D:\Anaconda\envs\Qwen2\python.exe D:\PycharmFile\Qwen2\模型微调.py
Traceback (most recent call last):
File "D:\PycharmFile\Qwen2\模型微调.py", line 1, in <module>
from datasets import Dataset
File "D:\Anaconda\envs\Qwen2\Lib\site-packages\datasets\__init__.py", line 18, in <module>
from .arrow_dataset import Dataset
File "D:\Anaconda\envs\Qwen2\Lib\site-packages\datasets\arrow_dataset.py", line 60, in <module>
import pyarrow as pa
File "D:\Anaconda\envs\Qwen2\Lib\site-packages\pyarrow\__init__.py", line 65, in <module>
import pyarrow.lib as _lib
ImportError: DLL load failed while importing lib: 找不到指定的程序。
码神们这是为什么啊?
源码如下:
from datasets import Dataset
import pandas as pd
from transformers import (AutoTokenizer,AutoModelForCausalLM,DataCollatorForSeq2Seq,TrainingArguments,
Trainer,GenerationConfig)
import torch
from peft import LoraConfig, TaskType, get_peft_model
df=pd.read_json(r'D:\数据集\huanhuan-100.json')
ds=Dataset.from_pandas(df)
tokenizer=AutoTokenizer.from_pretrained(r"D:\模型文件夹\Qwen2",
use_fast=False,trust_remote_code=True)
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(f"<|im_start|>system\n现在你要扮演皇帝身边的女人--甄嬛<|im_end|>\n<|im_start|>user\n{example['instruction'] + example['input']}<|im_end|>\n<|im_start|>assistant\n", add_special_tokens=False) # add_special_tokens 不在开头加 special_tokens
response = tokenizer(f"{example['output']}", add_special_tokens=False)
input_ids = instruction["input_ids"] + response["input_ids"] + [tokenizer.pad_token_id]
attention_mask = instruction["attention_mask"] + response["attention_mask"] + [1] # 因为eos token咱们也是要关注的所以 补充为1
labels = [-100] * len(instruction["input_ids"]) + response["input_ids"] + [tokenizer.pad_token_id]
if len(input_ids) > MAX_LENGTH: # 做一个截断
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
}
tokenized_id = ds.map(process_func, remove_columns=ds.column_names)
model = AutoModelForCausalLM.from_pretrained(r"D:\模型文件夹\Qwen2", device_map="auto",torch_dtype=torch.bfloat16)
model.enable_input_require_grads() # 开启梯度检查点时,要执行该方法
config = LoraConfig(
task_type=TaskType.CAUSAL_LM,
target_modules=["q_proj", "k_proj", "v_proj", "o_proj", "gate_proj", "up_proj", "down_proj"],
inference_mode=False, # 训练模式
r=8, # Lora 秩
lora_alpha=32,
lora_dropout=0.1# Dropout 比例
)
model = get_peft_model(model, config)
model.print_trainable_parameters()
args = TrainingArguments(
output_dir=r"D:\微调后的模型\Qwen2-7B-甄嬛",
per_device_train_batch_size=4,
gradient_accumulation_steps=4,
logging_steps=10,
num_train_epochs=3,
save_steps=100,
learning_rate=1e-4,
save_on_each_node=True,
gradient_checkpointing=True
)
trainer = Trainer(
model=model,
args=args,
train_dataset=tokenized_id,
data_collator=DataCollatorForSeq2Seq(tokenizer=tokenizer, padding=True),
)
trainer.train()