L2- InternVL 多模态模型部署微调实践

一、什么是InternVL

InternVL 是一种用于多模态任务的深度学习模型，旨在处理和理解多种类型的数据输入，如图像和文本。它结合了视觉和语言模型，能够执行复杂的跨模态任务，比如图文匹配、图像描述生成等。通过整合视觉特征和语言信息，InternVL 可以在多模态领域取得更好的表现。

二、InternVL 模型总览

1.Dynamic High Resolution

InternVL独特的预处理模块：动态高分辨率，是为了让ViT模型能够尽可能获取到更细节的图像信息，提高视觉特征的表达能力。对于输入的图片，首先resize成448的倍数，然后按照预定义的尺寸比例从图片上crop对应的区域。

2.Pixel Shuffle 

Pixel Shuffle在超分任务中是一个常见的操作，PyTorch中有官方实现，即nn.PixelShuffle(upscale_factor) 该类的作用就是将一个tensor中的元素值进行重排列，假设tensor维度为[B, C, H, W], PixelShuffle操作不仅可以改变tensor的通道数，也会改变特征图的大小。

三、InternVL 部署微调实践

1.准备InternVL模型

cd /root
mkdir -p model

# cp 模型

cp -r /root/share/new_models/OpenGVLab/InternVL2-2B /root/model/

  

2. 准备环境

①配置虚拟环境

conda create --name xtuner python=3.10 -y

# 激活虚拟环境（注意：后续的所有操作都需要在这个虚拟环境中进行）
conda activate xtuner

# 安装一些必要的库
conda install pytorch==2.1.2 torchvision==0.16.2 torchaudio==2.1.2 pytorch-cuda=12.1 -c pytorch -c nvidia -y
# 安装其他依赖
apt install libaio-dev
pip install transformers==4.39.3
pip install streamlit==1.36.0

 

 

 

②安装xtuner

# 创建一个目录，用来存放源代码
mkdir -p /root/InternLM/code

cd /root/InternLM/code

git clone -b v0.1.23  https://github.com/InternLM/XTuner

 

③进入XTuner目录

cd /root/InternLM/code/XTuner
pip install -e '.[deepspeed]'

  

④安装LMDeploy

pip install lmdeploy==0.5.3

 

⑤安装验证

xtuner version

##命令

xtuner help

 

3.准备微调数据集

①使用huggingface上的zhongshsh/CLoT-Oogiri-GO据集

@misc{zhong2023clot,
  title={Let's Think Outside the Box: Exploring Leap-of-Thought in Large Language Models with Creative Humor Generation},
  author={Zhong, Shanshan and Huang, Zhongzhan and Gao, Shanghua and Wen, Weushao and Lin, Liang and Zitnik, Marinka and Zhou, Pan},
  journal={arXiv preprint arXiv:2312.02439},
  year={2023}
}

 

②从share里挪出数据集 

## 首先让我们安装一下需要的包
pip install datasets matplotlib Pillow timm

## 让我们把数据集挪出来
cp -r /root/share/new_models/datasets/CLoT_cn_2000 /root/InternLM/datasets/

 

 

③选择jsonl里的第一条数据对应的图片，把这张图片挪动到InternLM文件夹下面 

cp InternLM/datasets/CLoT_cn_2000/ex_images/007aPnLRgy1hb39z0im50j30ci0el0wm.jpg InternLM/ 

 

四、InternVL 推理部署攻略  

1.使用pipeline进行推理

①新建一个文件

touch /root/InternLM/code/test_lmdeploy.py
cd /root/InternLM/code/

 

②打开文件，贴入代码

from lmdeploy import pipeline
from lmdeploy.vl import load_image

pipe = pipeline('/root/model/InternVL2-2B')

image = load_image('/root/InternLM/007aPnLRgy1hb39z0im50j30ci0el0wm.jpg')
response = pipe(('请你根据这张图片，讲一个脑洞大开的梗', image))
print(response.text)

 

③运行执行推理结果

python3 test_lmdeploy.py

 

