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

Commits

Commits on Feb 13, 2019

Fixing #405

ranjita-naik
authored and
ebrevdo
committed
Update NMT codebase to be python3-friendly.
ebrevdo
committed
Update min tf version to 1.12.0

lmthang
authored and
ebrevdo
committed
fix wrong spell

yangyaofei
authored and
ebrevdo
committed

Commits on Feb 12, 2019

Fix the seeding for `Dataset.shuffle(..., reshuffle_each_iteration=False)`.

mrry
authored and
ebrevdo
committed
Support coverage penalty for beam search decoder.

Anonymous
authored and
ebrevdo
committed

Commits on Aug 25, 2018

Add INFERENCE_KEYS and make extend_hparams() in nmt.py more robust
lmthang
committed
Minor improvements and fixes
lmthang
committed
Use LooseVersion for version check

Anonymous
authored and
lmthang
committed
Decouple model loading from inference code
lmthang
committed
[NMT] Adding support for sharded train sets.

eli7
authored and
lmthang
committed
Fix typo.

Anonymous
authored and
lmthang
committed
Remove unused standard hparams.

oahziur
authored and
lmthang
committed
Remove hparams.num_layers
lmthang
committed
Pretty print hparams when writing file.

Anonymous
authored and
lmthang
committed
Refactored *model.py files.
lmthang
committed
Add char-level embeddings for encoder only.

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

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

Anonymous
authored and
lmthang
committed
Allow embedding file to have some misformated entry. Simply ignore the entry

oahziur
authored and
lmthang
committed
Update vocab_utils.py to load embedding files under word2vec format.
lmthang
committed
Minor clean-ups, update docstring
lmthang
committed
Refactoring Model._build_encoder.

eli7
authored and
lmthang
committed
Refactoring internal and external eval to allow injection of placeholders tensors.

eli7
authored and
lmthang
committed
Add copyright text for model_helper.py
lmthang
committed
Update standard hparams as num_layers is no longer a valid hparam.

oahziur
authored and
lmthang
committed
Add sampled_softmax_loss and minor cleanup.

oahziur
authored and
lmthang
committed
Replace compute_encoder_states with build_encoder_states (no sess.run).
lmthang
committed
Add infer_mode option to specify which type of decoder to use during inference.

oahziur
authored and
lmthang
committed
Add language_model flag to train a language model by ignoring the encoder.
lmthang
committed
Set self.encoder_state_list in build_encoder() for gnmt_model.py
lmthang
committed
Make compute_encoder_states() in model.py more general
lmthang
committed
Internal change only
lmthang
committed
Add add_info_summaries to automatically add summaries from info dict.
lmthang
committed
NMT: Improving GPU availability debugging.

eli7
authored and
lmthang
committed