docs: clarify torch.arange floating-point rounding behavior

axsseldz · axsseldz · commit d2e673067725 · 2024-11-26T20:51:21.000-08:00
diff --git a/torch/_torch_docs.py b/torch/_torch_docs.py
@@ -9122,6 +9122,12 @@ def merge_dicts(*dicts):
 with values from the interval ``[start, end)`` taken with common difference
 :attr:`step` beginning from `start`.
 
+Note: When using floating-point dtypes (especially reduced precision types like ``bfloat16``),
+the results may be affected by floating-point rounding behavior. Some values in the sequence
+might not be exactly representable in certain floating-point formats, which can lead to
+repeated values or unexpected rounding. For precise sequences, it is recommended to use
+integer dtypes instead of floating-point dtypes.
+
 Note that non-integer :attr:`step` is subject to floating point rounding errors when
 comparing against :attr:`end`; to avoid inconsistency, we advise subtracting a small epsilon from :attr:`end`
 in such cases.
