TypeError: cannot astype a timedelta from [timedelta64[ns]] to [int32]

[andymancodes]

Hi,

while trying to run: fig = plot_cross_validation_metric(df_cv, metric='mape')

I am getting:

TypeError Traceback (most recent call last)
in ()
----> 1 fig = plot_cross_validation_metric(df_cv, metric='mape')

C:\Program Files\Anaconda3\envs\py35\lib\site-packages\fbprophet\plot.py in plot_cross_validation_metric(df_cv, metric, rolling_window, ax)
462 break
463
--> 464 x_plt = df_none['horizon'].astype('timedelta64[ns]').astype(int) / float(dt_conversions[i])
465 x_plt_h = df_h['horizon'].astype('timedelta64[ns]').astype(int) / float(dt_conversions[i])
466

C:\Program Files\Anaconda3\envs\py35\lib\site-packages\pandas\util_decorators.py in wrapper(*args, **kwargs)
176 else:
177 kwargs[new_arg_name] = new_arg_value
--> 178 return func(*args, **kwargs)
179 return wrapper
180 return _deprecate_kwarg

C:\Program Files\Anaconda3\envs\py35\lib\site-packages\pandas\core\generic.py in astype(self, dtype, copy, errors, **kwargs)
4995 # else, only a single dtype is given
4996 new_data = self._data.astype(dtype=dtype, copy=copy, errors=errors,
-> 4997 **kwargs)
4998 return self._constructor(new_data).finalize(self)
4999

C:\Program Files\Anaconda3\envs\py35\lib\site-packages\pandas\core\internals.py in astype(self, dtype, **kwargs)
3712
3713 def astype(self, dtype, **kwargs):
-> 3714 return self.apply('astype', dtype=dtype, **kwargs)
3715
3716 def convert(self, **kwargs):

C:\Program Files\Anaconda3\envs\py35\lib\site-packages\pandas\core\internals.py in apply(self, f, axes, filter, do_integrity_check, consolidate, **kwargs)
3579
3580 kwargs['mgr'] = self
-> 3581 applied = getattr(b, f)(**kwargs)
3582 result_blocks = _extend_blocks(applied, result_blocks)
3583

C:\Program Files\Anaconda3\envs\py35\lib\site-packages\pandas\core\internals.py in astype(self, dtype, copy, errors, values, **kwargs)
573 def astype(self, dtype, copy=False, errors='raise', values=None, **kwargs):
574 return self._astype(dtype, copy=copy, errors=errors, values=values,
--> 575 **kwargs)
576
577 def _astype(self, dtype, copy=False, errors='raise', values=None,

C:\Program Files\Anaconda3\envs\py35\lib\site-packages\pandas\core\internals.py in _astype(self, dtype, copy, errors, values, klass, mgr, **kwargs)
662
663 # _astype_nansafe works fine with 1-d only
--> 664 values = astype_nansafe(values.ravel(), dtype, copy=True)
665 values = values.reshape(self.shape)
666

C:\Program Files\Anaconda3\envs\py35\lib\site-packages\pandas\core\dtypes\cast.py in astype_nansafe(arr, dtype, copy)
694 raise TypeError("cannot astype a timedelta from [{from_dtype}] "
695 "to [{to_dtype}]".format(from_dtype=arr.dtype,
--> 696 to_dtype=dtype))
697
698 elif (np.issubdtype(arr.dtype, np.floating) and

TypeError: cannot astype a timedelta from [timedelta64[ns]] to [int32]

While, df_p = performance_metrics(df_cv) works perfectly fine.

I really need to plot the error.

[FGG100y]

Hi, @andymancodes the exactly same 'TypeError' appeared when I ran the same code. So if you just want to plot the error, maybe you could try to plot it with pd.DataFrame.plot(), it's not as pretty as the 'original ' output figure, but it work. May this would help.

[andymancodes]

@FGGsysu Hi, I am not getting how the evaluation of the model is goining on.

I have set my horizon to 30 days, yet I get

I get:

horizon mape
30 0.247988

	horizon	mse	rmse	mae	mape	coverage
3 days 01:00:00	1.196251	1.093733	0.808108	0.189526	0.929949
3 days 01:00:00	1.196157	1.093690	0.807951	0.189485	0.929949
3 days 01:00:00	1.192805	1.092156	0.806375	0.189264	0.929949
3 days 01:00:00	1.194557	1.092958	0.807047	0.189117	0.929949
3 days 01:30:00	1.194866	1.093099	0.807151	0.188838	0.929949

How to plot the error now? What does it mean? Any help would be wonderful! Thanks.

[bletham]

This is a bug that shows up on 32-bit systems (see also #591). It was fixed in #577, but the fix hasn't been pushed to PyPI yet. In the meantime you can download and install the current github version.

[bletham]

I'm going to leave this open until we push the next version to pip which will have this fix.

[ajoros]

I have latest fbprophet package and I am having this issue as well, and I am on 64-bit system. I even manually went into plot.py and modified lines 464 and 465 to use .astype(np.int64) , and still no luck. Error continues saying:
"cannot astype a timedelta from[timedelta64[ns]] to [int32]"

[bletham]

@ajoros we just pushed a new version (0.4) to PyPI that should fix this. Could you re-install fbprophet and check if it works now?

pip uninstall fbprophet
pip install fbprophet --no-cache-dir --no-binary :all:

[bletham]

The fix has been pushed to PyPI so I'll close this, but @ajoros keep me posted so we can make sure it is working in your case.