[AutoDiff][TF-1200] Adding derivatives for stdlib pow function.
#30580
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
vguerra
commented
Mar 23, 2020
dan-zheng
requested changes
Mar 23, 2020
vguerra
force-pushed
the
tf-1200
branch
2 times, most recently
from
1a7caab to
44d1a49
Compare
vguerra
commented
Apr 19, 2020
| @@ -38,17 +38,40 @@ func expectEqualWithTolerance<T>(_ expected: TestLiteralType, _ actual: T, | |||
| file: file, line: line) | |||
| } | |||
|
|
|||
| func computeDividedDifference<T: BinaryFloatingPoint> ( | |||
There was a problem hiding this comment.
I tried to use instead symmetric derivative formula to compute the numerical derivative, which gives a more precise result hence allowing to use lower values for ulps but some test cases for reminder broke, hence I dropped that change, but maybe we should use it in the future.
vguerra
commented
Apr 19, 2020
|
|
||
| // pow | ||
| let eps:${T} = 0.01 | ||
| let ulps:${T} = eps/eps.ulp |
There was a problem hiding this comment.
here I compute the needed ulps for comparing values as the resulting numerical derivative depends on the ε used to compute it.
|
hello @dan-zheng, sorry that it took me this long to address your comments but I finally had time this weekend to address them. I needed to rebase my PR due to conflicts, so I did :/. |
marcrasi
suggested changes
Apr 21, 2020
Adding JVP and VJP function derivatives for `pow` function defined in stdlib. Resolves TF-1200.
The versions of VJP and JVP match the results that TensorFlow woudl give. As well, this extends the test cases to include combinations of positive and negative values and corner cases.
We now have 3 sections for the tests: negative base, 0 base and positive base.
vguerra
changed the title
pow function.pow function.
marcrasi
approved these changes
May 1, 2020
|
@swift-ci please test |
|
No worries @marcrasi, thanks! |
|
@swift-ci please test |
rxwei
pushed a commit
to rxwei/swift
that referenced
this pull request
Jun 3, 2020
…iftlang#30580) Adding JVP and VJP function derivatives for pow function defined in stdlib. Resolves TF-1200.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Adding JVP and VJP function derivatives for
powfunctiondefined in stdlib.
Resolves TF-1200.