[special] Alias for special.expm1 and special.exp2 #54670
Conversation
facebook-github-bot
added
oncall: jit
💊 CI failures summary and remediationsAs of commit 33d7264 (more details on the Dr. CI page): 💚 💚 Looks good so far! There are no failures yet. 💚 💚 This comment was automatically generated by Dr. CI (expand for details).Follow this link to opt-out of these comments for your Pull Requests.Please report bugs/suggestions to the (internal) Dr. CI Users group. |
Codecov Report
@@ Coverage Diff @@
## master #54670 +/- ##
=======================================
Coverage 77.45% 77.45%
=======================================
Files 1893 1893
Lines 186023 186029 +6
=======================================
+ Hits 144086 144097 +11
+ Misses 41937 41932 -5 |
| @@ -3177,12 +3177,14 @@ def reference_lgamma(x): | |||
| dtypesIfCUDA=floating_types_and(torch.float16), | |||
| assert_autodiffed=True), | |||
| UnaryUfuncInfo('exp2', | |||
| aliases=('special.exp2', ), | |||
| ref=np_unary_ufunc_integer_promotion_wrapper(np.exp2), | |||
There was a problem hiding this comment.
Have kept np.exp2 as reference as it matches Pytorch promotion behavior for floats (preserves the input float type) (similar for expm1).
scipy.special.exp2 promotes float16 to float32 (thus there is a mismatch in result)
>>> import torch
>>> import numpy as np
>>> import scipy.special
>>> import math
>>> t = torch.tensor([math.pi/2], dtype=torch.float16)
>>> torch.exp2(t)
tensor([2.9688], dtype=torch.float16)
>>> np.exp2(t.numpy())
array([2.969], dtype=float16)
>>> scipy.special.exp2(t.numpy())
array([2.9696903], dtype=float32)
|
@mruberry Gentle Ping :) |
| @@ -3082,7 +3082,10 @@ def merge_dicts(*dicts): | |||
|
|
|||
| .. math:: | |||
| y_{i} = 2^{x_{i}} | |||
|
|
|||
| .. note:: Alias for :func:`torch.special.exp2`. | |||
There was a problem hiding this comment.
This is OK but eventually we should just replace these doc strings with the pointer to what they alias, like the arccos docs: https://pytorch.org/docs/master/generated/torch.arccos.html?highlight=alias.
| /// auto t = torch::randn(128, dtype=kDouble); | ||
| /// torch::special::exp2(t); | ||
| /// ``` | ||
| inline Tensor exp2(const Tensor& self) { |
There was a problem hiding this comment.
What about the out variants?
|
@mruberry has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Reference: #50345