1. 环境准备
将Tutorial仓库的资料克隆到本地
mkdir -p /root/InternLM/Tutorial
git clone -b camp3 https://github.com/InternLM/Tutorial /root/InternLM/Tutorial
创建一个叫做demo的虚拟环境
# 创建虚拟环境
conda create -n demo 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 demo
# 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
pip install transformers==4.39.3
pip install streamlit==1.36.0
然后,我们安装 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 v0.1.21 https://github.com/InternLM/XTuner /root/InternLM/code/XTuner
进入源码目录,执行安装
# 进入到源码目录
cd /root/InternLM/code/XTuner
conda activate xtuner0121
# 执行安装
pip install -e '.[deepspeed]' -i https://mirrors.aliyun.com/pypi/simple/
最后,我们可以验证一下安装结果。
我们还可以查看 XTuner 的用法
2.模型准备
由于使用的是书生浦语给的开发机,因此模型实现都已经下载好了,直接通过软链接调用就行
# 创建一个目录,用来存放微调的所有资料,后续的所有操作都在该路径中进行
mkdir -p /root/InternLM/XTuner
cd /root/InternLM/XTuner
mkdir -p Shanghai_AI_Laboratory
ln -s /root/share/new_models/Shanghai_AI_Laboratory/internlm2-chat-1_8b Shanghai_AI_Laboratory/internlm2-chat-1_8b
软连接好后,我们可以通过tree来查看,如果没有安装tree,可以通过以下命令安装
apt-get install -y tree
输入
tree -l
我们可以得到
3.快速开始
3.1 初始模型
首先,我们看一下微调前的回答,我们通过以下命令启动
streamlit run /root/InternLM/Tutorial/tools/xtuner_streamlit_demo.py
8501是服务器的端口,我们需要将8501端口映射到本地,这在之前已经介绍过了,这里就不介绍了。
我们在本地电脑上打开端口,可以看到
3.2 指令跟随微调
3.2.1 准数据文件
为了让模型能够认清自己的身份弟位,在询问自己是谁的时候按照我们预期的结果进行回复,我们就需要通过在微调数据集中大量加入这样的数据。我们准备一个数据集文件datas/assistant.json
,文件内容为对话数据。
cd /root/InternLM/XTuner
mkdir -p datas
touch datas/assistant.json
为了简化数据文件准备,我们也可以通过脚本生成的方式来准备数据。创建一个脚本文件 xtuner_generate_assistant.py
:
import json
# 设置用户的名字
name = 'universe'
# 设置需要重复添加的数据次数
n = 8000
# 初始化数据
data = [
{"conversation": [{"input": "请介绍一下你自己", "output": "我是{}的小助手,内在是上海AI实验室书生·浦语的1.8B大模型哦".format(name)}]},
{"conversation": [{"input": "你在实战营做什么", "output": "我在这里帮助{}完成XTuner微调个人小助手的任务".format(name)}]}
]
# 通过循环,将初始化的对话数据重复添加到data列表中
for i in range(n):
data.append(data[0])
data.append(data[1])
# 将data列表中的数据写入到'datas/assistant.json'文件中
with open('datas/assistant.json', 'w', encoding='utf-8') as f:
# 使用json.dump方法将数据以JSON格式写入文件
# ensure_ascii=False 确保中文字符正常显示
# indent=4 使得文件内容格式化,便于阅读
json.dump(data, f, ensure_ascii=False, indent=4)
然后执行脚本生成数据
python xtuner_generate_assistant.py
执行完后,我们可以通过tree查看目录结构如下:
3.2.2 准备配置文件
配置文件其实是一种用于定义和控制模型训练和测试过程中各个方面的参数和设置的工具。
3.2.2.1 列出支持的配置文件
xtuner list-cfg
命令用于列出内置的所有配置文件。参数 -p
或 --pattern
表示模式匹配,后面跟着的内容将会在所有的配置文件里进行模糊匹配搜索,然后返回最有可能得内容。比如我们这里微调的是书生·浦语的模型,我们就可以匹配搜索 internlm2
。
xtuner list-cfg -p internlm2
可以得到
3.2.2.2 复制一个预设的配置文件
由于我们是对internlm2-chat-1_8b
模型进行指令微调,所以与我们的需求最匹配的配置文件是 internlm2_chat_1_8b_qlora_alpaca_e3
,这里就复制该配置文件。
xtuner copy-cfg
命令用于复制一个内置的配置文件。该命令需要两个参数:CONFIG
代表需要复制的配置文件名称,SAVE_PATH
代表复制的目标路径。在我们的输入的这个命令中,我们的CONFIG
对应的是上面搜索到的internlm2_chat_1_8b_qlora_alpaca_e3
,而SAVE_PATH
则是当前目录 。
我们打开internlm2_chat_1_8b_qlora_alpaca_e3_copy.py文件,进行修改
#######################################################################
# PART 1 Settings #
#######################################################################
- pretrained_model_name_or_path = 'internlm/internlm2-chat-1_8b'
