fast.ai课程Lesson1报错：'Sequential'对象无'fine_tune'属性

问题重现 #

代码：

dls = DataBlock(
    blocks=[ImageBlock, CategoryBlock],
    get_items=get_image_files,
    splitter=RandomSplitter(valid_pct=0.2, seed=42),
    get_y=parent_label,
    item_tfms=[Resize(192, method='squish')]
).dataloaders(path, bs=32)

learn = vision_learner(dls, resnet18, metrics=error_rate)
learn.fine_tune(3)

错误信息：

Traceback (most recent call last):
  File "/Projects/fastai/l1-image-classification/l1_image_classification/build_model.py", line 39, in <module>
    build_model()
  File "/Projects/fastai/l1-image-classification/l1_image_classification/build_model.py", line 25, in build_model
    learn.fine_tune(3)
    ^^^^^^^^^^^^^^^
  File "/Users/sys/Library/Caches/pypoetry/virtualenvs/l1-image-classification-ApaWH_9y-py3.11/lib/python3.11/site-packages/fastcore/basics.py", line 496, in __getattr__
    if attr is not None: return getattr(attr,k)
                                ^^^^^^^^^^^^^^^
  File "/Users/sys/Library/Caches/pypoetry/virtualenvs/l1-image-classification-ApaWH_9y-py3.11/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1614, in __getattr__
    raise AttributeError("'{}' object has no attribute '{}'".format(
AttributeError: 'Sequential' object has no attribute 'fine_tune'

问题原因 #

你传入vision_learner的resnet18是PyTorch原生的模型对象（通常从torchvision.models导入），而非fastai要求的模型名称字符串或适配后的模型构造器。当vision_learner接收原生PyTorch模型时，会直接返回该模型（类型为Sequential），而原生模型没有fastai的fine_tune方法。

修复方案 #

有两种正确的使用方式：

方式1：使用模型名称字符串 #

直接传入模型名称字符串，让fastai自动加载并封装为Learner实例：

learn = vision_learner(dls, 'resnet18', metrics=error_rate)

方式2：导入fastai适配的模型 #

从fastai.vision.models导入resnet18，确保是fastai适配后的版本：

from fastai.vision.models import resnet18
learn = vision_learner(dls, resnet18, metrics=error_rate)

注意：避免从torchvision.models导入resnet18，这会导致传入原生模型对象，触发上述错误。

内容的提问来源于stack exchange，提问作者clay