Implement logaddexp (#38384)

Summary:
Resolve https://github.com/pytorch/pytorch/issues/38377
Related https://github.com/pytorch/pytorch/issues/38349

This op should be disambiguated with `logsumexp` which do a reduction on a tensor over a specific axis.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/38384

Differential Revision: D21737336

Pulled By: mruberry

fbshipit-source-id: 7864d04ca304c0fb2937bb083583e3e3d6ef205d
This commit is contained in:
Cloud Han
2020-05-27 20:24:21 -07:00
committed by Facebook GitHub Bot
parent 90a8cdfdbf
commit 05f097b5bb
13 changed files with 242 additions and 1 deletions

View File

@ -1742,6 +1742,20 @@ log2_() -> Tensor
In-place version of :meth:`~Tensor.log2`
""")
add_docstr_all('logaddexp',
r"""
logaddexp(other) -> Tensor
See :func:`torch.logaddexp`
""")
add_docstr_all('logaddexp2',
r"""
logaddexp2(other) -> Tensor
See :func:`torch.logaddexp2`
""")
add_docstr_all('log_normal_', r"""
log_normal_(mean=1, std=2, *, generator=None)