1--BitFit高效微调
BitFit,全称是 bias-term fine-tuning,其高效微调只去微调带有 bias 的参数,其余参数全部固定;
2--实例代码
from datasets import load_from_disk
from transformers import AutoTokenizer, AutoModelForCausalLM, DataCollatorForSeq2Seq
from transformers import pipeline, TrainingArguments, Trainer
# 分词器
tokenizer = AutoTokenizer.from_pretrained("Langboat/bloom-1b4-zh")
# 函数内将instruction和response拆开分词的原因是:
# 为了便于mask掉不需要计算损失的labels, 即代码labels = [-100] * len(instruction["input_ids"]) + response["input_ids"]
def process_func(example):
MAX_LENGTH = 256
input_ids, attention_mask, labels = [], [], []
instruction = tokenizer("\n".join(["Human: " + example["instruction"], example["input"]]).strip() + "\n\nAssistant: ")
response = tokenizer(example["output"] + tokenizer.eos_token)
input_ids = instruction["input_ids"] + response["input_ids"]
attention_mask = instruction["attention_mask"] + response["attention_mask"]
labels = [-100] * len(instruction["input_ids"]) + response["input_ids"]
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
}
if __name__ == "__main__":
# 加载数据集
dataset = load_from_disk("./PEFT/data/alpaca_data_zh")
# 处理数据
tokenized_ds = dataset.map(process_func, remove_columns = dataset.column_names)
# print(tokenizer.decode(tokenized_ds[1]["input_ids"]))
# print(tokenizer.decode(list(filter(lambda x: x != -100, tokenized_ds[1]["labels"]))))
# 创建模型
model = AutoModelForCausalLM.from_pretrained("Langboat/bloom-1b4-zh", low_cpu_mem_usage=True)
# 基于bitfit只训练带有bias的参数
for name, param in model.named_parameters():
if "bias" not in name:
param.requires_grad = False
# 训练参数
args = TrainingArguments(
output_dir = "./chatbot",
per_device_train_batch_size = 1,
gradient_accumulation_steps = 8,
logging_steps = 10,
num_train_epochs = 1
)
# trainer
trainer = Trainer(
model = model,
args = args,
train_dataset = tokenized_ds,
data_collator = DataCollatorForSeq2Seq(tokenizer=tokenizer, padding=True)
)
# 训练模型
trainer.train()
# 模型推理
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, device=0)
ipt = "Human: {}\n{}".format("考试有哪些技巧?", "").strip() + "\n\nAssistant: "
output = pipe(ipt, max_length=256, do_sample=True)
print(output)
结果: