Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Sep 18, 2019

• Add post training loop api override for learners.

  Show description for fcdffc0
  

  rohan-anil

  authored and

  copybara-github

  committed
  fcdffc0

  Copy full SHA for fcdffc0

• Initial support for ExecutorTpu on Cloud TPU.

  Show description for 585edf3
  

  will-cromar

  authored and

  copybara-github

  committed
  585edf3

  Copy full SHA for 585edf3

• Add tf.summary.image to summary_utils.

  Show description for c0b4784
  

  lingvo-bot

  authored and

  copybara-github

  committed
  c0b4784

  Copy full SHA for c0b4784

• Removed calls to the deprecated ops.RegisterShape

  Show description for aac9b37
  

  lingvo-bot

  authored and

  copybara-github

  committed
  aac9b37

  Copy full SHA for aac9b37

• Add README for car directory to show how to reproduce the StarNet

  Show description for b698d5e
  

  Vijay Vasudevan

  authored and

  copybara-github

  committed
  b698d5e

  Copy full SHA for b698d5e

Commits on Sep 17, 2019

• Internal change

  Show description for b936ae8
  

  lingvo-bot

  authored and

  Jonathan Shen

  committed
  b936ae8

  Copy full SHA for b936ae8

• Remove assertion for filter_shapes in conv_layers as it's already asserted when creating layer variable.

  Show description for 37a25aa
  

  anmolgulati

  authored and

  copybara-github

  committed
  37a25aa

  Copy full SHA for 37a25aa

• Add py_utils.IsCompatible()

  Show description for f8dbbd8
  

  protoget

  authored and

  copybara-github

  committed
  f8dbbd8

  Copy full SHA for f8dbbd8

• Adds Optimizer.ComputeGradients(), which defaults to py_utils.ComputeGradient(), so that subclasses can override gradient computation.

  Show description for 7be9373
  

  Ruoming Pang

  authored and

  copybara-github

  committed
  7be9373

  Copy full SHA for 7be9373

Commits on Sep 16, 2019

• Put predictions before input_batch.

  Show description for 9aa4631
  

  Jonathan Shen

  authored and

  copybara-github

  committed
  9aa4631

  Copy full SHA for 9aa4631

• Add "lib" dockerfile that preinstalls lingvo pip and dependencies.

  Show description for a2f3261
  

  Vijay Vasudevan

  authored and

  copybara-github

  committed
  a2f3261

  Copy full SHA for a2f3261

• Adds EmbeddingStep.

  Show description for 411768a
  

  lingvo-bot

  authored and

  copybara-github

  committed
  411768a

  Copy full SHA for 411768a

Commits on Sep 14, 2019

• Consolidate base_model.ComputeLoss and base_decoder.ComputeLoss interfaces.

  Show description for 0e0ba2d
  

  Jonathan Shen

  authored and

  copybara-github

  committed
  0e0ba2d

  Copy full SHA for 0e0ba2d

• Better error message when --model isn't provided to trainer.py main.

  Show description for 3171c28
  

  Vijay Vasudevan

  authored and

  copybara-github

  committed
  3171c28

  Copy full SHA for 3171c28

• Params: Suggest similar key names when returning key mismatch errors.

  Show description for ed9e741
  

  lingvo-bot

  authored and

  copybara-github

  committed
  ed9e741

  Copy full SHA for ed9e741

• More lingvo pip support.

  Show description for 603017f
  

  Vijay Vasudevan

  authored and

  copybara-github

  committed
  603017f

  Copy full SHA for 603017f

• Modify SplitInputBatch, which is typically used when the model is not run on TPU, to also rely on GetPreprocessedInputBatch to get input batch, similar to CreateTPUInputfeed. This makes the prepro…

  Show description for edb5803
  

  lingvo-bot

  authored and

  copybara-github

  committed
  edb5803

  Copy full SHA for edb5803

• Strip 'params.' instead of '.params' in model registry

  Show description for cc63de6
  

  Vijay Vasudevan

  authored and

  copybara-github

  committed
  cc63de6

  Copy full SHA for cc63de6

Commits on Sep 13, 2019

• Add StarnetPed model with sparse input fusion preprocessor.

