This repository was archived by the owner on Dec 11, 2023. It is now read-only.

Added num_encoder_layers/num_decoder_layers to WMT16 standard hparams.#269

Closed

tiberiu92 wants to merge 39 commits intotensorflow:masterfrom

tiberiu92:master

Commits on Mar 12, 2018

Added num_encoder_layers/num_decoder_layers to WMT16 standard hparams.
tiberiuscarlat
committed

Commits on Aug 25, 2018

Add TrainOutputTuple, EvalOutputTuple, InferOutputTuple
lmthang
committed
Wrap try/except for saver.restore and print variables in loaded checkpoints
lmthang
committed
Improve several behaviors regarding loading hparams for train/inference:
lmthang
committed
pretrained our models on

Anonymous
authored and
lmthang
committed
Factor out model creation code in train.py and inference.py.
lmthang
committed
Standardize vocab in test to use <unk>, <s>, </s>
lmthang
committed
Clean up inference_test.py and add more sharing code to _createTestInferCheckpoint();
lmthang
committed
NMT: Improving GPU availability debugging.

eli7
authored and
lmthang
committed
Add add_info_summaries to automatically add summaries from info dict.
lmthang
committed
Internal change only
lmthang
committed
Make compute_encoder_states() in model.py more general
lmthang
committed
Set self.encoder_state_list in build_encoder() for gnmt_model.py
lmthang
committed
Add language_model flag to train a language model by ignoring the encoder.
lmthang
committed
Add infer_mode option to specify which type of decoder to use during inference.

oahziur
authored and
lmthang
committed
Replace compute_encoder_states with build_encoder_states (no sess.run).
lmthang
committed
Add sampled_softmax_loss and minor cleanup.

oahziur
authored and
lmthang
committed
Update standard hparams as num_layers is no longer a valid hparam.

oahziur
authored and
lmthang
committed
Add copyright text for model_helper.py
lmthang
committed
Refactoring internal and external eval to allow injection of placeholders tensors.

eli7
authored and
lmthang
committed
Refactoring Model._build_encoder.

eli7
authored and
lmthang
committed
Minor clean-ups, update docstring
lmthang
committed
Update vocab_utils.py to load embedding files under word2vec format.
lmthang
committed
Allow embedding file to have some misformated entry. Simply ignore the entry

oahziur
authored and
lmthang
committed
Unify reading and writing of hparams file.

Anonymous
authored and
lmthang
committed
Colocate output_layer with last LSTM cell to improve training speed.

oahziur
authored and
lmthang
committed
Add char-level embeddings for encoder only.

oahziur
authored and
lmthang
committed
Refactored *model.py files.
lmthang
committed
Pretty print hparams when writing file.

Anonymous
authored and
lmthang
committed
Remove hparams.num_layers
lmthang
committed
Remove unused standard hparams.

oahziur
authored and
lmthang
committed
Fix typo.

Anonymous
authored and
lmthang
committed
[NMT] Adding support for sharded train sets.

eli7
authored and
lmthang
committed
Decouple model loading from inference code
lmthang
committed
Use LooseVersion for version check

Anonymous
authored and
lmthang
committed
Minor improvements and fixes
lmthang
committed
Add INFERENCE_KEYS and make extend_hparams() in nmt.py more robust
lmthang
committed

Added num_encoder_layers/num_decoder_layers to WMT16 standard hparams.#269tiberiu92 wants to merge 39 commits intotensorflow:mastertensorflow/nmt:masterfrom tiberiu92:mastertiberiu92/nmt:masterCopy head branch name to clipboard

Commits

Commits on Mar 12, 2018

Commits on Aug 25, 2018

Commits on Oct 17, 2018

Added num_encoder_layers/num_decoder_layers to WMT16 standard hparams.#269
tiberiu92 wants to merge 39 commits intotensorflow:masterfrom
tiberiu92:master