无法在小数据集上训练多标签分类器

Question

我有一个包含 3 个类的数据集 animals、landscapes 和 buildings。每个数据集只有 100 张图像，我正在尝试使用 ResNet34 + fastai 训练分类器，但遇到了几个问题。

第一个是我认为我的模型训练不正确。当我使用 lr_finder 时，我的验证损失是 na，除非这是预期的：

然后我运行 10 个 epoch。它看起来要么学得很好，要么像地狱一样过度拟合。

然后我解冻找到另一个学习率，我的验证损失仍然是na：

现在当我训练 10 个 epoch 时，图表开始振荡，这告诉我它不能再学习了。准确度看起来不错（除非它过度拟合），所以我决定使用我的验证集来看看我的模型实际上是如何做的：

因为我有 300 张图片，所以 240 张在 train 文件夹中，60 张在 test 文件夹中。我的测试图像被标记是因为我想对其进行准确度评分。这是我运行的单元格：

    path = '/content/dataset/'

    data_test = ImageList.from_folder(path).split_by_folder(train='train', valid='test').label_from_re(file_parse).transform(size=512).databunch().normalize(imagenet_stats)

    learn = cnn_learner(data, models.resnet50, metrics=[accuracy, top_1],callback_fns=ShowGraph)

    learn.load('stage-2')

这是我遇到的错误：

    RuntimeError                              Traceback (most recent call last)
    <ipython-input-28-f1232eb8cf47> in <module>()
      4 
      5 learn = cnn_learner(data, models.resnet50, metrics=[accuracy, top_1],callback_fns=ShowGraph)
    ----> 6 learn.load('stage-2')

    1 frames
    /usr/local/lib/python3.7/dist-packages/torch/nn/modules/module.py in load_state_dict(self, state_dict, strict)
    828         if len(error_msgs) > 0:
    829             raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
    --> 830                                self.__class__.__name__, "\n\t".join(error_msgs)))
    831         return _IncompatibleKeys(missing_keys, unexpected_keys)
    832 