  Show description for 0740cfc
  

  Vijay Vasudevan

  authored and

  copybara-github

  committed
  0740cfc

  Copy full SHA for 0740cfc

• Split pillar network into contracting and expanding part.

  Show description for 7341c83
  

  lingvo-bot

  authored and

  copybara-github

  committed
  7341c83

  Copy full SHA for 7341c83

• Removes 'bias_c' when enable_ln_on_c=False to match the behavior of LayerNormalizedLSTMCellSimple.

  Show description for a83c94f
  

  Ruoming Pang

  authored and

  copybara-github

  committed
  a83c94f

  Copy full SHA for a83c94f

• Adds input_preprocessors.SparseSampler which uses the fused sampler op.

  Show description for e4e320a
  

  Zhifeng Chen

  authored and

  copybara-github

  committed
  e4e320a

  Copy full SHA for e4e320a

Commits on Sep 12, 2019

• Add bias term to Conv2DLayerWithPadding

  Show description for 4ed6ae6
  

  protoget

  authored and

  copybara-github

  committed
  4ed6ae6

  Copy full SHA for 4ed6ae6

• Adds error metrics to StarNet model for bounding box dimensions and rotation.

  Show description for e59ff64
  

  lingvo-bot

  authored and

  copybara-github

  committed
  e59ff64

  Copy full SHA for e59ff64

• Adds LinearRampupSqrtDecayByBatchSizeAndReplicas schedule.

  Show description for e71b8ed
  

  lingvo-bot

  authored and

  copybara-github

  committed
  e71b8ed

  Copy full SHA for e71b8ed

• make it possible to call FProp when input is None

  Show description for 6a15901
  

  lingvo-bot

  authored and

  copybara-github

  committed
  6a15901

  Copy full SHA for 6a15901

• Internal change.

  Show description for 111c91d
  

  Vijay Vasudevan

  authored and

  copybara-github

  committed
  111c91d

  Copy full SHA for 111c91d

• Allow to leave points_padding blank in PointsToGrid preprocessor.

  Show description for db5d6fb
  

  lingvo-bot

  authored and

  copybara-github

  committed
  db5d6fb

  Copy full SHA for db5d6fb

• Adds RnnStep and RnnStackStep.

  Show description for 57fa285
  

  lingvo-bot

  authored and

  copybara-github

  committed
  57fa285

  Copy full SHA for 57fa285

Commits on Sep 11, 2019

• Let LayerNorm function in SRU cell apply bias to make sure bias is always applied after layer normalization.

  Show description for c7d6337
  

  lingvo-bot

  authored and

  copybara-github

  committed
  c7d6337

  Copy full SHA for c7d6337

• Adds ability to filter based on weather/location/time of day to Metadata extractor.

  Show description for 3a8294a
  

  lingvo-bot

  authored and

  copybara-github

  committed
  3a8294a

  Copy full SHA for 3a8294a

• py_utils.CumSum does not work on TPU for non-float types.

  Show description for 7154e69
  

  Jonathan Shen

  authored and

  copybara-github

  committed
  7154e69

  Copy full SHA for 7154e69

• Fix py_utils.ReadFileLines to work with Python 2.

  Show description for fe46570
  

  will-cromar

  authored and

  copybara-github

  committed
  fe46570

  Copy full SHA for fe46570

• Implements GraphStep, StackStep, and StatelessLayerStep.

  Show description for aafbf7a
  

  lingvo-bot

  authored and

  copybara-github

  committed
  aafbf7a

  Copy full SHA for aafbf7a

• Add waymo open dataset starnet model params file.

  Show description for 8f3dca6
  

  Vijay Vasudevan

  authored and

  copybara-github

  committed
  8f3dca6

  Copy full SHA for 8f3dca6

Pagination

Previous

Next