Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Feb 26, 2019

• Follow up to cr/235549624: Set cluster fields from flags first.

  Show description for 37b411b
  

  lingvo-bot

  authored and

  copybara-github

  committed
  37b411b

  Copy full SHA for 37b411b

• Update Lingvo paper link.

  Show description for b28a1ac
  

  Jonathan Shen

  authored and

  copybara-github

  committed
  b28a1ac

  Copy full SHA for b28a1ac

• Merge pull request #19 from ghchinoy:master

  Show description for 370c6bd
  copybara-github

  committed
  370c6bd

  Copy full SHA for 370c6bd

Commits on Feb 25, 2019

• Updates first bash script to reference LINGVO_DIR

  ghchinoy

  committed
  ea68b69

  Copy full SHA for ea68b69

• Update cluster job name before fetching dataset, so that dataset

  Show description for 1322ffd
  

  lingvo-bot

  authored and

  copybara-github

  committed
  1322ffd

  Copy full SHA for 1322ffd

• Export Send/Recv ops.

  Show description for ee95161
  

  bignamehyp

  authored and

  copybara-github

  committed
  ee95161

  Copy full SHA for ee95161

Commits on Feb 23, 2019

• Set random seeds for sampled softmax unittest.

  Show description for 7273eb5
  

  Jonathan Shen

  authored and

  copybara-github

  committed
  7273eb5

  Copy full SHA for 7273eb5

Commits on Feb 22, 2019

• Add Kaiming / He and UniformPositive initializations to WeightInit.

  Show description for 376f8d7
  

  lingvo-bot

  authored and

  copybara-github

  committed
  376f8d7

  Copy full SHA for 376f8d7

• Add LingvoPaper.pdf.

  Show description for 9eae00c
  

  Jonathan Shen

  authored and

  copybara-github

  committed
  9eae00c

  Copy full SHA for 9eae00c

• Fix accidental automatic line wrapping in codelab.

  Show description for 57f4fb4
  

  Jonathan Shen

  authored and

  copybara-github

  committed
  57f4fb4

  Copy full SHA for 57f4fb4

• Update paper link in codelab.

  Show description for 4fb51a1
  

  Jonathan Shen

  authored and

  copybara-github

  committed
  4fb51a1

  Copy full SHA for 4fb51a1

• Add Lingvo paper.

  jonathanasdf

  authored
  8977731

  Copy full SHA for 8977731

• Fix InferenceGraphExporter in exporting frozen graph.

  Show description for 8940b19
  

  protoget

  authored and

  copybara-github

  committed
  8940b19

  Copy full SHA for 8940b19

• Adding missing dtypes for Transformer decoder.

  Show description for 07f8bfc
  

  Mia Chen

  authored and

  copybara-github

  committed
  07f8bfc

  Copy full SHA for 07f8bfc

Commits on Feb 21, 2019

• Add LayerNormalization to SRU.

  Show description for 154556a
  

  lingvo-bot

  authored and

  copybara-github

  committed
  154556a

  Copy full SHA for 154556a

• Update Lingvo Codelab.

  Show description for 4f3ebf9
  

  Jonathan Shen

  authored and

  copybara-github

  committed
  4f3ebf9

  Copy full SHA for 4f3ebf9

Commits on Feb 20, 2019

• Better shape inference in SpeechOnlineEncoderV1

  Show description for bf12638
  

  protoget

  authored and

  copybara-github

  committed
  bf12638

  Copy full SHA for bf12638

• Use int32 seeds on TPU.

  Show description for c9842ae
  

  Jonathan Shen

  authored and

  copybara-github

  committed
  c9842ae

  Copy full SHA for c9842ae

Commits on Feb 19, 2019

• Minor changes to predictor.

  Show description for eb3bcd2
  

  protoget

  authored and

  copybara-github

  committed
  eb3bcd2

  Copy full SHA for eb3bcd2

• Fix FRNNWithAttention bug with use_zero_atten_state and a NestedMap attention state.

