pm.auto_arima "zero-size array to reduction operation maximum which has no identity" only for some columns

Question

The training split of my time series of weekly data has 105 data points, i.e. 2 years 1 week, and 7 variables. My goal is to fit a VAR. To determine the order of p and q I want to do an auto arima for each of the variables with the following code:

...
for name in train.columns:
    ...
    stepwise_model = auto_arima(train[name],start_p=1, start_q=1,max_p=10, max_q=10, seasonal=True, m=52,
        trace=True,error_action='ignore',suppress_warnings=True, stepwise=True,maxiter=1000)
...

The problem: For 5 variables, this works just fine. For 2 variables, this throws an error:

<__array_function__ internals> in ptp(*args, **kwargs)

/usr/local/lib/python3.7/dist-packages/numpy/core/_methods.py in _ptp(a, axis, out, keepdims)
    274 def _ptp(a, axis=None, out=None, keepdims=False):
    275     return um.subtract(
--> 276         umr_maximum(a, axis, None, out, keepdims),
    277         umr_minimum(a, axis, None, None, keepdims),
    278         out

ValueError: zero-size array to reduction operation maximum which has no identity

My questions:

1. What does this error mean (couldn't find anything so far)?
2. Why do I only get it for two out of my seven variables although each series has the same length?