    RuntimeError: Error(s) in loading state_dict for Sequential:
        Missing key(s) in state_dict: "0.4.0.conv3.weight", "0.4.0.bn3.weight", "0.4.0.bn3.bias", "0.4.0.bn3.running_mean", "0.4.0.bn3.running_var", "0.4.0.downsample.0.weight", "0.4.0.downsample.1.weight", "0.4.0.downsample.1.bias", "0.4.0.downsample.1.running_mean", "0.4.0.downsample.1.running_var", "0.4.1.conv3.weight", "0.4.1.bn3.weight", "0.4.1.bn3.bias", "0.4.1.bn3.running_mean", "0.4.1.bn3.running_var", "0.4.2.conv3.weight", "0.4.2.bn3.weight", "0.4.2.bn3.bias", "0.4.2.bn3.running_mean", "0.4.2.bn3.running_var", "0.5.0.conv3.weight", "0.5.0.bn3.weight", "0.5.0.bn3.bias", "0.5.0.bn3.running_mean", "0.5.0.bn3.running_var", "0.5.1.conv3.weight", "0.5.1.bn3.weight", "0.5.1.bn3.bias", "0.5.1.bn3.running_mean", "0.5.1.bn3.running_var", "0.5.2.conv3.weight", "0.5.2.bn3.weight", "0.5.2.bn3.bias", "0.5.2.bn3.running_mean", "0.5.2.bn3.running_var", "0.5.3.conv3.weight", "0.5.3.bn3.weight", "0.5.3.bn3.bias", "0.5.3.bn3.running_mean", "0.5.3.bn3.running_var", "0.6.0.conv3.weight", "0.6.0.bn3.weight", "0.6.0.bn3.bias", "0.6.0.bn3.running_mean", "0.6.0.bn3.running_var", "0.6.1.conv3.weight", "0.6.1.bn3.weight", "0.6.1.bn3.bias", "0.6.1.bn3.running_mean", "0.6.1.bn3.running_var", "0.6.2.conv3.weight", "0.6.2.bn3.weight", "0.6.2.bn3.bias", "0.6.2.bn3.running_mean", "0.6.2.bn3.running_var", "0.6.3.conv3.weight", "0.6.3.bn3.weight", "0.6.3.bn3.bias", "0.6.3.bn3.running_mean", "0.6.3.bn3.running_var", "0.6.4.conv3.weight", "0.6.4.bn3.weight", "0.6.4.bn3.bias", "0.6.4.bn3.running_mean", "0.6....
        size mismatch for 0.4.0.conv1.weight: copying a param with shape torch.Size([64, 64, 3, 3]) from checkpoint, the shape in current model is torch.Size([64, 64, 1, 1]).
        size mismatch for 0.4.1.conv1.weight: copying a param with shape torch.Size([64, 64, 3, 3]) from checkpoint, the shape in current model is torch.Size([64, 256, 1, 1]).
        size mismatch for 0.4.2.conv1.weight: copying a param with shape torch.Size([64, 64, 3, 3]) from checkpoint, the shape in current model is torch.Size([64, 256, 1, 1]).
        size mismatch for 0.5.0.conv1.weight: copying a param with shape torch.Size([128, 64, 3, 3]) from checkpoint, the shape in current model is torch.Size([128, 256, 1, 1]).
        size mismatch for 0.5.0.downsample.0.weight: copying a param with shape torch.Size([128, 64, 1, 1]) from checkpoint, the shape in current model is torch.Size([512, 256, 1, 1]).
        size mismatch for 0.5.0.downsample.1.weight: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([512]).
        size mismatch for 0.5.0.downsample.1.bias: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([512]).
        size mismatch for 0.5.0.downsample.1.running_mean: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([512]).
        size mismatch for 0.5.0.downsample.1.running_var: copying a param with shape torch.Size([128]) from checkpoint, the shape in current model is torch.Size([512]).
        size mismatch for 0.5.1.conv1.weight: copying a param with shape torch.Size([128, 128, 3, 3]) from checkpoint, the shape in current model is torch.Size([128, 512, 1, 1]).
        size mismatch for 0.5.2.conv1.weight: copying a param with shape torch.Size([128, 128, 3, 3]) from checkpoint, the shape in current model is torch.Size([128, 512, 1, 1]).
        size mismatch for 0.5.3.conv1.weight: copying a param with shape torch.Size([128, 128, 3, 3]) from checkpoint, the shape in current model is torch.Size([128, 512, 1, 1]).
        size mismatch for 0.6.0.conv1.weight: copying a param with shape torch.Size([256, 128, 3, 3]) from checkpoint, the shape in current model is torch.Size([256, 512, 1, 1]).
        size mismatch for 0.6.0.downsample.0.weight: copying a param with shape torch.Size([256, 128, 1, 1]) from checkpoint, the shape in current model is torch.Size([1024, 512, 1, 1]).
        size mismatch for 0.6.0.downsample.1.weight: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([1024]).
        size mismatch for 0.6.0.downsample.1.bias: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([1024]).
        size mismatch for 0.6.0.downsample.1.running_mean: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([1024]).
        size mismatch for 0.6.0.downsample.1.running_var: copying a param with shape torch.Size([256]) from checkpoint, the shape in current model is torch.Size([1024]).
        size mismatch for 0.6.1.conv1.weight: copying a param with shape torch.Size([256, 256, 3, 3]) from checkpoint, the shape in current model is torch.Size([256, 1024, 1, 1]).
        size mismatch for 0.6.2.conv1.weight: copying a param with shape torch.Size([256, 256, 3, 3]) from checkpoint, the shape in current model is torch.Size([256, 1024, 1, 1]).
        size mismatch for 0.6.3.conv1.weight: copying a param with shape torch.Size([256, 256, 3, 3]) from checkpoint, the shape in current model is torch.Size([256, 1024, 1, 1]).
        size mismatch for 0.6.4.conv1.weight: copying a param with shape torch.Size([256, 256, 3, 3]) from checkpoint, the shape in current model is torch.Size([256, 1024, 1, 1]).
        size mismatch for 0.6.5.conv1.weight: copying a param with shape torch.Size([256, 256, 3, 3]) from checkpoint, the shape in current model is torch.Size([256, 1024, 1, 1]).
        size mismatch for 0.7.0.conv1.weight: copying a param with shape torch.Size([512, 256, 3, 3]) from checkpoint, the shape in current model is torch.Size([512, 1024, 1, 1]).
        size mismatch for 0.7.0.downsample.0.weight: copying a param with shape torch.Size([512, 256, 1, 1]) from checkpoint, the shape in current model is torch.Size([2048, 1024, 1, 1]).
        size mismatch for 0.7.0.downsample.1.weight: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.Size([2048]).
        size mismatch for 0.7.0.downsample.1.bias: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.Size([2048]).
        size mismatch for 0.7.0.downsample.1.running_mean: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.Size([2048]).
        size mismatch for 0.7.0.downsample.1.running_var: copying a param with shape torch.Size([512]) from checkpoint, the shape in current model is torch.Size([2048]).
        size mismatch for 0.7.1.conv1.weight: copying a param with shape torch.Size([512, 512, 3, 3]) from checkpoint, the shape in current model is torch.Size([512, 2048, 1, 1]).
        size mismatch for 0.7.2.conv1.weight: copying a param with shape torch.Size([512, 512, 3, 3]) from checkpoint, the shape in current model is torch.Size([512, 2048, 1, 1]).
        size mismatch for 1.2.weight: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([4096]).
        size mismatch for 1.2.bias: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([4096]).
        size mismatch for 1.2.running_mean: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([4096]).
        size mismatch for 1.2.running_var: copying a param with shape torch.Size([1024]) from checkpoint, the shape in current model is torch.Size([4096]).
        size mismatch for 1.4.weight: copying a param with shape torch.Size([512, 1024]) from checkpoint, the shape in current model is torch.Size([512, 4096]).

Answer 1

您的方法存在一些问题：

1. 您提到您使用 resnet 34 进行训练，然后保存其权重。然而，在测试时，您正在使用 resnet 50 并尝试加载 resnet 34 的权重，由于拱门（层和参数）不同，这将无法正常工作。

2. 如果您在同一个 NB 中进行测试，您在创建训练和有效数据集本身时添加了测试数据集，并使用 2 行代码对整个数据集进行预测。快速示例：

   test_imgs = (path/'cars_test/').ls()

   data.add_test(test_imgs)

   learn.data = 数据

   preds = learn.get_preds(ds_type=DatasetType.Test)

3. 如果您要在不同的 nb 中使用它，请尝试使用 model.export 来保存模型、其权重、数据和所有信息。我认为 jeremy 在他的 NB 和课程中做到了这一点。