+ pretrained_model_name_or_path = '/root/InternLM/XTuner/Shanghai_AI_Laboratory/internlm2-chat-1_8b'
- alpaca_en_path = 'tatsu-lab/alpaca'
+ alpaca_en_path = 'datas/assistant.json'
evaluation_inputs = [
- '请给我介绍五个上海的景点', 'Please tell me five scenic spots in Shanghai'
+ '请介绍一下你自己', 'Please introduce yourself'
]
#######################################################################
# PART 3 Dataset & Dataloader #
#######################################################################
alpaca_en = dict(
type=process_hf_dataset,
- dataset=dict(type=load_dataset, path=alpaca_en_path),
+ dataset=dict(type=load_dataset, path='json', data_files=dict(train=alpaca_en_path)),
tokenizer=tokenizer,
max_length=max_length,
- dataset_map_fn=alpaca_map_fn,
+ dataset_map_fn=None,
template_map_fn=dict(
type=template_map_fn_factory, template=prompt_template),
remove_unused_columns=True,
shuffle_before_pack=True,
pack_to_max_length=pack_to_max_length,
use_varlen_attn=use_varlen_attn)
修改后
# Copyright (c) OpenMMLab. All rights reserved.
import torch
from datasets import load_dataset
from mmengine.dataset import DefaultSampler
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 (AutoModelForCausalLM, AutoTokenizer,
BitsAndBytesConfig)
from xtuner.dataset import process_hf_dataset
from xtuner.dataset.collate_fns import default_collate_fn
from xtuner.dataset.map_fns import alpaca_map_fn, template_map_fn_factory
from xtuner.engine.hooks import (DatasetInfoHook, EvaluateChatHook,
VarlenAttnArgsToMessageHubHook)
from xtuner.engine.runner import TrainLoop
from xtuner.model import SupervisedFinetune
from xtuner.parallel.sequence import SequenceParallelSampler
from xtuner.utils import PROMPT_TEMPLATE, SYSTEM_TEMPLATE
#######################################################################
# PART 1 Settings #
#######################################################################
# Model
pretrained_model_name_or_path = '/root/InternLM/XTuner/Shanghai_AI_Laboratory/internlm2-chat-1_8b'
use_varlen_attn = False
# Data
alpaca_en_path = 'datas/assistant.json'
prompt_template = PROMPT_TEMPLATE.internlm2_chat
max_length = 2048
pack_to_max_length = True
# parallel
sequence_parallel_size = 1
# Scheduler & Optimizer
batch_size = 1 # per_device
accumulative_counts = 16
accumulative_counts *= sequence_parallel_size
dataloader_num_workers = 0
max_epochs = 3
optim_type = AdamW
lr = 2e-4
betas = (0.9, 0.999)
weight_decay = 0
max_norm = 1 # grad clip
warmup_ratio = 0.03
# Save
save_steps = 500
save_total_limit = 2 # Maximum checkpoints to keep (-1 means unlimited)
# Evaluate the generation performance during the training
evaluation_freq = 500
SYSTEM = SYSTEM_TEMPLATE.alpaca
evaluation_inputs = [
'请介绍一下你自己', 'Please introduce yourself'
]
#######################################################################
# PART 2 Model & Tokenizer #
#######################################################################
tokenizer = dict(
type=AutoTokenizer.from_pretrained,
pretrained_model_name_or_path=pretrained_model_name_or_path,
trust_remote_code=True,
padding_side='right')
model = dict(
