Does this algorithm support batch training

[ghost]

When training large dataset，it is very slow，and I found that the default batch size is 1，does it support other batch size.

[lemonwaitting]

我之前就遇到过global变量用在多进程使用的时候，经常出现问题，然后我跑了一下FISM，global变量也出现了问题，pool.map(get_train_batch, range(_num_batch)，就是在get_train_batch函数中，全局变量是None，请问您跑的时候有遇到过这个问题吗