Baichuan-13B 保姆级微调范例

lyhue1991

发布于 2023-09-05 13:39:06

1.3K0

发布于 2023-09-05 13:39:06

文章被收录于专栏：Python与算法之美

干货预警：这可能是你能够找到的最容易懂的，最完整的，适用于各种NLP任务的Baichuan-13B-Chat的finetune教程~

Baichuan-13B是百川智能于2023年7月11日发布的开源中英双语LLM，各项指标经评测在开源LLM中同尺寸模型中位居前列。

Baichuan-13B包括Baichuan-13B-Base和Baichuan-13B-chat两个不同模型。前者仅仅是预训练模型，后者在前者基础上增加了SFT,RLHF等偏好对齐过程。

本范例微调的模型是Baichuan-13B-Chat，我们使用非常简单的，外卖评论数据集来实施微调，对一段外卖评论区分是好评还是差评。

可以发现，经过微调后的模型，相比直接 3-shot-prompt 可以取得明显更好的效果(0.89->0.90)。

虽然Baichuan-13B-Chat是一个百亿级的LLM，但由于我们使用非常节约显存的QLoRA微调算法，具备32G左右显存的GPU即可实施本过程。

值得注意的是，尽管我们以文本分类任务为例，实际上，任何NLP任务，例如，命名实体识别，翻译，聊天对话等等，都可以通过加上合适的上下文，转换成一个对话问题，并针对我们的使用场景，设计出合适的数据集来微调Baichuan-13B-Chat.

注，本教程是 ChatGLM2-6b保姆级微调范例的兄弟版本~ 😋

60分钟吃掉ChatGLM2-6b微调范例~

waimai数据集简单评测对比：

〇，预训练模型

我们需要从 https://huggingface.co/baichuan-inc/Baichuan-13B-Chat 下载baichuan-13b-chat的模型。

国内可能速度会比较慢，总共有25个G左右，网速不太好的话，大概可能需要两到三个小时。

如果网络不稳定，也可以手动从这个页面一个一个下载全部文件然后放置到一个文件夹中例如 'baichuan-13b' 以便读取。

import warnings
warnings.filterwarnings('ignore')

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM,AutoConfig, AutoModel, BitsAndBytesConfig
from transformers.generation.utils import GenerationConfig
import torch.nn as nn


#使用QLoRA引入的 NF4量化数据类型以节约显存
model_name_or_path ='../baichuan-13b' #远程 'baichuan-inc/Baichuan-13B-Chat'

bnb_config=BitsAndBytesConfig(
            load_in_4bit=True,
            bnb_4bit_compute_dtype=torch.float16,
            bnb_4bit_use_double_quant=True,
            bnb_4bit_quant_type="nf4",
            llm_int8_threshold=6.0,
            llm_int8_has_fp16_weight=False,
        )

tokenizer = AutoTokenizer.from_pretrained(
   model_name_or_path, trust_remote_code=True)

model = AutoModelForCausalLM.from_pretrained(model_name_or_path,
                quantization_config=bnb_config,
                trust_remote_code=True) 

model.generation_config = GenerationConfig.from_pretrained(model_name_or_path)

from IPython.display import clear_output 
messages = []
messages.append({"role": "user",
                 "content": "世界上第二高的山峰是哪座?"})
response = model.chat(tokenizer,messages=messages,stream=True)
for res in response:
    print(res)
    clear_output(wait=True)

下面我们设计一个3-shot-prompt方法，使用外卖数据集测试一下BaiChuan13b的文本分类能力。

prefix = """外卖评论文本分类任务:
下面是一些范例:

味道真不错 -> 好评
太辣了，吃不下都  -> 差评

请对下述评论进行分类。返回'好评'或者'差评'。
"""

def get_prompt(text):
    return prefix+text+' -> '

messages  = [{"role": "user", "content": get_prompt('味道不错，下次再来')}]
response = model.chat(tokenizer, messages)
print(response)

好评

messages = messages+[{"role": "assistant", "content": response}]
print(messages)

def get_message(prompt,response):
    return [{"role": "user", "content": f'{prompt} -> '},
            {"role": "assistant", "content": response}]

messages.extend(get_message('太贵了','差评'))
messages.extend(get_message('非常快，味道好','好评'))
messages.extend(get_message('这么咸，真的是醉了','差评'))

messages

def predict(text,temperature=0.01):
    model.generation_config.temperature=temperature
    response = model.chat(tokenizer, 
                          messages = messages+[{'role':'user','content':f'{text} -> '}])
    return response

