Numpy.dot TypeError: Cannot cast array data from dtype('float64') to dtype('S32') according to the rule 'safe'

Question

Why am I getting this error when using np.dot(a,b.T):

TypeError: Cannot cast array data from dtype('float64') 
               to dtype('S32') according to the rule 'safe'

a and b are of type numpy.ndarray. My NumPy version is 1.11.0.

You need to show a self-contained example with sample data. The error message is saying that some of your data are floats and some are strings. — BrenBarn
– BrenBarn, Commented Dec 9, 2015 at 7:39
'S32' means one of your arrays is an array of strings, not numbers. Take a closer look at your arrays and how you create them. In particular, check a.dtype and b.dtype. — Warren Weckesser
– Warren Weckesser, Commented Dec 9, 2015 at 7:58

Romeo Kienzler · Accepted Answer · 2017-08-20 12:21:52Z

27

Just taking the input from BrenBarn and Warren Weckesser to provide a code snippet which should run (by converting your strings to float):

a = map(lambda x: float(x),a)
b = map(lambda x: float(x),b)
np.dot(a,b.T)

or simpler as suggested by @JLT

a = map(float,a)
b = map(float,b)
np.dot(a,b.T)

But as Warren Weckesser already said, you should check the types of the array, most likely one already contains floats.

answered Sep 5, 2016 at 4:14

Romeo Kienzler

3,5634 gold badges43 silver badges63 bronze badges

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

1 Comment

wouldn't map(float, a) be simpler?

Gaurav Gandhi · Accepted Answer · 2018-10-06 14:49:14Z

18

Try converting whole numpy array into float Example:

train = train.astype(float)
train_target = train_target.astype(float)

Gaurav Gandhi

3,2112 gold badges30 silver badges40 bronze badges

answered Oct 31, 2017 at 6:17

VishnuVardhanA

6276 silver badges7 bronze badges