2.推理后（推理结果随机） 

五、InternVL 微调攻略 

1.准备数据集

数据集是复制进InternLM/datasets的数据

# 为了高效训练，请确保数据格式为：
{
    "id": "000000033471",
    "image": ["coco/train2017/000000033471.jpg"], # 如果是纯文本，则该字段为 None 或者不存在
    "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."
      }
    ]
  }

2.配置微调参数

修改XTuner下 InternVL的config，

文件在： /root/InternLM/code/XTuner/xtuner/configs/internvl/v2/internvl_v2_internlm2_2b_qlora_finetune.py

①需要修改的部分

②总体config文件

# Copyright (c) OpenMMLab. All rights reserved.
from mmengine.hooks import (CheckpointHook, DistSamplerSeedHook, IterTimerHook,
                            LoggerHook, ParamSchedulerHook)
from mmengine.optim import AmpOptimWrapper, CosineAnnealingLR, LinearLR
from peft import LoraConfig
from torch.optim import AdamW
from transformers import AutoTokenizer

from xtuner.dataset import InternVL_V1_5_Dataset
from xtuner.dataset.collate_fns import default_collate_fn
from xtuner.dataset.samplers import LengthGroupedSampler
from xtuner.engine.hooks import DatasetInfoHook
from xtuner.engine.runner import TrainLoop
from xtuner.model import InternVL_V1_5
from xtuner.utils import PROMPT_TEMPLATE

#######################################################################
#                          PART 1  Settings                           #
#######################################################################
# Model
path = '/root/model/InternVL2-2B'

# Data
data_root = '/root/InternLM/datasets/CLoT_cn_2000/'
data_path = data_root + 'ex_cn.json'
image_folder = data_root
prompt_template = PROMPT_TEMPLATE.internlm2_chat
max_length = 6656

# Scheduler & Optimizer
batch_size = 4  # per_device
accumulative_counts = 4
dataloader_num_workers = 4
max_epochs = 6
optim_type = AdamW
# official 1024 -> 4e-5
lr = 2e-5
betas = (0.9, 0.999)
weight_decay = 0.05
max_norm = 1  # grad clip
warmup_ratio = 0.03

# Save
save_steps = 1000
save_total_limit = 1  # Maximum checkpoints to keep (-1 means unlimited)

#######################################################################
#            PART 2  Model & Tokenizer & Image Processor              #
#######################################################################
model = dict(
    type=InternVL_V1_5,
    model_path=path,
    freeze_llm=True,
    freeze_visual_encoder=True,
    quantization_llm=True,  # or False
    quantization_vit=False,  # or True and uncomment visual_encoder_lora
    # comment the following lines if you don't want to use Lora in llm
    llm_lora=dict(
        type=LoraConfig,
        r=128,
        lora_alpha=256,
        lora_dropout=0.05,
        target_modules=None,
        task_type='CAUSAL_LM'),
    # uncomment the following lines if you don't want to use Lora in visual encoder # noqa
    # visual_encoder_lora=dict(
    #     type=LoraConfig, r=64, lora_alpha=16, lora_dropout=0.05,
    #     target_modules=['attn.qkv', 'attn.proj', 'mlp.fc1', 'mlp.fc2'])
)

#######################################################################
#                      PART 3  Dataset & Dataloader                   #
#######################################################################
llava_dataset = dict(
    type=InternVL_V1_5_Dataset,
    model_path=path,
    data_paths=data_path,
    image_folders=image_folder,
    template=prompt_template,
    max_length=max_length)

train_dataloader = dict(
    batch_size=batch_size,
    num_workers=dataloader_num_workers,
    dataset=llava_dataset,
    sampler=dict(
        type=LengthGroupedSampler,
        length_property='modality_length',
        per_device_batch_size=batch_size * accumulative_counts),
    collate_fn=dict(type=default_collate_fn))

#######################################################################
#                    PART 4  Scheduler & Optimizer                    #
#######################################################################
# optimizer
optim_wrapper = dict(
    type=AmpOptimWrapper,
    optimizer=dict(
        type=optim_type, lr=lr, betas=betas, weight_decay=weight_decay),
    clip_grad=dict(max_norm=max_norm, error_if_nonfinite=False),
    accumulative_counts=accumulative_counts,
    loss_scale='dynamic',
    dtype='float16')

