a55c4cf | Lukasz Kaiser | 27 July 2017, 19:26:30 UTC | Merge pull request #185 from rsepassi/push v1.1.2 | 27 July 2017, 19:26:30 UTC |
36766d8 | Ryan Sepassi | 27 July 2017, 18:49:11 UTC | internal-external fixes and enable tests PiperOrigin-RevId: 163370562 | 27 July 2017, 19:05:52 UTC |
175a125 | Ryan Sepassi | 27 July 2017, 02:19:52 UTC | v1.1.2 PiperOrigin-RevId: 163290663 | 27 July 2017, 19:03:56 UTC |
93b325f | Ryan Sepassi | 27 July 2017, 01:18:04 UTC | Baseline model for GeneExpression problem PiperOrigin-RevId: 163286026 | 27 July 2017, 19:03:53 UTC |
5242ac6 | Ryan Sepassi | 27 July 2017, 00:33:54 UTC | Rm num_shards from Problem. Problems specify sharding themselves. PiperOrigin-RevId: 163281576 | 27 July 2017, 19:03:50 UTC |
c01617e | T2T Team | 26 July 2017, 22:01:19 UTC | Use TensorFlow idiom for importing six.moves.xrange. PiperOrigin-RevId: 163261434 | 27 July 2017, 19:03:47 UTC |
cff9f43 | T2T Team | 26 July 2017, 20:52:47 UTC | Allow building a subword vocab from a word vocab file and add tests. PiperOrigin-RevId: 163250427 | 27 July 2017, 19:03:44 UTC |
28eb48f | Ryan Sepassi | 26 July 2017, 19:42:01 UTC | Limit number of concurrent processes in GeneExpressionProblem PiperOrigin-RevId: 163241281 | 27 July 2017, 19:03:41 UTC |
92101af | Niki Parmar | 26 July 2017, 16:58:41 UTC | Bug fix, specify axis for squeeze when computing BLEU score PiperOrigin-RevId: 163219501 | 27 July 2017, 19:03:38 UTC |
b9fcd66 | Lukasz Kaiser | 26 July 2017, 01:12:16 UTC | Back to wmt16 on one set not downloadable from wmt17, internal merges. PiperOrigin-RevId: 163152415 | 27 July 2017, 19:03:34 UTC |
ca08ad9 | Lukasz Kaiser | 26 July 2017, 00:26:22 UTC | Un-reorder of timing signals to make trained models work. PiperOrigin-RevId: 163147659 | 27 July 2017, 19:03:32 UTC |
fd1a87d | Ryan Sepassi | 26 July 2017, 00:06:12 UTC | Problem.eval_metrics PiperOrigin-RevId: 163145397 | 27 July 2017, 19:03:27 UTC |
7de63bd | Noam Shazeer | 25 July 2017, 22:58:10 UTC | Character-level version of lm1b. PiperOrigin-RevId: 163136520 | 27 July 2017, 19:03:24 UTC |
d190b79 | Noam Shazeer | 25 July 2017, 22:20:01 UTC | Update inspect.py to allow decoding with TokenTextEncoder and ByteTextEncoder. PiperOrigin-RevId: 163131045 | 27 July 2017, 19:03:21 UTC |
83a757d | Ashish Vaswani | 25 July 2017, 19:04:03 UTC | Adding encoder conv attention. A query block attends to a neighborhood to the left and the right of it. Pair programmed (Ashish + Niki) PiperOrigin-RevId: 163103460 | 27 July 2017, 19:03:17 UTC |
bea4993 | T2T Team | 25 July 2017, 16:05:13 UTC | Clean up some code around escaping/unescaping tokens and add tests. PiperOrigin-RevId: 163077617 | 27 July 2017, 19:03:14 UTC |
315647c | Lukasz Kaiser | 25 July 2017, 01:51:17 UTC | Download newer WMT dev set. PiperOrigin-RevId: 163020223 | 27 July 2017, 19:03:11 UTC |
7a3c35d | Ryan Sepassi | 25 July 2017, 01:03:03 UTC | GeneExpression Problem, RealModality, and Problem.preprocessing PiperOrigin-RevId: 163016460 | 27 July 2017, 19:03:08 UTC |
