mirror of
https://github.com/zebrajr/pytorch.git
synced 2025-12-06 12:20:52 +01:00
Essentially not allow ops on self-overlapping outputs, by adding
`at::assert_no_internal_overlap(self);` check that already used in CPU
and CUDA builds, see
|
||
|---|---|---|
| .. | ||
| _internal | ||
| __init__.py | ||
| _comparison.py | ||
| _creation.py | ||
| _utils.py | ||