PyTorch Lightning快速学习教程三：迁移学习

介绍：本期介绍Lightning的迁移学习

一、使用预训练的LightningModule

使用AutoEncoder作为特征提取器，同时其也作为模型的一部分

class Encoder(torch.nn.Module):
    ...

class AutoEncoder(LightningModule):
    def __init__(self):
        self.encoder = Encoder()
        self.decoder = Decoder()

class CIFAR10Classifier(LightningModule):
    def __init__(self):
        # 初始化预训练权重
        self.feature_extractor = AutoEncoder.load_from_checkpoint(PATH)
        self.feature_extractor.freeze()

        # 输出是CIFAR10分类
        self.classifier = nn.Linear(100, 10)

    def forward(self, x):
        representations = self.feature_extractor(x)
        x = self.classifier(representations)
        ...

通过上述方法来实现迁移学习

栗子1：ImageNet（计算机视觉）

import torchvision.models as models

class ImagenetTransferLearning(LightningModule):
    def __init__(self):
        super().__init__()

        # 初始化一个预训练好的resnet50
        backbone = models.resnet50(weights="DEFAULT")
        num_filters = backbone.fc.in_features
        layers = list(backbone.children())[:-1]
        self.feature_extractor = nn.Sequential(*layers)

        # 使用预训练模型对CIFAR10进行分类，用的是ImageNet的权重
        num_target_classes = 10
        self.classifier = nn.Linear(num_filters, num_target_classes)

    def forward(self, x):
        self.feature_extractor.eval()
        with torch.no_grad():
            representations = self.feature_extractor(x).flatten(1)
        x = self.classifier(representations)
        ...

Finetune（微调），进行训练

model = ImagenetTransferLearning()
trainer = Trainer()
trainer.fit(model)

进行预测

model = ImagenetTransferLearning.load_from_checkpoint(PATH)
model.freeze()

x = some_images_from_cifar10()
predictions = model(x)

imagenet的预训练模型，在CIFAR10上进行微调，以在CIFAR10上进行预测。在非学术领域，一般会对小数据集进行微调，并对数据集进行预测。一个意思。

栗子2：BERT（自然语言处理）

推荐一个transformer的git：hugging face

class BertMNLIFinetuner(LightningModule):
    def __init__(self):
        super().__init__()

        self.bert = BertModel.from_pretrained("bert-base-cased", output_attentions=True)
        self.W = nn.Linear(bert.config.hidden_size, 3)
        self.num_classes = 3

    def forward(self, input_ids, attention_mask, token_type_ids):
        h, _, attn = self.bert(input_ids=input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids)

        h_cls = h[:, 0]
        logits = self.W(h_cls)
        return logits, attn