如何修复 AttributeError 错误：“列表”对象没有属性“大小”？ [已关闭]

Question

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
from datasets import load_dataset
from torch.utils.data import DataLoader
import torch
from tqdm import tqdm

# Загрузка модели и токенизатора
model_name = "google/flan-t5-base"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)

# Загрузка и подготовка данных
def load_and_prepare_data(dataset_name="squad"):
    dataset = load_dataset(dataset_name, split='train')
    dataset = dataset.filter(lambda example: len(example['context']) <= 512)
    dataset = dataset.map(
        lambda example: {
            'input_ids': tokenizer(example['question'], example['context'], truncation=True, padding="max_length", return_tensors="pt")['input_ids'].squeeze(0),
            'attention_mask': tokenizer(example['question'], example['context'], truncation=True, padding="max_length", return_tensors="pt")['attention_mask'].squeeze(0),
            'labels': tokenizer(example['answers']['text'][0], truncation=True, padding="max_length", return_tensors="pt")['input_ids'].squeeze(0)
        }
    )
    
    # Разделяем данные на тренировочный и тестовый наборы
    train_size = int(0.8 * len(dataset))
    train_data, test_data = torch.utils.data.random_split(dataset, [train_size, len(dataset) - train_size])
    
    # Создаем DataLoader для тренировочных и тестовых данных
    train_dataloader = DataLoader(train_data, batch_size=8, shuffle=True)
    test_dataloader = DataLoader(test_data, batch_size=8)
    
    return train_dataloader, test_dataloader

# Обучение модели
def train_model(model, train_dataloader, test_dataloader, epochs=3):
    optimizer = torch.optim.Adam(model.parameters(), lr=1e-5)

    for epoch in range(epochs):
        model.train()
        for batch in tqdm(train_dataloader, desc=f"Epoch {epoch+1}"):
            optimizer.zero_grad()
            input_ids = batch['input_ids']
            attention_mask = batch['attention_mask']
            labels = batch['labels']
            outputs = model(input_ids=input_ids, attention_mask=attention_mask, labels=labels)
            loss = outputs.loss
            loss.backward()
            optimizer.step()

        model.eval()
        total_loss = 0
        for batch in tqdm(test_dataloader, desc=f"Epoch {epoch+1} - Validation"):
            with torch.no_grad():
                input_ids = batch['input_ids']
                attention_mask = batch['attention_mask']
                labels = batch['labels']
                outputs = model(input_ids=input_ids, attention_mask=attention_mask, labels=labels)
                total_loss += outputs.loss.item()
        avg_loss = total_loss / len(test_dataloader)
        print(f"Epoch {epoch+1} - Validation Loss: {avg_loss:.4f}")

# Загрузка и подготовка данных
train_dataloader, test_dataloader = load_and_prepare_data()

# Обучение модели
train_model(model, train_dataloader, test_dataloader)

# Сохранение модели
torch.save(model.state_dict(), "chatbot_model.pth")

# Основной цикл
while True:
    user_input = input("You: ")

    if user_input.lower() == "exit":
        break
    else:
        # Генерация ответа
        inputs = tokenizer(user_input, return_tensors="pt")
        outputs = model.generate(**inputs, max_length=50)
        response = tokenizer.decode(outputs[0], skip_special_tokens=True)
        print(f"Bot: {response}")

Error:
AttributeError                            Traceback (most recent call last)
<ipython-input-6-fadd0eb00ebb> in <cell line: 66>()
     64 
     65 # Обучение модели
---> 66 train_model(model, train_dataloader, test_dataloader)
     67 
     68 # Сохранение модели

/usr/local/lib/python3.10/dist-packages/transformers/models/t5/modeling_t5.py in forward(self, input_ids, attention_mask, encoder_hidden_states, encoder_attention_mask, inputs_embeds, head_mask, cross_attn_head_mask, past_key_values, use_cache, output_attentions, output_hidden_states, return_dict)
    999             )
   1000         elif input_ids is not None:
-> 1001             input_shape = input_ids.size()
   1002             input_ids = input_ids.view(-1, input_shape[-1])
   1003         elif inputs_embeds is not None:

AttributeError: 'list' object has no attribute 'size'

我还是不明白这是什么错误。

Answer 1

发生此错误是因为

input_ids

（以及其他输入，如

attention_mask

和

labels

）在

train_model

函数中作为列表而不是张量传递。 PyTorch 模型需要张量，而不是列表，因此您需要确保所有数据在传递给模型之前都采用正确的格式。

这是代码：

def train_model(model, train_dataloader, test_dataloader, epochs=3):
    optimizer = torch.optim.Adam(model.parameters(), lr=1e-5)

    for epoch in range(epochs):
        model.train()
        for batch in tqdm(train_dataloader, desc=f"Epoch {epoch+1}"):
            optimizer.zero_grad()
            input_ids = torch.stack(batch['input_ids']).to(model.device)  # Convert list to tensor
            attention_mask = torch.stack(batch['attention_mask']).to(model.device)  # Convert list to tensor
            labels = torch.stack(batch['labels']).to(model.device)  # Convert list to tensor
            outputs = model(input_ids=input_ids, attention_mask=attention_mask, labels=labels)
            loss = outputs.loss
            loss.backward()
            optimizer.step()

        model.eval()
        total_loss = 0
        for batch in tqdm(test_dataloader, desc=f"Epoch {epoch+1} - Validation"):
            with torch.no_grad():
                input_ids = torch.stack(batch['input_ids']).to(model.device)  # Convert list to tensor
                attention_mask = torch.stack(batch['attention_mask']).to(model.device)  # Convert list to tensor
                labels = torch.stack(batch['labels']).to(model.device)  # Convert list to tensor
                outputs = model(input_ids=input_ids, attention_mask=attention_mask, labels=labels)
                total_loss += outputs.loss.item()
        avg_loss = total_loss / len(test_dataloader)
        print(f"Epoch {epoch+1} - Validation Loss: {avg_loss:.4f}")

希望这对你有一点帮助。