type=SupervisedFinetune,
use_varlen_attn=use_varlen_attn,
llm=dict(
type=AutoModelForCausalLM.from_pretrained,
pretrained_model_name_or_path=pretrained_model_name_or_path,
trust_remote_code=True,
torch_dtype=torch.float16,
quantization_config=dict(
type=BitsAndBytesConfig,
load_in_4bit=True,
load_in_8bit=False,
llm_int8_threshold=6.0,
llm_int8_has_fp16_weight=False,
bnb_4bit_compute_dtype=torch.float16,
bnb_4bit_use_double_quant=True,
bnb_4bit_quant_type='nf4')),
lora=dict(
type=LoraConfig,
r=64,
lora_alpha=16,
lora_dropout=0.1,
bias='none',
task_type='CAUSAL_LM'))
#######################################################################
# PART 3 Dataset & Dataloader #
#######################################################################
alpaca_en = dict(
type=process_hf_dataset,
dataset=dict(type=load_dataset, path='json', data_files=dict(train=alpaca_en_path)),
tokenizer=tokenizer,
max_length=max_length,
dataset_map_fn=None,
template_map_fn=dict(
type=template_map_fn_factory, template=prompt_template),
remove_unused_columns=True,
shuffle_before_pack=True,
pack_to_max_length=pack_to_max_length,
use_varlen_attn=use_varlen_attn)
sampler = SequenceParallelSampler \
if sequence_parallel_size > 1 else DefaultSampler
train_dataloader = dict(
batch_size=batch_size,
num_workers=dataloader_num_workers,
dataset=alpaca_en,
sampler=dict(type=sampler, shuffle=True),
collate_fn=dict(type=default_collate_fn, use_varlen_attn=use_varlen_attn))
#######################################################################
# 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
custom_hooks = [
dict(type=DatasetInfoHook, tokenizer=tokenizer),
dict(
type=EvaluateChatHook,
tokenizer=tokenizer,
every_n_iters=evaluation_freq,
evaluation_inputs=evaluation_inputs,
system=SYSTEM,
prompt_template=prompt_template)
]
if use_varlen_attn:
custom_hooks += [dict(type=VarlenAttnArgsToMessageHubHook)]
# 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,
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)
然后,我们就可以进行训练了
3.2.3 模型训练
xtuner train ./internlm2_chat_1_8b_qlora_alpaca_e3_copy.py
训练完成后的目录如下:
3.2.4 模型转换
# 先获取最后保存的一个pth文件
pth_file=`ls -t ./work_dirs/internlm2_chat_1_8b_qlora_alpaca_e3_copy/*.pth | head -n 1`
export MKL_SERVICE_FORCE_INTEL=1
export MKL_THREADING_LAYER=GNU
xtuner convert pth_to_hf ./internlm2_chat_1_8b_qlora_alpaca_e3_copy.py ${pth_file} ./hf
转换后的目录如下
3.2.5 模型合并
对于 LoRA 或者 QLoRA 微调出来的模型其实并不是一个完整的模型,而是一个额外的层(Adapter),训练完的这个层最终还是要与原模型进行合并才能被正常的使用。
使用以下命令进行合并
cd /root/InternLM/XTuner
conda activate xtuner0121
export MKL_SERVICE_FORCE_INTEL=1
export MKL_THREADING_LAYER=GNU
xtuner convert merge /root/InternLM/XTuner/Shanghai_AI_Laboratory/internlm2-chat-1_8b ./hf ./merged --max-shard-size 2GB
模型合并后的目录:
3.3 微调后的模型对话
修改/root/InternLM/Tutorial/tools/xtuner_streamlit_demo.py
# model_name_or_path = "/root/InternLM/XTuner/Shanghai_AI_Laboratory/internlm2-chat-1_8b"
model_name_or_path ="/root/InternLM/XTuner/merged"
然后我们就可以启动微调后的大模型了
streamlit run /root/InternLM/Tutorial/tools/xtuner_streamlit_demo.py
在浏览器中输入地址,就可以了,然后我们发现回答的就已经是微调之后的数据了