This repository was archived by the owner on Jan 21, 2025. It is now read-only.

Commits

History for

Commits on Nov 17, 2023

Migrate references and remove legacy target tpu:tpu_estimator.

jam14j
authored and
Mesh TensorFlow Team
committed

Commits on Mar 25, 2022

Explicitly import estimator from tensorflow as a separate import instead of accessing it via tf.estimator and depend on the tensorflow estimator target.

qlzh727
authored and
Mesh TensorFlow Team
committed

Commits on Mar 24, 2022

Remove unused comments related to Python 2 compatibility.

yilei
authored and
Mesh TensorFlow Team
committed

Commits on Jan 19, 2022

fix typo in logging statement.
Mesh TensorFlow Team
committed

Commits on Mar 5, 2021

[Mesh-TF] Add is_training as an arg to mtf.dropout

afrozenator
authored and
Mesh TensorFlow Team
committed

Commits on Jan 5, 2021

Fix decoding bug for Funnel Transformer

hwchung27
authored and
Mesh TensorFlow Team
committed

Commits on Jun 5, 2020

enable non-tpu training of mtf-bert

nshazeer
authored and
Mesh TensorFlow Team
committed

Commits on Feb 24, 2020

Internal change
Mesh TensorFlow Team
committed
Internal change
Mesh TensorFlow Team
committed
Internal change
Mesh TensorFlow Team
committed

Commits on Feb 7, 2020

1. Add a flag to differentiate train and eval sets. Although we will still use the same set.

Mesh TensorFlow Team
authored and
copybara-github
committed

Commits on Feb 5, 2020

internal

toponado-zz
authored and
copybara-github
committed

Commits on Feb 4, 2020

internal

toponado-zz
authored and
copybara-github
committed

Commits on Feb 3, 2020

Internal

toponado-zz
authored and
copybara-github
committed

Commits on Jan 27, 2020

Dump inputs to another file during eval.

sharannarang
authored and
copybara-github
committed

Commits on Nov 19, 2019

move tpu test into internal folder

toponado-zz
authored and
copybara-github
committed
Fix build by removing google_type_annotations import.

adarob
authored and
copybara-github
committed

Commits on Nov 8, 2019

Enable custom kernel initializers in mtf.layers.dense that depend on reduced_dims and new_dims.

nshazeer
authored and
copybara-github
committed

Commits on Nov 6, 2019

Enable lots of different options in mtf_bert.

nshazeer
authored and
copybara-github
committed
Add z_loss to BERT pre-training.

toponado-zz
authored and
copybara-github
committed

Commits on Nov 4, 2019

Speed up training by customized ring constructor.

toponado-zz
authored and
copybara-github
committed

Commits on Nov 3, 2019

Opensource MeshTensorFlow BERT.

toponado-zz
authored and
copybara-github
committed

Loading