在 keras 的 model.fit 中随机播放

Question

在keras的model.fit中，有一个shuffle参数，

shuffle: Boolean (whether to shuffle the training data before each epoch) or str (for 'batch'). 'batch' is a special option for dealing with the limitations of HDF5 data; it shuffles in batch-sized chunks. Has no effect when steps_per_epoch is not  None.

假设训练集是一个带有50000元素的列表，那么整个列表会在每个epoch之前随机排列吗？如果批量大小是250，只有属于每个批次的元素会被置换？正确的理解应该是什么？

Answer 1

根据您传递给fit 的batch_size 参数，它将先shuffle your entire dataset (x, y and sample_weight together) first 然后make batches。

编辑

正如@yuk 在评论中指出的那样，代码自 2018 年以来发生了重大变化。shuffle 参数的文档现在似乎更清楚了。您可以选择打乱整个训练数据或只是打乱批次：

        shuffle: Boolean (whether to shuffle the training data
            before each epoch) or str (for 'batch'). This argument is ignored
            when `x` is a generator. 'batch' is a special option for dealing
            with the limitations of HDF5 data; it shuffles in batch-sized
            chunks. Has no effect when `steps_per_epoch` is not `None`.