Chapter 37
Lora 实战
Notebooktransformers41 cells
Lora 实战
Step1 导入相关包
In [ ]python · cell 3
python
from datasets import Dataset
from transformers import AutoTokenizer, AutoModelForCausalLM, DataCollatorForSeq2Seq, TrainingArguments, TrainerStep2 加载数据集
In [ ]python · cell 5
python
ds = Dataset.load_from_disk("../data/alpaca_data_zh/")
dsIn [ ]python · cell 6
python
ds[:3]In [ ]python · cell 7
python
# print(len("以下是保持健康的三个提示：\n\n1. 保持身体活动。每天做适当的身体运动，如散步、跑步或游泳，能促进心血管健康，增强肌肉力量，并有助于减少体重。\n\n2. 均衡饮食。每天食用新鲜的蔬菜、水果、全谷物和脂肪含量低的蛋白质食物，避免高糖、高脂肪和加工食品，以保持健康的饮食习惯。\n\n3. 睡眠充足。睡眠对人体健康至关重要，成年人每天应保证 7-8 小时的睡眠。良好的睡眠有助于减轻压力，促进身体恢复，并提高注意力和记忆力。"))Step3 数据集预处理
In [ ]python · cell 9
python
tokenizer = AutoTokenizer.from_pretrained("D:/Pretrained_models/modelscope/Llama-2-7b-ms")
tokenizerIn [ ]python · cell 10
python
tokenizer.padding_side = "right" # 一定要设置padding_side为right，否则batch大于1时可能不收敛In [ ]python · cell 11
python
tokenizer.pad_token_id = 2In [ ]python · cell 12
python
def process_func(example):
MAX_LENGTH = 384 # Llama分词器会将一个中文字切分为多个token，因此需要放开一些最大长度，保证数据的完整性
input_ids, attention_mask, labels = [], [], []
instruction = tokenizer("\n".join(["Human: " + example["instruction"], example["input"]]).strip() + "\n\nAssistant: ", add_special_tokens=False)
response = tokenizer(example["output"], add_special_tokens=False)
input_ids = instruction["input_ids"] + response["input_ids"] + [tokenizer.eos_token_id]
attention_mask = instruction["attention_mask"] + response["attention_mask"] + [1]
labels = [-100] * len(instruction["input_ids"]) + response["input_ids"] + [tokenizer.eos_token_id]
if len(input_ids) > MAX_LENGTH:
input_ids = input_ids[:MAX_LENGTH]
attention_mask = attention_mask[:MAX_LENGTH]
labels = labels[:MAX_LENGTH]
return {
"input_ids": input_ids,
"attention_mask": attention_mask,
"labels": labels
}In [ ]python · cell 13
python
tokenized_ds = ds.map(process_func, remove_columns=ds.column_names)
tokenized_dsIn [ ]python · cell 14
python
print(tokenized_ds[0]["input_ids"])In [ ]python · cell 15
python
# tokenizer("abc " + tokenizer.eos_token)In [ ]python · cell 16
python
tokenizer.decode(tokenized_ds[0]["input_ids"])In [ ]python · cell 17
python
# tokenizer("呀", add_special_tokens=False) # Llama分词器会将一个中文字切分为多个token，因此需要放开一些最大长度，保证数据的完整性In [ ]python · cell 18
python
tokenizer.decode(list(filter(lambda x: x != -100, tokenized_ds[1]["labels"])))Step4 创建模型
In [ ]python · cell 20
python
import torch
# 多卡情况，可以去掉device_map="auto"，否则会将模型拆开
model = AutoModelForCausalLM.from_pretrained("D:/Pretrained_models/modelscope/Llama-2-7b-ms", low_cpu_mem_usage=True,
torch_dtype=torch.bfloat16, device_map="auto", load_in_8bit=True)In [ ]python · cell 21
python
for name, param in model.named_parameters():
print(name, param.shape, param.dtype)In [ ]python · cell 22
python
model.configLora
PEFT Step1 配置文件
In [ ]python · cell 25
python
from peft import LoraConfig, TaskType, get_peft_model
config = LoraConfig(task_type=TaskType.CAUSAL_LM,)
configPEFT Step2 创建模型
In [ ]python · cell 27
python
model = get_peft_model(model, config)In [ ]python · cell 28
python
configIn [ ]python · cell 29
python
model.enable_input_require_grads() # 开启梯度检查点时，要执行该方法In [ ]python · cell 30
python
# model = model.half() # 当整个模型都是半精度时，需要将adam_epsilon调大
# torch.tensor(1e-8).half() In [ ]python · cell 31
python
model.print_trainable_parameters()Step5 配置训练参数
In [ ]python · cell 33
python
args = TrainingArguments(
output_dir="./chatbot",
per_device_train_batch_size=1,
gradient_accumulation_steps=32,
logging_steps=10,
num_train_epochs=1,
gradient_checkpointing=True
)Step6 创建训练器
In [ ]python · cell 35
python
trainer = Trainer(
model=model,
args=args,
tokenizer=tokenizer,
train_dataset=tokenized_ds.select(range(6000)),
data_collator=DataCollatorForSeq2Seq(tokenizer=tokenizer, padding=True),
)Step7 模型训练
In [ ]python · cell 37
python
trainer.train()Step8 模型推理
In [ ]python · cell 39
python
model.eval()
ipt = tokenizer("Human: {}\n{}".format("你好", "").strip() + "\n\nAssistant: ", return_tensors="pt").to(model.device)
tokenizer.decode(model.generate(**ipt, max_length=128, do_sample=True, eos_token_id=tokenizer.eos_token_id)[0], skip_special_tokens=True)In [ ]python · cell 40
python
model.merge_and_unload() In [ ]python · cell 41
python