我们拿外卖数据集来测试一下未经微调，预训练模型的效果。

import pandas as pd 
import numpy as np 
import datasets 
from tqdm import tqdm

#数据集加载
dftrain = pd.read_parquet('../data/dftrain.parquet')[['text','label','tag']]
dftest = pd.read_parquet('../data/dftest.parquet')[['text','label','tag']]
ds_train,ds_val = datasets.Dataset.from_pandas(dftrain).train_test_split(
    test_size=1000,seed=42).values()\

dftrain,dfval = ds_train.to_pandas(), ds_val.to_pandas()

dftest['pred'] = [predict(text) for text in tqdm(dftest['text'])]

一，准备数据

我们仿照百川模型的 model._build_chat_input 方法来进行token编码，同时把需要学习的内容添加label.

1，token编码

import torch 

#将messages编码成 token, 同时返回labels, 该函数适用于多轮对话数据
#注意baichuan-13b通过插入tokenizer.user_token_id和tokenizer.assistant_token_id 来区分用户和机器人会话内容

# reference@ model._build_chat_input?
def build_chat_input(messages, model=model,
                     tokenizer=tokenizer, 
                     max_new_tokens: int=0):
    max_new_tokens = max_new_tokens or model.generation_config.max_new_tokens
    max_input_tokens = model.config.model_max_length - max_new_tokens
    max_input_tokens = max(model.config.model_max_length // 2, max_input_tokens)
    
    total_input, round_input, total_label, round_label = [], [], [], []
    
    for i, message in enumerate(messages[::-1]):
        content_tokens = tokenizer.encode(message['content'])
        if message['role'] == 'user':
            round_input = [model.generation_config.user_token_id] + content_tokens + round_input
            round_label = [-100]+[-100 for _ in content_tokens]+ round_label
            
            if total_input and len(total_input) + len(round_input) > max_input_tokens:
                break
            else:
                total_input = round_input + total_input
                total_label = round_label + total_label
                if len(total_input) >= max_input_tokens:
                    break
                else:
                    round_input = []
                    round_label = []
                    
        elif message['role'] == 'assistant':
            round_input = [
                model.generation_config.assistant_token_id
            ] + content_tokens + [
                model.generation_config.eos_token_id
            ] + round_input

            round_label = [
                -100
            ] + content_tokens + [
                model.generation_config.eos_token_id
            ]+ round_label
        else:
            raise ValueError(f"message role not supported yet: {message['role']}")
            
    total_input = total_input[-max_input_tokens:]  # truncate left
    total_label = total_label[-max_input_tokens:]
    
    total_input.append(model.generation_config.assistant_token_id)
    total_label.append(-100)
    
    return total_input,total_label

2，做数据集

from torch.utils.data import Dataset,DataLoader 
class MyDataset(Dataset):
    def __init__(self,df,
                 prefix=prefix
                ):
        self.df = df 
        self.prefix=prefix
        
    def __len__(self):
        return len(self.df)
        
    def get_samples(self,index):
        samples = []
        d = dict(self.df.iloc[index])
        samples.append(d)
        return samples
    
    def get_messages(self,index):
        samples = self.get_samples(index)
        messages = []
        for i,d in enumerate(samples):
            if i==0:
                messages.append({'role':'user','content':self.prefix+d['text']+' -> '})
            else:
                messages.append({'role':'user','content':d['text']+' -> '})
            
            messages.append({'role':'assistant','content':d['tag']})
        return messages
        
    def __getitem__(self,index):
        messages = self.get_messages(index)
        input_ids, labels = build_chat_input(messages)
        return {'input_ids':input_ids,'labels':labels}

    def show_sample(self,index):
        samples = self.get_samples(index)
        print(samples)

ds_train = MyDataset(dftrain)
ds_val = MyDataset(dfval)

3，创建管道

def data_collator(examples: list):
    len_ids = [len(example["input_ids"]) for example in examples]
    longest = max(len_ids) #之后按照batch中最长的input_ids进行padding
    
    input_ids = []
    labels_list = []
    
    for length, example in sorted(zip(len_ids, examples), key=lambda x: -x[0]):
        ids = example["input_ids"]
        labs = example["labels"]
        
        ids = ids + [tokenizer.pad_token_id] * (longest - length)
        labs = labs + [-100] * (longest - length)
        
        input_ids.append(torch.LongTensor(ids))
        labels_list.append(torch.LongTensor(labs))
          
