from libcity.model.abstract_model import AbstractModel
[docs]class AbstractTrafficStateModel(AbstractModel):
def __init__(self, config, data_feature):
self.data_feature = data_feature
super().__init__(config, data_feature)
[docs] def predict(self, batch):
"""
输入一个batch的数据，返回对应的预测值，一般应该是**多步预测**的结果，一般会调用nn.Moudle的forward()方法
Args:
batch (Batch): a batch of input
Returns:
torch.tensor: predict result of this batch
"""
[docs] def calculate_loss(self, batch):
"""
输入一个batch的数据，返回训练过程的loss，也就是需要定义一个loss函数
Args:
batch (Batch): a batch of input
Returns:
torch.tensor: return training loss
"""