convert list of strings to numpy list of lists

Question

I'm reading data (numbers) from a file into a list, as follows:

    weight_file = open(model_name, 'r').readlines()
    weights = weight_file[6:]

It seems that I can't read them straight into a numpy.array because the first rows of the file contains words.

So now, in weights I have a list as follows: [['1 2 3'] ['4 5 6']]

Now I want to convert it into a numpy.array. I tried this:

    weights_np = np.array([])
    for weight in weights:
        weights_np = np.append(weights_np, weight.split())

What this does is create a single vector: [1, 2, 3, 4, 5, 6], but I need it to be represented as some kind of matrix - like a list of lists. How can I do that?

Thanks

unutbu · Accepted Answer · 2014-12-12 13:06:41Z

2

If weights_np is this:

In [23]: weights_np = np.array([1, 2, 3, 4, 5, 6])

then you could use reshape to make it 2-dimensional with 3 columns:

In [24]: weights_np = weights_np.reshape((-1, 3))

In [25]: weights_np
Out[25]: 
array([[1, 2, 3],
       [4, 5, 6]])

But a more satisfying method would be to parse the file properly with np.loadtxt or np.genfromtxt. (This would be much much faster than using Python loops and calling np.append.) Note that these functions have a skiprows parameter which you could use to skip the first few rows:

weights = np.loadtxt(model_name, skiprows=6)

answered Dec 12, 2014 at 13:06

unutbu

886k197 gold badges1.9k silver badges1.7k bronze badges

Sign up to request clarification or add additional context in comments.

Comments

Pierre · Accepted Answer · 2014-12-12 14:45:31Z

2

To get a "kind of matrix", simply use:

numpy.array([[int(val) for val in line.split()] for line in open(model_name)])

To get a simple vector, use:

numpy.array([int(val) for line in open(model_name) for val in line.split()])

edited Dec 12, 2014 at 14:45

answered Dec 12, 2014 at 13:08

Pierre

6,3071 gold badge34 silver badges52 bronze badges

Comments

a5kin · Accepted Answer · 2014-12-12 13:14:38Z

1

>>> import numpy as np
>>> weights = ['1 2 3', '4 5 6']
>>> weights_np = np.array(map(lambda x: map(int, x.split()), weights))
>>> weights_np
array([[1, 2, 3],
       [4, 5, 6]])

answered Dec 12, 2014 at 13:14

a5kin

1,44518 silver badges20 bronze badges

Comments

Irshad Bhat · Accepted Answer · 2014-12-12 13:29:23Z

1

Try it this way:

weights_np = np.empty((0,len(weights[0].split())),int)
for weight in weights:
    weights_np = np.append(weights_np, [map(int,weight.split())], axis=0)

Demo:

>>> weights = ['1 2 3','4 5 6']
>>> weights_np = np.empty((0,len(weights[0].split())),int)
>>> for weight in weights:
...     weights_np = np.append(weights_np, [map(int,weight.split())], axis=0)
... 
>>> weights_np
array([[1, 2, 3],
       [4, 5, 6]])

edited Dec 12, 2014 at 13:29

answered Dec 12, 2014 at 13:24

Irshad Bhat

8,7792 gold badges31 silver badges37 bronze badges

Collectives™ on Stack Overflow

convert list of strings to numpy list of lists

4 Answers 4

Comments

Comments

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

4 Answers 4

Comments

Comments

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related