some bugs in CategoricalFocalLoss function

[CyrilZhao-sudo]

super(BinaryFocalLoss, self).init(reduction=reduction, name=name)
the BinaryFocalLoss may be CategoricalFocalLoss

[jackguagua]

@CyrilZhao-sudo This is fixed now, thanks.

[CyrilZhao-sudo]

@CyrilZhao-sudo This is fixed now, thanks.

For the layer，MultiheadAttention， the output shape is always (batch_size, field_size, embedding_size), not (batch_size, field_size, embedding_size*num_head). en ?

[CyrilZhao-sudo]

@CyrilZhao-sudo This is fixed now, thanks.

[CyrilZhao-sudo]

Here, class DefaultPreprocessor, fit_transform method,
{if copy: # TODO bug here
X = copy.deepcopy(X)
y = copy.deepcopy(y)} . the copy may be copy_data.