Why are the dimensions of lookup_parameters flipped?

[Xiang Yu]

Hi,

I'm wondering why are the dimensions of lookup_parameters not consistent with normal parameters?

For example:

    m = dy.Model()

    p = m.add_lookup_parameters((2,3))

    print p[0].value()  # get a vector of length 3

    print p.expr()[0].value() # get a vector of length 2

Cheers,

Xiang

[Junjie Yang]

Hi,

For your example, p is of type LookupParameters. If you want to get the value of p, i recommend the use of method .as_array() instead of .expr() which is a method of type Parameters.

I hope that helps.

在 2018年4月18日星期三 UTC+8下午9:53:48，Xiang Yu写道：

[Xiang Yu]

Hi Junjie, 

thanks for the explanation! Do you also know is there any difference in terms of efficiency between these two methods?