Skip to content

Prepare branch for the TFP 0.12.0rc2 release #1174

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 39 commits into from
Nov 20, 2020
Merged

Prepare branch for the TFP 0.12.0rc2 release #1174

merged 39 commits into from
Nov 20, 2020

Conversation

@jburnim
Copy link
Member

@jburnim jburnim commented Nov 20, 2020

No description provided.

davmre and others added 30 commits November 20, 2020 10:11
@davmre @jburnim
Track parameters in STS subclasses of LinearGaussianStateSpaceModel.
e74b9cb 
Also adds a test to the harness in `structural_time_series_test`.

This fixes a bug where `ssm.copy()` didn't work.

PiperOrigin-RevId: 341714072
@tensorflower-gardener @jburnim
Merge pull request #4863 from apaszke:pmap-in-axes
16c6093 
PiperOrigin-RevId: 341715953
@axch @jburnim
Rewrite RunningMean as a CompositeTensor with methods, instead of the
1d5b092 
former split into separate controller and state classes.

PiperOrigin-RevId: 341846197
@brianwa84 @jburnim
Bug fix: Parabolic was incorrectly mapped to Epanechnikov, lacked a t…
3602bfe 
…est.

PiperOrigin-RevId: 341867916
@davmre @jburnim
Add a general-purpose bijector to numerically invert any bijective sc…
c901a98 
…alar function.

Putting this under experimental for now, since it's not clear whether this is compatible with the bijector contract.

PiperOrigin-RevId: 341889649
@davmre @jburnim
Add Chandrupatla's method for finding roots of scalar functions.
c7915d6 
It's not clear that we *really* need this, but I got most of the way through implementing it before I noticed that `secant_root` already existed. :-) So I figured I might as well check it in, and it does has somewhat different properties.

I tried to mostly mimic the interface of `secant_root`. I think the main practical distinction is that Chandrupatla's method takes a pair of bounds, instead of just an initial position, and is guaranteed to converge if those bounds bracket a root.

PiperOrigin-RevId: 341901341
@davmre @jburnim
Update the style guide to provide guidance on overloaded operators.
c90dbee 
PiperOrigin-RevId: 341939581
@emilyfertig @jburnim
Relax tolerance on HalfStudentT test due to JAX mode failure.
24c6c1a 
PiperOrigin-RevId: 342087045
@hartikainen @jburnim
Fix log_prob example in RealNVP docstring
fb6706b
@srvasude @jburnim
Fix ContinuousBernoulli quantile and sampler so that it does not emit
0b863a8 
`NaN` in the extreme cases where `p == 0` or `p == 1`.

PiperOrigin-RevId: 342127106
@jburnim
Give tfp.math.value_and_gradient more natural calling convention fo…
34144a1 
…r multiple

arg functions. Also support variables and variable tracking via tf.Module.

PiperOrigin-RevId: 342133664
@axch @jburnim
Cleaning pass on tests of potential scale reduction reducer.
8fbc527 
- Use text_fixtures instead of copying.
- Delete redundant tests
- Use `tf.while_loop` instead of Python `for` to iterate over large
  arrays.  This speeds the test up considerably, by saving on graph
  construction time.

PiperOrigin-RevId: 342273733
@hartikainen @jburnim
Change the RationalQuadraticSpline example to use non-unit batch
35c7e57
@hartikainen @jburnim
Fix bin position/slope reshaping in RationalQuadraticSpline example
4700190
@axch @jburnim
Cleaning pass on tests of covariance reducer.
6161342 
- Delete redundant tests
- Use `tf.while_loop` instead of Python `for` to iterate over large
  arrays.  This speeds the test up about 2x, by saving on graph
  construction time.

PiperOrigin-RevId: 342699805
@axch @jburnim
Use Python rather than numpy to compute constants in HalfNormal so
06166b6 
they don't defaulted to float64 precision in float64-enabled JAX.

PiperOrigin-RevId: 342720623
@hartikainen @jburnim
Use list instead of tuple for nvp.log_prob inputs
748c222
@jburnim
Bug fix for additive coupling and test to catch this bug.
ecb9e43 
PiperOrigin-RevId: 342884167
@srvasude @jburnim
Add hyp2f1 for computation of Gauss' Hypergeometric Function for inpu…
13884d1 
…ts in the unit ball.

PiperOrigin-RevId: 342899955
@axch @jburnim
Delete redundant tests of tracing reducer.
a88a623 
PiperOrigin-RevId: 342909438
@axch @jburnim
Remove redundant tests of expectations reducer
698d3e5 
PiperOrigin-RevId: 342910093
@brianwa84 @jburnim
Make the tensorfloat32 warning more targeted to only the mcmc package.
58332f9 
PiperOrigin-RevId: 342923768
@jburnim
Remove deprecated tfb.ScaleTrilL. Use tfb.FillScaleTriL instead.
9a51412 
PiperOrigin-RevId: 342924894
@axch @jburnim
Rewrite RunningCentralMoments as a CompositeTensor with methods, inst…
c9eb03a 
…ead of the

former split into separate controller and state classes.

PiperOrigin-RevId: 342956217
@SiegeLordEx @jburnim
Oryx: Allow reloading the distributions/bijectors modules.
d8f67b3 
The previous re-import logic clobbered too many builtin globals, breaking the
importlib reload functionality.

PiperOrigin-RevId: 343133516
@davmre @jburnim
Fix a numeric bug that sometimes caused Chandrapatla's method to fail…
29d9e54 
… to converge.

PiperOrigin-RevId: 343141509
@hawkinsp @jburnim
[JAX] Delete jax.source_info_util.
9b93ab9 
Fix Oryx to point to the internal name jax._src.source_info_util.

PiperOrigin-RevId: 343278153
@hawkinsp @jburnim
[JAX] Delete jax.lax_linalg.
dcde695 
Use jax.lax.linalg instead.

PiperOrigin-RevId: 343289024
@brianwa84 @jburnim
Use prefer_static in ordered_logistic.
105ccec 
PiperOrigin-RevId: 343322613
@srvasude @jburnim
Add tfp.random.spherical_uniform and make it robust in dimensions 1 a…
d649037 
…nd 2.

PiperOrigin-RevId: 343330154
brianwa84 and others added 9 commits November 20, 2020 10:27
@brianwa84 @jburnim
Fix two broadcast-not-yet-supported errors tickled by the addition of…
249adeb 
… DeterminantalPointProcess.

PiperOrigin-RevId: 343376467
@brianwa84 @jburnim
Catch another class of "rank not supported" errors from Argmin/Argmax.
1ca2d42 
PiperOrigin-RevId: 343378461
@brianwa84 @jburnim
Update numpy/jax generated tensorshape rewrite, which changed upstrea…
04cd7a9 
…m in TF.

PiperOrigin-RevId: 343380177
@brianwa84 @jburnim
Increase test size.
972387d 
PiperOrigin-RevId: 343380724
@srvasude @jburnim
Improve numerics for bessel_ive, bessel_kve by moving more computatio…
d1795f8 
…ns to log space.

PiperOrigin-RevId: 343419035
@ColCarroll @jburnim
Allow RunningVariance to be initialized from num_counts, mean, variance.
33881c0 
PiperOrigin-RevId: 343488163
@brianwa84 @jburnim
DPP has a non-trainable parameter eigenvectors (unless/until we add…
3e2ebc4 
… Expm+FillSkewSymmetric bijectors).

PiperOrigin-RevId: 343490438
@brianwa84 @jburnim
Bump atol for vmap Beta log_prob.
48ab5b0 
PiperOrigin-RevId: 343500115
@jburnim
Set the version for the TFP 0.12-rc2 release.
1fd985d
@google-cla
Copy link

google-cla bot commented Nov 20, 2020

All (the pull request submitter and all commit authors) CLAs are signed, but one or more commits were authored or co-authored by someone other than the pull request submitter.

We need to confirm that all authors are ok with their commits being contributed to this project. Please have them confirm that by leaving a comment that contains only @googlebot I consent. in this pull request.

Note to project maintainer: There may be cases where the author cannot leave a comment, or the comment is not properly detected as consent. In those cases, you can manually confirm consent of the commit author(s), and set the cla label to yes (if enabled on your project).

ℹ️ Googlers: Go here for more info.

1 similar comment
@googlebot
Copy link

googlebot commented Nov 20, 2020

All (the pull request submitter and all commit authors) CLAs are signed, but one or more commits were authored or co-authored by someone other than the pull request submitter.

We need to confirm that all authors are ok with their commits being contributed to this project. Please have them confirm that by leaving a comment that contains only @googlebot I consent. in this pull request.

Note to project maintainer: There may be cases where the author cannot leave a comment, or the comment is not properly detected as consent. In those cases, you can manually confirm consent of the commit author(s), and set the cla label to yes (if enabled on your project).

ℹ️ Googlers: Go here for more info.

@googlebot googlebot added the cla: no Declares that the user has not signed CLA label Nov 20, 2020
@jburnim jburnim added cla: yes Declares that the user has signed CLA and removed cla: no Declares that the user has not signed CLA labels Nov 20, 2020
@google-cla
Copy link

google-cla bot commented Nov 20, 2020

A Googler has manually verified that the CLAs look good.

(Googler, please make sure the reason for overriding the CLA status is clearly documented in these comments.)

ℹ️ Googlers: Go here for more info.

1 similar comment
@googlebot
Copy link

googlebot commented Nov 20, 2020

A Googler has manually verified that the CLAs look good.

(Googler, please make sure the reason for overriding the CLA status is clearly documented in these comments.)

ℹ️ Googlers: Go here for more info.

@jburnim
Copy link
Member Author

jburnim commented Nov 20, 2020

NOTE: All commits in the PR (besides 1fd985d) are cherry-picked from TFP's master banch.

@jburnim jburnim merged commit ed47dda into tensorflow:r0.12 Nov 20, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

cla: yes Declares that the user has signed CLA

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.