的Image Projector即iter_2181.pth进行进一步的训练。

2.4.1训练数据构建

数据集格式

[
    {
        "id": "随便什么字符串",
        "image": "图片文件的相对位置。相对谁?相对你后面config文件里指定的image_folder参数的路径。",
        "conversation": [
            {
                "from": "human",
                "value": "<image>\n第1个问题。"
            },
            {
                "from": "gpt",
                "value": "第1个回答"
            },
            {
                "from": "human",
                "value": "第2个问题。"
            },
            {
                "from": "gpt",
                "value": "第2个回答"
            },
            # ......
            {
                "from": "human",
                "value": "第n个问题。"
            },
            {
                "from": "gpt",
                "value": "第n个回答"
            },
        ]
    },

    # 下面是第2组训练数据了。

    {
        "id": "随便什么字符串",
        "image": "图片文件的相对位置。相对谁?相对你后面config文件里指定的image_folder参数的路径。",
        "conversation": [
            {
                "from": "human",
                "value": "<image>\n第1个问题。"
            },
            # ......
            {
                "from": "gpt",
                "value": "第n个回答"
            }
        ]
    }
]

注意:每组训练数据的第1个来自human的问题前,要加上图片占位符,即<image>

示例如下:

[
  {
    "id": "000000033471",
    "image": "coco/train2017/000000033471.jpg",
    "conversations": [
      {
        "from": "human",
        "value": "<image>\nWhat are the colors of the bus in the image?"
      },
      {
        "from": "gpt",
        "value": "The bus in the image is white and red."
      },
      {
        "from": "human",
        "value": "What feature can be seen on the back of the bus?"
      },
      {
        "from": "gpt",
        "value": "The back of the bus features an advertisement."
      },
      {
        "from": "human",
        "value": "Is the bus driving down the street or pulled off to the side?"
      },
      {
        "from": "gpt",
        "value": "The bus is driving down the street, which is crowded with people and other vehicles."
      }
    ]
  },
  {
    "id": "000000052846",
    "image": "coco/train2017/000000052846.jpg",
    "conversations": [
      {
        "from": "human",
        "value": "<image>\nWhere is the cat positioned in the image?"
      },
      {
        "from": "gpt",
        "value": "The cat is positioned on top of the back of the couch in the living room."
      }
    ]
  }
 ]

2.4.1数据集制作

        我们可以效法LLaVA作者的做法,将自己的图片发送给GPT,要求其按照上述格式生成若干条问答对。

prompts

oph.jpg

Create a dataset for me, following this format.(按照以下格式为我创建一个数据集。)

[
  {
    "id": "<random_number_string>",
    "image": "test_img/oph.jpg",
    "conversations": [
      {
        "from": "human",
        "value": "<image>\nDescribe this image."
      },
      {
        "from": "gpt",
        "value": "<answer1>"
      },
      {
        "from": "human",
        "value": "<question2>"
      },
      {
        "from": "gpt",
        "value": "<answer2>"
      },
      {
        "from": "human",
        "value": "<question3>"
      },
      {
        "from": "gpt",
        "value": "<answer3>"
      }
    ]
  }
]

        The questions and answers, please generate for me, based on the image I sent to you. Thes questions should be from the shallow to the deep, and the answers should be as detailed and correct as possible. The questions and answers should be stick to the contents in the image itself, like objects, peoples, equipment, environment, purpose, color, attitude, etc. 5 question and answer pairs.(请根据我发给你的图片为我生成问题和答案。问题应由浅入深,答案应尽可能详细和正确。问题和答案应紧扣图片本身的内容,如物体、人物、设备、环境、目的、颜色、态度等。5 对问答题。)

        为了方便大家跟随课程,针对这张示例图片的问答对数据(repeat_data.json),大家按照下面的脚本运行就可以生成啦~(重复200次)

cd ~ && git clone https://github.com/InternLM/tutorial -b camp2 && conda activate xtuner0.1.17 && cd tutorial

python /root/tutorial/xtuner/llava/llava_data/repeat.py \
  -i /root/tutorial/xtuner/llava/llava_data/unique_data.json \
  -o /root/tutorial/xtuner/llava/llava_data/repeated_data.json \
  -n 200

2.5创建配置文件

# 查询xtuner内置配置文件
xtuner list-cfg -p llava_internlm2_chat_1_8b

# 拷贝配置文件到当前目录
xtuner copy-cfg \
  llava_internlm2_chat_1_8b_qlora_clip_vit_large_p14_336_lora_e1_gpu8_finetune \
  /root/tutorial/xtuner/llava

