This repository was archived by the owner on Jul 7, 2023. It is now read-only.

1.6.2#771

Merged

lukaszkaiser merged 12 commits intotensorflow:masterfrom

lukaszkaiser:push

May 8, 2018

Commits on May 8, 2018

Adding Lambada dataset as a new problem to T2T.

T2T Team
authored and
lukaszkaiser
committed
Enable bfloat16 activation for Image Transformer model.

T2T Team
authored and
lukaszkaiser
committed
Zero-pad speech utterances to bring closer training (with zero padding from batching) and testing conditions

T2T Team
authored and
lukaszkaiser
committed
Creation of ASR with transformer demo notebook.

T2T Team
authored and
lukaszkaiser
committed
Using _slow_greedy_infer and _beam_decode_slow in _greey_infer and _beam_decode for R_transformer models.

T2T Team
authored and
lukaszkaiser
committed
Removed the following from transformer_vae :

Ashish Vaswani
authored and
lukaszkaiser
committed
Internal decoding merge and update Travis

Ryan Sepassi
authored and
lukaszkaiser
committed
Allow for decompress outside do_mask. It seems to be independent of that.

Ashish Vaswani
authored and
lukaszkaiser
committed
Corrections to autoencoders and atari ROM paths.
Lukasz Kaiser
committed
Bump version and link ASR notebook and tutorial.
Lukasz Kaiser
committed
Update base hparams for tpu

Niki Parmar
authored and
lukaszkaiser
committed
internal
Lukasz Kaiser
committed