Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on May 27, 2022

• Remove support for older TensorFlow versions 2.4 & 2.5 (#946)

  Show description for 60c89fb
  guillaumekln

  authored
  60c89fb

  Copy full SHA for 60c89fb

Commits on May 25, 2022

• Set a minimum elapsed time of 1 ms for logging metrics (#947)

  guillaumekln

  authored
  0feba73

  Copy full SHA for 0feba73

Commits on May 24, 2022

• Integrate the new CTranslate2 converter for OpenNMT-tf (#945)

  guillaumekln

  authored
  c23a96d

  Copy full SHA for c23a96d

Commits on May 23, 2022

• Support TensorFlow 2.9 (#944)

  guillaumekln

  authored
  e42c1f8

  Copy full SHA for e42c1f8

Commits on May 16, 2022

• Fix training batch size documentation

  guillaumekln

  committed
  767d674

  Copy full SHA for 767d674

Commits on May 2, 2022

• Fix property formatting in docs

  guillaumekln

  committed
  75801bd

  Copy full SHA for 75801bd

Commits on Apr 11, 2022

• Update docs requirements (#940)

  Show description for 5cfe9da
  guillaumekln

  authored
  5cfe9da

  Copy full SHA for 5cfe9da

• Rollback to tf.nn.embedding_lookup in non TensorFlow Lite mode (#942)

  guillaumekln

  authored
  d9838ab

  Copy full SHA for d9838ab

Commits on Mar 31, 2022

• Bump version to 2.26.1

  guillaumekln

  committed
  63c7248

  Copy full SHA for 63c7248

• Fix documentation build (#938)

  guillaumekln

  authored
  f4818c9

  Copy full SHA for f4818c9

• Bump version to 2.26.0

  guillaumekln

  committed
  579926d

  Copy full SHA for 579926d

Commits on Mar 16, 2022

• Clarify the inference buffering log (#935)

  guillaumekln

  authored
  a4c1da9

  Copy full SHA for a4c1da9

• Enable CTranslate2 conversion for models with GELU or Swish activations (#936)

  guillaumekln

  authored
  4f82de6

  Copy full SHA for 4f82de6

• Fix WordNoiser on inputs with ranks > 2 (#934)

  guillaumekln

  authored
  4b2f010

  Copy full SHA for 4b2f010

Commits on Mar 9, 2022

• Add learning rate schedule InvSqrtDecay (#929)

  Show description for 94914ed
  guillaumekln

  authored
  94914ed

  Copy full SHA for 94914ed

Commits on Mar 3, 2022

• Use PyPI API token for authentification

  guillaumekln

  committed
  1fb6839

  Copy full SHA for 1fb6839

Commits on Feb 21, 2022

• Bump version to 2.25.0

  guillaumekln

  committed
  55b16c3

  Copy full SHA for 55b16c3

• Support TensorFlow 2.8 (#924)

  Show description for a0e570f
  guillaumekln

  authored
  a0e570f

  Copy full SHA for a0e570f

Commits on Feb 8, 2022

• Allow setting all learning rate schedules parameters (#925)

  guillaumekln

  authored
  bfa5b07

  Copy full SHA for bfa5b07

Commits on Feb 2, 2022

• Add missing word in docstring

  guillaumekln

  committed
  3661220

  Copy full SHA for 3661220

Commits on Feb 1, 2022

• Add flag to simplify continuing the training in another model directory (#922)

  guillaumekln

  authored
  38263bf

  Copy full SHA for 38263bf

Commits on Jan 31, 2022

• Fix unknowns replacement when the source has BOS or EOS tokens (#920)

  Show description for eb2dd7a
  guillaumekln

  authored
  eb2dd7a

  Copy full SHA for eb2dd7a

• Handle multi source length constraints in base Transformer config (#919)

  guillaumekln

  authored
  36d7593

  Copy full SHA for 36d7593

• Update black to stable release (#921)

  guillaumekln

  authored
  991eda5

  Copy full SHA for 991eda5

Commits on Dec 17, 2021

• Bump version to 2.24.0

  guillaumekln

  committed
  59a4dfd

  Copy full SHA for 59a4dfd

• Declare alignment layers and heads when converting to CTranslate2 (#914)

  guillaumekln

  authored
  5650492

  Copy full SHA for 5650492

• Add option to mask high loss values considered as outliers (#907)

  Show description for c6805d0
  guillaumekln

  authored
  c6805d0

  Copy full SHA for c6805d0

• Replace custom implementation with tf.nn.gelu (#913)

  Show description for 2ea436a
  guillaumekln

  authored
  2ea436a

  Copy full SHA for 2ea436a

• Fix minor typo in docstring

  guillaumekln

  committed
  6aadc73

  Copy full SHA for 6aadc73

• Automatically pad weights to enable Tensor Cores (#911)

  Show description for 8cb7214
  guillaumekln

  authored
  8cb7214

  Copy full SHA for 8cb7214

Commits on Dec 15, 2021

• Warn about duplicate tokens in loaded vocabulary (#910)

  guillaumekln

  authored
  1165074

  Copy full SHA for 1165074

Commits on Dec 14, 2021

• Allow layers to detect TensorFlow Lite tracing (#909)

  guillaumekln

  authored
  40897c7

  Copy full SHA for 40897c7

• Simplify loading of lines in file

  guillaumekln

  committed
  28ac29f

  Copy full SHA for 28ac29f

• Only strip newline characters when loading lines for sacreBLEU

  guillaumekln

  committed
  71cea4f

  Copy full SHA for 71cea4f

• Check dataset sizes before the training (#908)

  guillaumekln

  authored
  3d98f76

  Copy full SHA for 3d98f76

Pagination

Previous

Next