Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Jan 16, 2020

• Fix initial value of steps per second when continuing a training (#594)

  guillaumekln

  authored
  26b6dec

  Copy full SHA for 26b6dec

• Bump version to 2.5.0

  guillaumekln

  committed
  fa6a6ce

  Copy full SHA for fa6a6ce

• Update CHANGELOG.md

  guillaumekln

  authored
  58ed169

  Copy full SHA for 58ed169

• Cleanup gradient accumulator using the ON_READ synchronization policy (#593)

  guillaumekln

  authored
  dcf3a70

  Copy full SHA for dcf3a70

• Define base Trainer class (#592)

  guillaumekln

  authored
  d7db4b1

  Copy full SHA for d7db4b1

Commits on Jan 14, 2020

• Do not synchronize accumulation steps counter (#590)

  guillaumekln

  authored
  e383c0d

  Copy full SHA for e383c0d

Commits on Jan 13, 2020

• Add missing import

  guillaumekln

  committed
  6abfe18

  Copy full SHA for 6abfe18

• Use non experimental symbols for devices configuration (#589)

  guillaumekln

  authored
  5fa8dfa

  Copy full SHA for 5fa8dfa

Commits on Jan 11, 2020

• Update CHANGELOG.md

  guillaumekln

  authored
  ec711eb

  Copy full SHA for ec711eb

• Clarify that the new tensorflow package has GPU support

  guillaumekln

  authored
  34b800b

  Copy full SHA for 34b800b

Commits on Jan 10, 2020

• Fix small typo in changelog

  guillaumekln

  committed
  7fa1d47

  Copy full SHA for 7fa1d47

• Update changelog

  guillaumekln

  committed
  b1c2be7

  Copy full SHA for b1c2be7

• Export vocabularies as assets by default (#587)

  Show description for d786ddd
  guillaumekln

  authored
  d786ddd

  Copy full SHA for d786ddd

• Update to TensorFlow 2.1 and TensorFlow Addons 0.7 (#586)

  Show description for 26e9a78
  guillaumekln

  authored
  26e9a78

  Copy full SHA for 26e9a78

• Inline function_on_next to improve clarity

  guillaumekln

  committed
  1743778

  Copy full SHA for 1743778

• Add a generator for gradient accumulation

  guillaumekln

  committed
  3d267a1

  Copy full SHA for 3d267a1

• Synchronize words counters in tf.function

  guillaumekln

  committed
  6f7f0ab

  Copy full SHA for 6f7f0ab

• Synchronize words counters across replicas only on read (#585)

  guillaumekln

  authored
  28daea1

  Copy full SHA for 28daea1

• Make dataset cardinality a multiple of the number of replicas (#584)

  Show description for a039a9a
  guillaumekln

  authored
  a039a9a

  Copy full SHA for a039a9a

• Fix reporting loss bug (#583)

  

  nilboy

  authored and

  guillaumekln

  committed
  8abffa5

  Copy full SHA for 8abffa5

Commits on Jan 9, 2020

• Update links to opennmt.net

  guillaumekln

  committed
  35db24b

  Copy full SHA for 35db24b

Commits on Dec 18, 2019

• Do not use device_map to get the replica variable (#578)

  Show description for 886a56c
  guillaumekln

  authored
  886a56c

  Copy full SHA for 886a56c

• Remove unused utility function

  guillaumekln

  committed
  9c807cf

  Copy full SHA for 9c807cf

Commits on Dec 17, 2019

• as_numpy_dtype is a property not a function

  guillaumekln

  authored
  03ce204

  Copy full SHA for 03ce204

Commits on Dec 16, 2019

• Pass Numpy dtype instead of TensorFlow dtype to Numpy function (#577)

  guillaumekln

  authored
  56a0298

  Copy full SHA for 56a0298

• Remove Python 2 guards before using sacreBLEU (#575)

  guillaumekln

  authored
  ea09d0f

  Copy full SHA for ea09d0f

Commits on Dec 13, 2019

• Cleanup some codes related to Python 2 support (#574)

  guillaumekln

  authored
  5809c29

  Copy full SHA for 5809c29

Commits on Dec 10, 2019

• Bump version to 2.4.0

  guillaumekln

  committed
  4869ddf

  Copy full SHA for 4869ddf

• Update changelog

  guillaumekln

  committed
  9d4e394

  Copy full SHA for 9d4e394

• Fixes for vocabulary update (#571)

  Show description for 2035840
  guillaumekln

  authored
  2035840

  Copy full SHA for 2035840

Commits on Dec 8, 2019

• Convert numpy array to Python list before running detokenization (#569)

  guillaumekln

  authored
  f844f24

  Copy full SHA for f844f24

Commits on Dec 4, 2019

• Refactor trainer functions for clarity (#566)

  guillaumekln

  authored
  4ce448a

  Copy full SHA for 4ce448a

• Export relative Transformers in models module (#565)

  guillaumekln

  authored
  d652764

  Copy full SHA for d652764

• Fix retrieval of variable from object-based name in checkpoint averaging (#564)

  Show description for 8fb3df8
  guillaumekln

  authored
  8fb3df8

  Copy full SHA for 8fb3df8

• Add relative position representations (#563)

  guillaumekln

  authored
  7ef62c8

  Copy full SHA for 7ef62c8

Pagination

Previous

Next