    input_ids = torch.stack(input_ids)
    labels = torch.stack(labels_list)
    return {
        "input_ids": input_ids,
        "labels": labels,
    }

import torch 
dl_train = torch.utils.data.DataLoader(ds_train,num_workers=2,batch_size=4,
                                       pin_memory=True,shuffle=True,
                                       collate_fn = data_collator)

dl_val = torch.utils.data.DataLoader(ds_val,num_workers=2,batch_size=4,
                                    pin_memory=True,shuffle=False,
                                     collate_fn = data_collator)

for batch in dl_train:
    break

二，定义模型

下面我们将使用QLoRA(实际上用的是量化的AdaLoRA）算法来微调Baichuan-13b模型。

from peft import get_peft_config, get_peft_model, TaskType
model.supports_gradient_checkpointing = True  #
model.gradient_checkpointing_enable()
model.enable_input_require_grads()

model.config.use_cache = False  # silence the warnings. Please re-enable for inference!

import bitsandbytes as bnb 
def find_all_linear_names(model):
    """
    找出所有全连接层，为所有全连接添加adapter
    """
    cls = bnb.nn.Linear4bit
    lora_module_names = set()
    for name, module in model.named_modules():
        if isinstance(module, cls):
            names = name.split('.')
            lora_module_names.add(names[0] if len(names) == 1 else names[-1])

    if 'lm_head' in lora_module_names:  # needed for 16-bit
        lora_module_names.remove('lm_head')
    return list(lora_module_names)

from peft import prepare_model_for_kbit_training 
model = prepare_model_for_kbit_training(model)

lora_modules = find_all_linear_names(model)
print(lora_modules)

['down_proj', 'o_proj', 'up_proj', 'W_pack', 'gate_proj']

from peft import AdaLoraConfig
peft_config = AdaLoraConfig(
    task_type=TaskType.CAUSAL_LM, inference_mode=False,
    r=64,
    lora_alpha=16, lora_dropout=0.05,
    target_modules= lora_modules
)

peft_model = get_peft_model(model, peft_config)

peft_model.is_parallelizable = True
peft_model.model_parallel = True
peft_model.print_trainable_parameters()

三，训练模型

from torchkeras import KerasModel 
from accelerate import Accelerator 

class StepRunner:
    def __init__(self, net, loss_fn, accelerator=None, stage = "train", metrics_dict = None, 
                 optimizer = None, lr_scheduler = None
                 ):
        self.net,self.loss_fn,self.metrics_dict,self.stage = net,loss_fn,metrics_dict,stage
        self.optimizer,self.lr_scheduler = optimizer,lr_scheduler
        self.accelerator = accelerator if accelerator is not None else Accelerator() 
        if self.stage=='train':
            self.net.train() 
        else:
            self.net.eval()
    
    def __call__(self, batch):
        
        #loss
        with self.accelerator.autocast():
            loss = self.net.forward(**batch)[0]

        #backward()
        if self.optimizer is not None and self.stage=="train":
            self.accelerator.backward(loss)
            if self.accelerator.sync_gradients:
                self.accelerator.clip_grad_norm_(self.net.parameters(), 1.0)
            self.optimizer.step()
            if self.lr_scheduler is not None:
                self.lr_scheduler.step()
            self.optimizer.zero_grad()
            
        all_loss = self.accelerator.gather(loss).sum()
        
        #losses (or plain metrics that can be averaged)
        step_losses = {self.stage+"_loss":all_loss.item()}
        
        #metrics (stateful metrics)
        step_metrics = {}
        
        if self.stage=="train":
            if self.optimizer is not None:
                step_metrics['lr'] = self.optimizer.state_dict()['param_groups'][0]['lr']
            else:
                step_metrics['lr'] = 0.0
        return step_losses,step_metrics
    
KerasModel.StepRunner = StepRunner 

#仅仅保存QLora可训练参数
def save_ckpt(self, ckpt_path='checkpoint', accelerator = None):
    unwrap_net = accelerator.unwrap_model(self.net)
    unwrap_net.save_pretrained(ckpt_path)
    
def load_ckpt(self, ckpt_path='checkpoint'):
    import os
    self.net.load_state_dict(
        torch.load(os.path.join(ckpt_path,'adapter_model.bin')),strict =False)
    self.from_scratch = False
    
KerasModel.save_ckpt = save_ckpt 
KerasModel.load_ckpt = load_ckpt

optimizer = bnb.optim.adamw.AdamW(peft_model.parameters(),
                                  lr=6e-05,is_paged=True)  #'paged_adamw'
keras_model = KerasModel(peft_model,loss_fn =None,
        optimizer=optimizer) 
ckpt_path = 'baichuan13b_waimai'

