Changing the shape of a numpy array

I guess you'd better create an empty 2D array and fill it later...

import numpy as np

N = how_many()
L = how_long()

complete = np.empty((N, L), dtype=float)

for i in range(N):
    complete(i, :) = extract_feature(i)

In place of dtype=float you can use a different numeric type, as required by your problem.

E.g.,

In [14]: a = np.empty((2,4), dtype=float)

In [15]: for i in (0, 1):
    ...:     a[i,:] = np.ones(4)*i
    ...:     

In [16]: a
Out[16]: 
array([[0., 0., 0., 0.],
       [1., 1., 1., 1.]])

Addendum

Comment on efficiency   If one knows beforehand the dimensions of an array they are going to construct row-by-row, the approach sketched above is better, because it avoids constantly alllocating new memory, copying the temporary result and the new row to the new memory and deallocating the memory used to hold the previous temporary result.

Alternative using concatenation   If, on the other hand, one doesn't know beforehand how many rows will be produced during the creation of the array, or if one insists on using a sub-optimal solution, they can use np.vstack, encapsulating the productions of new rows in a generator

import numpy as np

def features_factory(stuff):
    while true:
        feature = new_feature(stuff)
        if feature:
            yield feature
        else:
            return

complete = np.vstack(features_factory(todays_stuff))

E.g.,

In [1]: import numpy as np

In [2]: np.random.seed((2018+7+8)) # today's stuff... ;)

In [3]: def features_factory(stuff):
   ...:     n = 0
   ...:     while True:
   ...:         if n<stuff:
   ...:             yield np.ones(5)*n
   ...:             n = n+1
   ...:         else:
   ...:             return        

In [4]: complete = np.vstack(features_factory(np.random.randint(5,11)))

In [5]: complete
Out[5]: 
array([[0., 0., 0., 0., 0.],
       [1., 1., 1., 1., 1.],
       [2., 2., 2., 2., 2.],
       [3., 3., 3., 3., 3.],
       [4., 4., 4., 4., 4.],
       [5., 5., 5., 5., 5.],
       [6., 6., 6., 6., 6.],
       [7., 7., 7., 7., 7.]])

In [6]: