GitLab

PiperOrigin-RevId: 365678315

PiperOrigin-RevId: 365665625

PiperOrigin-RevId: 365639389

adding `beta_1` parameter to `create_optimizer` in `optimization.py` to allow for a tunable `beta_1`. Defaults to the original `0.9` to ensure existing compatibility.

PiperOrigin-RevId: 365634937

PiperOrigin-RevId: 365582100

PiperOrigin-RevId: 365507990

PiperOrigin-RevId: 365275913

PiperOrigin-RevId: 365254703

PiperOrigin-RevId: 365242134

PiperOrigin-RevId: 365188568

Change "push" to "pull_request" to trigger the workflow on pull request events only

PiperOrigin-RevId: 365177695

PiperOrigin-RevId: 365133146

* Create GitHub Action for pylint script.

* Fix indent

* Add back incremental flag from TF, remove allowlist.

Running a full-repo pylint with Github Actions CI takes >40 minutes,
too long for use on presubmit.

Allowlist is per-file, so the TF allowlist offers nothing here.

* Fetch master so our script has something to diff against.

* Checkout as branch rather than as merge.

This should (?) get us only our modified files.

* Diff to common ancestor.

* Terrible no-good commit introducing linter error.

* Point to locally fetched master.

* debug logging

* HEAD missing?

* Fetch full branches.

* Fetch master again.

* Revert "debug logging"

This reverts commit 6edde819537a162a2fe4010b0e66c82b088ea7a9.

* Revert "Terrible no-good commit introducing linter error."

This reverts commit ac41b830342a21a1711a9d77a11777276602fb96.

PiperOrigin-RevId: 365085378

PiperOrigin-RevId: 364952108

PiperOrigin-RevId: 364924351

PiperOrigin-RevId: 364895608

PiperOrigin-RevId: 364736401

PiperOrigin-RevId: 364734183

PiperOrigin-RevId: 364731599

PiperOrigin-RevId: 364726354

PiperOrigin-RevId: 364722225

PiperOrigin-RevId: 364722131

PiperOrigin-RevId: 364671967

PiperOrigin-RevId: 364634581

PiperOrigin-RevId: 364625895

PiperOrigin-RevId: 364620438

Updated the single instance postprocessing logics such that it uses the score to distance ratio scoring method. The new logics also avoid using expensive ops, e.g. reduce_max/maximum.

PiperOrigin-RevId: 364607976

PiperOrigin-RevId: 364488384

PiperOrigin-RevId: 364483380

PiperOrigin-RevId: 364458994

PiperOrigin-RevId: 364443016

PiperOrigin-RevId: 364433023

PiperOrigin-RevId: 364395434

PiperOrigin-RevId: 364388006

PiperOrigin-RevId: 364378436

PiperOrigin-RevId: 364363163

In current implementation of StandardEvaluator, the iterator of evaluate dataset will go back (actually create a new iterator) to the beginning when evaluate() is called every time.

In some case, the iterator need to go ahead forever. So add the new option "recreate_iterator_for_each_eval" to control this behavior.

In some case, the iterator creation is time consuming. For example, careting an iterator of distributed dataset for multi worker in eager mode. So it's better to make a repeat dataset instead of recreating an iterator.

If "recreate_iterator_for_each_eval" is True, then the iterator will go back; Otherwise, keep going ahead.

PiperOrigin-RevId: 364356169