# keras_model.load_ckpt(ckpt_path) #支持加载微调后的权重继续训练(断点续训)
keras_model.fit(train_data = dl_train,
                val_data = dl_val,
                epochs=100,patience=10,
                monitor='val_loss',mode='min',
                ckpt_path = ckpt_path
               )

四，保存模型

为减少GPU压力，此处可重启kernel释放显存

import warnings 
warnings.filterwarnings('ignore')

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM,AutoConfig, AutoModel, BitsAndBytesConfig
from transformers.generation.utils import GenerationConfig
import torch.nn as nn
model_name_or_path ='../baichuan-13b'
ckpt_path = 'baichuan13b_waimai'
tokenizer = AutoTokenizer.from_pretrained(
    model_name_or_path,
    trust_remote_code=True
)
model_old = AutoModelForCausalLM.from_pretrained(
    model_name_or_path,
    trust_remote_code=True,
    low_cpu_mem_usage=True,
    torch_dtype=torch.float16,
    device_map='auto'
)

from peft import PeftModel

#可能需要5分钟左右
peft_model = PeftModel.from_pretrained(model_old, ckpt_path)
model_new = peft_model.merge_and_unload()

from transformers.generation.utils import GenerationConfig
model_new.generation_config = GenerationConfig.from_pretrained(model_name_or_path)

from IPython.display import clear_output
messages = []
messages.append({"role": "user",
                 "content": "世界上第二高的山峰是什么？"})
response = model_new.chat(tokenizer,messages=messages,stream=True)
for res in response:
    print(res)
    clear_output(wait=True)

save_path = 'baichuan-13b-waimai'

tokenizer.save_pretrained(save_path)
model_new.save_pretrained(save_path)

!cp baichuan-13b/*.py  baichuan-13b-waimai

五，使用模型

为减少GPU压力，此处可再次重启kernel释放显存。

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM,AutoConfig, BitsAndBytesConfig
from transformers.generation.utils import GenerationConfig
import torch.nn as nn

import warnings
warnings.filterwarnings('ignore')

model_name_or_path = 'baichuan-13b-waimai'

bnb_config=BitsAndBytesConfig(
            load_in_4bit=True,
            bnb_4bit_compute_dtype=torch.float16,
            bnb_4bit_use_double_quant=True,
            bnb_4bit_quant_type="nf4",
            llm_int8_threshold=6.0,
            llm_int8_has_fp16_weight=False,
        )

tokenizer = AutoTokenizer.from_pretrained(
   model_name_or_path, trust_remote_code=True)

model = AutoModelForCausalLM.from_pretrained(model_name_or_path,
                quantization_config=bnb_config,
                trust_remote_code=True) 

model.generation_config = GenerationConfig.from_pretrained(model_name_or_path)

from IPython.display import clear_output
messages = []
messages.append({"role": "user",
                 "content": "世界上第二高的山峰是什么？"})
response = model.chat(tokenizer,messages=messages,stream=True)
for res in response:
    print(res)
    clear_output(wait=True)

乔戈里峰。世界第二高峰———乔戈里峰
海拔高度：8610米
坐标纬度：35°49′15′′n,76°21′24′′e
地理位置：喀喇昆仑山脉中巴边境上

我们测试一下微调后的效果。

import pandas as pd 
import numpy as np 
import datasets 
from tqdm import tqdm

prefix = """外卖评论文本分类任务:
下面是一些范例:

味道真不错 -> 好评
太辣了，吃不下都  -> 差评

请对下述评论进行分类。返回'好评'或者'差评'。
"""

def get_prompt(text):
    return prefix+text+' -> '

messages  = [{"role": "user", "content": get_prompt('味道不错，下次再来')}]
response = model.chat(tokenizer, messages)
print(response)

好评

messages = messages+[{"role": "assistant", "content": response}]
print(messages)

def get_message(prompt,response):
    return [{"role": "user", "content": f'{prompt} -> '},
            {"role": "assistant", "content": response}]

messages.extend(get_message('太贵了','差评'))
messages.extend(get_message('非常快，味道好','好评'))
messages.extend(get_message('这么咸，真的是醉了','差评'))

def predict(text,temperature=0.01):
    model.generation_config.temperature=temperature
    response = model.chat(tokenizer, 
                          messages = messages+[{'role':'user','content':f'{text} -> '}])
    return response