  Show description for 88e19f0
  

  Colin Cherry

  authored and

  copybara-github

  committed
  88e19f0

  Copy full SHA for 88e19f0

Commits on Feb 18, 2019

• * Adds NestedMap.FilterKeyVal.

  Show description for 35abf0d
  

  Zhifeng Chen

  authored and

  copybara-github

  committed
  35abf0d

  Copy full SHA for 35abf0d

Commits on Feb 15, 2019

• Split py_binary into py_binary and py_library to avoid having py_binary in deps.

  Show description for 89f095b
  

  lingvo-bot

  authored and

  copybara-github

  committed
  89f095b

  Copy full SHA for 89f095b

• Corrects the comment about in_done_hyps for TopKTerminatedHyps to make it clear that the shape is [t, h*b] rather than [t, b*h] and in_done_hyps[t, i * num_beams + j] represents the i-th hypothesis…

  Show description for 60fff8d
  

  Ruoming Pang

  authored and

  copybara-github

  committed
  60fff8d

  Copy full SHA for 60fff8d

Commits on Feb 13, 2019

• Only create lr_schedule and optimizer for trainer jobs.

  Show description for 66c76b5
  

  Jonathan Shen

  authored and

  copybara-github

  committed
  66c76b5

  Copy full SHA for 66c76b5

• Adds DoubleProjectionLSTMCell, which supports both input and output projections.

  Show description for a4972c8
  

  Ruoming Pang

  authored and

  copybara-github

  committed
  a4972c8

  Copy full SHA for a4972c8

• Fixes the behavior of the tf.data based input generator when the batch size is

  Show description for 004235c
  

  lnzsy

  authored and

  copybara-github

  committed
  004235c

  Copy full SHA for 004235c

Commits on Feb 12, 2019

• Make punctuator model_test size = large.

  Show description for 452c5a3
  

  Stella Laurenzo

  authored and

  copybara-github

  committed
  452c5a3

  Copy full SHA for 452c5a3

• Factor out metrics computation into helper method so child classes can easily override logic of Decode while keeping same metrics computation.

  Show description for ce70d47
  

  Anjuli Kannan

  authored and

  copybara-github

  committed
  ce70d47

  Copy full SHA for ce70d47

• Automated rollback of commit 53148b30799e3fdad2b217dfbc755919bdc36c8a

  Show description for 849fde8
  

  freeblee

  authored and

  copybara-github

  committed
  849fde8

  Copy full SHA for 849fde8

Commits on Feb 11, 2019

• * Makes ConvSoftmax handle both 2-D and 3-D inputs.

  Show description for ecc7ebd
  

  Ruoming Pang

  authored and

  copybara-github

  committed
  ecc7ebd

  Copy full SHA for ecc7ebd

• Minor changes to predictor.

  Show description for 53148b3
  

  protoget

  authored and

  copybara-github

  committed
  53148b3

  Copy full SHA for 53148b3

• Adds a variant of Xavier initialization that uses geometric mean, instead of arithmetic mean, of fan_in and fan_out.

  Show description for b983d24
  

  Ruoming Pang

  authored and

  copybara-github

  committed
  b983d24

  Copy full SHA for b983d24

• Rework tf.data based input generator to have a distinct name and take the dataset class directly.

  Show description for 134e9fd
  

  Stella Laurenzo

  authored and

  copybara-github

  committed
  134e9fd

  Copy full SHA for 134e9fd

• Recover attention probability's static shape.

  Show description for f2feaa7
  

  Colin Cherry

  authored and

  copybara-github

  committed
  f2feaa7

  Copy full SHA for f2feaa7

Commits on Feb 9, 2019

• Add more detailed instructions on how to get started using Docker.

  Show description for fd3bc4a
  

  Jonathan Shen

  authored and

  copybara-github

  committed
  fd3bc4a

  Copy full SHA for fd3bc4a

Pagination

Previous

Next