# learning policy
# More information: https://github.com/open-mmlab/mmengine/blob/main/docs/en/tutorials/param_scheduler.md  # noqa: E501
param_scheduler = [
    dict(
        type=LinearLR,
        start_factor=1e-5,
        by_epoch=True,
        begin=0,
        end=warmup_ratio * max_epochs,
        convert_to_iter_based=True),
    dict(
        type=CosineAnnealingLR,
        eta_min=0.0,
        by_epoch=True,
        begin=warmup_ratio * max_epochs,
        end=max_epochs,
        convert_to_iter_based=True)
]

# train, val, test setting
train_cfg = dict(type=TrainLoop, max_epochs=max_epochs)

#######################################################################
#                           PART 5  Runtime                           #
#######################################################################
# Log the dialogue periodically during the training process, optional
tokenizer = dict(
    type=AutoTokenizer.from_pretrained,
    pretrained_model_name_or_path=path,
    trust_remote_code=True)

custom_hooks = [
    dict(type=DatasetInfoHook, tokenizer=tokenizer),
]

# configure default hooks
default_hooks = dict(
    # record the time of every iteration.
    timer=dict(type=IterTimerHook),
    # print log every 10 iterations.
    logger=dict(type=LoggerHook, log_metric_by_epoch=False, interval=10),
    # enable the parameter scheduler.
    param_scheduler=dict(type=ParamSchedulerHook),
    # save checkpoint per `save_steps`.
    checkpoint=dict(
        type=CheckpointHook,
        save_optimizer=False,
        by_epoch=False,
        interval=save_steps,
        max_keep_ckpts=save_total_limit),
    # set sampler seed in distributed evrionment.
    sampler_seed=dict(type=DistSamplerSeedHook),
)

# configure environment
env_cfg = dict(
    # whether to enable cudnn benchmark
    cudnn_benchmark=False,
    # set multi process parameters
    mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0),
    # set distributed parameters
    dist_cfg=dict(backend='nccl'),
)

# set visualizer
visualizer = None

# set log level
log_level = 'INFO'

# load from which checkpoint
load_from = None

# whether to resume training from the loaded checkpoint
resume = False

# Defaults to use random seed and disable `deterministic`
randomness = dict(seed=None, deterministic=False)

# set log processor
log_processor = dict(by_epoch=False)

3.开始训练

使用之前的configs进行训练，要调整一下batch size，并且使用qlora

cd XTuner

NPROC_PER_NODE=1 xtuner train /root/InternLM/code/XTuner/xtuner/configs/internvl/v2/internvl_v2_internlm2_2b_qlora_finetune.py  --work-dir /root/InternLM/work_dir/internvl_ft_run_8_filter  --deepspeed deepspeed_zero1

 

4.合并权重&&模型转换

①用官方脚本进行权重合并

cd XTuner
# transfer weights
python3 xtuner/configs/internvl/v1_5/convert_to_official.py xtuner/configs/internvl/v2/internvl_v2_internlm2_2b_qlora_finetune.py /root/InternLM/work_dir/internvl_ft_run_8_filter/iter_3000.pth /root/InternLM/InternVL2-2B/

 

②最后的模型

在：/root/InternLM/convert_model/，文件格式 

.
|-- added_tokens.json
|-- config.json
|-- configuration_intern_vit.py
|-- configuration_internlm2.py
|-- configuration_internvl_chat.py
|-- conversation.py
|-- generation_config.json
|-- model.safetensors
|-- modeling_intern_vit.py
|-- modeling_internlm2.py
|-- modeling_internvl_chat.py
|-- special_tokens_map.json
|-- tokenization_internlm2.py
|-- tokenizer.model
`-- tokenizer_config.json

六、微调后效果对比

①把下面的代码替换进test_lmdeploy.py中

from lmdeploy import pipeline
from lmdeploy.vl import load_image

pipe = pipeline('/root/InternLM/InternVL2-2B')

image = load_image('/root/InternLM/007aPnLRgy1hb39z0im50j30ci0el0wm.jpg')
response = pipe(('请你根据这张图片，讲一个脑洞大开的梗', image))
print(response.text)

  

②运行

cd /root/InternLM/code

python3 test_lmdeploy.py