v1.4.1

TODO(afrozm): Write this up.

v1.4.0

TODO(afrozm): Write this up.

v1.3.9

Some bug fixes.

v1.3.8

TODO(afrozm): Describe

v1.3.7

Lots of documentation, bugs squashed and misc changes.

Features

• Load checkpoint for fine-tuning in 0349c3f by @henrykmichalewski !
• Added generic input pipeline for GLUE tasks in 2f490e8 by @henrykmichalewski - thanks a lot!
• Some nascent support for bfloat16s!

Models

• Performer's Favor and CausalFavor - 77db199 thanks @lukaszkaiser !
• Funnel-Transformer in #1156 thanks a lot @mvxxx !
• BERT for Trax in #1254 , #1223, etc by @piotrekp1 - thanks a lot!
• Residual Exchange Network by @kkanska in 3a8f402 ! Thanks!

PRs Merged

• Fixing broken example links in #1263 thanks @amtagrwl !
• Added WideResnet, Deconv etc Example Notebook in #1259 , #1202, #1232 thanks a lot! @SauravMaheshkar !
• Remove implicit object from the base class in #1228 thanks @HarshCasper !
• Fashion MNIST example in #1199 thanks @jimexist !
• Fix PretrainedBERT init in #1135 by @hepaajan !
• Typo in the TransformerDecoder input parameters description in #1100 by @kujaomega !

v1.3.6

TODO(afrozm): Write something.

v1.3.5

PRs:
Thanks @NathanHowell for fixing a bunch of typos in #962
Thanks @DarrenZhang01 for contributing to the TF-Numpy extensions code in #954

ReversibleSerialTrainer for memory efficient, layer by layer training.

Miscellaneous other issues.

v1.3.4

v1.3.4

v1.3.3

Minor Fixes:

• Rename gumbel_sample to logsoftmax_sample in a2497cb
• A fix to storing checkpoints in Loop in 88b033c

v1.3.2

Framework:

• Data pipeline combinators in 17d4471
• Multi-host training in the Loop api 524321e
• Auto-regressive sampling in fe0fa78
• T2T Tokenizers 6306231
• Early work on multi-task training in 423d664 thanks to @koz4k

Tasks / Gin files:

• IMDB dataset in 08bdb50
• Machine Translation en-pl in 5fb8aa8