Why do I get the trouble with “cannot cast array data from dtype('float64') to dtype('

Question

The error

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

occurred when I use np.dot(a,b.T) in NumPy version 1.11.0. I've tried to fix it for some days but have not resolved it yet. Can someone suggest a relevant solution to fix the cannot cast array data from dtype('float64') to dtype('<u32') according to the rule 'safe' issue? Much appreciate your support.

lyytutoria · Accepted Answer

The cause: You get trouble with ''cannot cast array data from dtype(‘float64’) to dtype(‘ Solution: You can solve this problem by changing your strings to floats, the following code snippet works well based on the input from BrenBarn and Warren Weckesser:

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

Or you can do simply as below:

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

However, you should check all the sorts of the array, probably one of them includes floats as Warren Weckesser has already indicated.

Gabriel Castel · Answer

You can convert whole numpy array to float example:

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

Sign Up

Sign In

Forgot Password

ITtutoria Latest Questions

Why do I get the trouble with “cannot cast array data from dtype(‘float64’) to dtype(‘<u32’) according to the rule ‘safe’”?