WillianFuks/tfcausalimpact

TypeError: ufunc 'isfinite' not supported for the input types

Queely1 opened this issue · 5 comments

Queely1 commented

Good afternoon, after the update, the following error occurred.

TypeError: ufunc 'isfinite' not supported for the input types, and the inputs could not be safely coerced to any supported types according to the casting rule ''safe''.

The code worked fine before. Summary and report are also shown, the error appears only when you try to draw a graph.

Rollback to the previous version could not, the error now appears on any versions.
image
image

WillianFuks commented

Hi @Queely1 ,

I noticed it yesterday as well and opened an issue for the numpy repository, this is a bug for version 1.24.

Just reinstall previous version:

pip install -U numpy==1.23.4

And things should be back to normal.

They are already working on a fix for this issue so I'm hoping this will be solved soon.

Let me know how this works for you.

  • 👍1
Queely1 commented

Works, thanks a lot!

  • 👍1
ecarruda commented

It is also breaking float64 types:

SingleBlockManager
Items: Float64Index([ -2.408224685280692, -2.3739711708557603, -2.3397176564308286,
               -2.305464142005897,  -2.271210627580966,  -2.236957113156034,
              -2.2027035987311026,  -2.168450084306171, -2.1341965698812393,
              -2.0999430554563077,
              ...
                4.099943055456308,   4.134196569881238,   4.168450084306171,
                4.202703598731102,   4.236957113156034,   4.271210627580965,
                4.305464142005897,   4.339717656430828,   4.373971170855761,
               4.4082246852806914],
             dtype='float64', length=200)
NumericBlock: 200 dtype: float64

  res = masked_where(~(np.isfinite(getdata(a))), a, copy=copy)

E TypeError: ufunc 'isfinite' not supported for the input types, and the inputs could not be safely coerced to any supported types according to the casting rule ''safe''

WillianFuks commented

Hi @ecarruda ,

Just to confirm, is it still breaking after downgrading the numpy version?

waral commented

Hi,

I just ran into the same issue and using .values when plotting confidence intervals, e.g.:

ax.fill_between(
            pre_post_index[1:],
            inferences['complete_preds_lower'].iloc[1:].values,
            inferences['complete_preds_upper'].iloc[1:].values,
            color=color,
            alpha=0.4
)

seems to solve the issue without downgrading the numpy version.

  • 👍1