print(pd.Series([1.0], dtype="float16"))
# RuntimeWarning: overflow encountered in cast np.array([1.0], dtype="float16") > 1e6
# 0 1.0
# dtype: float16
This occurs here:
https://github.com/pandas-dev/pandas/blob/bc34e2497f1471409f144fde89330a71bb8ba892/pandas/io/formats/format.py#L1446
Perhaps we just set has_large_values
to be false for float16?
Comment From: Pedro-Santos04
take
Comment From: mroeschke
I know we disallow float16 in Index
and in many places upcast float16 inputs to float32/float64. I think eventually we should deprecate float16 supports everywhere
Comment From: rhshadrach
Thanks @mroeschke - searching through the issues I see now that it was never really supported in pandas. I'll remove it from the docs in #60987.