        当前你的/root/tutorial/xtuner/llava/目录下的文件结构应该是这样:

|-- llava_data
|   |-- repeat.py
|   |-- repeated_data.json
|   |-- test_img
|   |   `-- oph.jpg
|   `-- unique_data.json
`-- llava_internlm2_chat_1_8b_qlora_clip_vit_large_p14_336_lora_e1_gpu8_finetune_copy.py

2.6修改配置文件

修改llava_internlm2_chat_1_8b_qlora_clip_vit_large_p14_336_lora_e1_gpu8_finetune_copy.py文件中的:

  • pretrained_pth
  • llm_name_or_path
  • visual_encoder_name_or_path
  • data_root
  • data_path
  • image_folder
# Model
- llm_name_or_path = 'internlm/internlm2-chat-1_8b'
+ llm_name_or_path = '/root/share/new_models/Shanghai_AI_Laboratory/internlm2-chat-1_8b'
- visual_encoder_name_or_path = 'openai/clip-vit-large-patch14-336'
+ visual_encoder_name_or_path = '/root/share/new_models/openai/clip-vit-large-patch14-336'

# Specify the pretrained pth
- pretrained_pth = './work_dirs/llava_internlm2_chat_1_8b_clip_vit_large_p14_336_e1_gpu8_pretrain/iter_2181.pth'  # noqa: E501
+ pretrained_pth = '/root/share/new_models/xtuner/iter_2181.pth'

# Data
- data_root = './data/llava_data/'
+ data_root = '/root/tutorial/xtuner/llava/llava_data/'
- data_path = data_root + 'LLaVA-Instruct-150K/llava_v1_5_mix665k.json'
+ data_path = data_root + 'repeated_data.json'
- image_folder = data_root + 'llava_images'
+ image_folder = data_root

# Scheduler & Optimizer
- batch_size = 16  # per_device
+ batch_size = 1  # per_device


# evaluation_inputs
- evaluation_inputs = ['请描述一下这张图片','Please describe this picture']
+ evaluation_inputs = ['Please describe this picture','What is the equipment in the image?']

2.7开始Finetune

cd /root/tutorial/xtuner/llava/
xtuner train /root/tutorial/xtuner/llava/llava_internlm2_chat_1_8b_qlora_clip_vit_large_p14_336_lora_e1_gpu8_finetune_copy.py --deepspeed deepspeed_zero2

2.8对比Finetune前后的性能差异

2.8.1微调前

即:加载 1.8B 和 Pretrain阶段产物(iter_2181) 到显存。

# 解决小bug
export MKL_SERVICE_FORCE_INTEL=1
export MKL_THREADING_LAYER=GNU

# pth转huggingface
xtuner convert pth_to_hf \
  llava_internlm2_chat_1_8b_clip_vit_large_p14_336_e1_gpu8_pretrain \
  /root/share/new_models/xtuner/iter_2181.pth \
  /root/tutorial/xtuner/llava/llava_data/iter_2181_hf

# 启动!
xtuner chat /root/share/new_models/Shanghai_AI_Laboratory/internlm2-chat-1_8b \
  --visual-encoder /root/share/new_models/openai/clip-vit-large-patch14-336 \
  --llava /root/tutorial/xtuner/llava/llava_data/iter_2181_hf \
  --prompt-template internlm2_chat \
  --image /root/tutorial/xtuner/llava/llava_data/test_img/oph.jpg
  

Q1: Describe this image.
Q2: What is the equipment in the image?

2.8.2微调后

即:加载 1.8B 和 Fintune阶段产物 到显存。

# 解决小bug
export MKL_SERVICE_FORCE_INTEL=1
export MKL_THREADING_LAYER=GNU

# pth转huggingface
xtuner convert pth_to_hf \
  /root/tutorial/xtuner/llava/llava_internlm2_chat_1_8b_qlora_clip_vit_large_p14_336_lora_e1_gpu8_finetune_copy.py \
  /root/tutorial/xtuner/llava/work_dirs/llava_internlm2_chat_1_8b_qlora_clip_vit_large_p14_336_lora_e1_gpu8_finetune_copy/iter_1200.pth \
  /root/tutorial/xtuner/llava/llava_data/iter_1200_hf

# 启动!
xtuner chat /root/share/new_models/Shanghai_AI_Laboratory/internlm2-chat-1_8b \
  --visual-encoder /root/share/new_models/openai/clip-vit-large-patch14-336 \
  --llava /root/tutorial/xtuner/llava/llava_data/iter_1200_hf \
  --prompt-template internlm2_chat \
  --image /root/tutorial/xtuner/llava/llava_data/test_img/oph.jpg

Q1: Describe this image.
Q2: What is the equipment in the image?

Finetune前后效果对比:

Finetune前:只会打标题 

ft_before

Finetune后:会回答问题了 

ft_after

三、本地Finetune后效果展示

评论 1
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值