c422b98 | Ryan Sepassi | 24 July 2017, 16:27:44 UTC | Add task_id to Problem for possibly distributed data gen PiperOrigin-RevId: 162946584 | 27 July 2017, 19:03:04 UTC |
e43ce96 | T2T Team | 24 July 2017, 16:27:21 UTC | Set `allow_defun` to False, allowing export to tf.SavedModel PiperOrigin-RevId: 162946551 | 27 July 2017, 19:03:00 UTC |
8f624db | T2T Team | 21 July 2017, 22:15:04 UTC | Don't repeatedly concatenate strings in a loop. PiperOrigin-RevId: 162791277 | 27 July 2017, 19:02:56 UTC |
13b02ca | T2T Team | 21 July 2017, 19:30:48 UTC | Refactoring of get_or_generate_vocab* functions. PiperOrigin-RevId: 162771691 | 27 July 2017, 19:02:51 UTC |
62a0ee7 | Lukasz Kaiser | 25 July 2017, 00:50:53 UTC | Merge pull request #177 from cshanbo/pos_emb fix positional embedding | 25 July 2017, 00:50:53 UTC |
8cc9a5e | Lukasz Kaiser | 25 July 2017, 00:48:37 UTC | Merge pull request #179 from stefan-it/remove-input-sentence-from-decoded-file Remove input sentence from decoded file | 25 July 2017, 00:48:37 UTC |
3913c1e | Lukasz Kaiser | 25 July 2017, 00:37:46 UTC | Merge pull request #181 from martinpopel/add_czech2 Simplify adding new translation language pairs | 25 July 2017, 00:37:46 UTC |
f793de2 | Martin Popel | 24 July 2017, 21:58:55 UTC | Simplify adding new translation language pairs * simplify wmt.py: `train_generator` is now a method (not a property returning a function) and WMTProblem subclasses implement it directly (not via one-purpose functions) * add WMT English-Czech translation problems * introduce `vocab_name` property, so e.g. `tokens.vocab-en-cs.32768` is used instead of `tokens.vocab.32768` if only English+Czech datasets were used | 24 July 2017, 21:58:55 UTC |
5418429 | Stefan Schweter | 22 July 2017, 22:13:39 UTC | utils: input sentence is no longer shown in decoded file (correct bleu calculation script for en-de) | 22 July 2017, 22:23:22 UTC |
422aac7 | Stefan Schweter | 22 July 2017, 22:11:29 UTC | trainer-utils: do not show input sentence in decoded file | 22 July 2017, 22:11:29 UTC |
ac038c5 | Shanbo Cheng | 20 July 2017, 06:19:24 UTC | fix positional embedding | 20 July 2017, 06:19:24 UTC |
668e385 | Lukasz Kaiser | 20 July 2017, 03:09:46 UTC | Merge pull request #172 from rsepassi/push v1.1.1 | 20 July 2017, 03:09:46 UTC |
2fd79ec | Ryan Sepassi | 20 July 2017, 03:01:18 UTC | Update readme and make genetics module optional PiperOrigin-RevId: 162570620 | 20 July 2017, 03:02:25 UTC |
a7339cd | Ryan Sepassi | 20 July 2017, 02:36:10 UTC | v1.1.1 PiperOrigin-RevId: 162569525 | 20 July 2017, 03:02:20 UTC |
60dd5e0 | Ryan Sepassi | 20 July 2017, 02:35:47 UTC | Add tests for genetics problems PiperOrigin-RevId: 162569505 | 20 July 2017, 03:02:16 UTC |
84445cc | Lukasz Kaiser | 20 July 2017, 02:31:47 UTC | Change the format of generated vocab files to include languages, put them in data_dir, add --generate_data option. PiperOrigin-RevId: 162569315 | 20 July 2017, 03:02:10 UTC |
293b5f6 | Ryan Sepassi | 19 July 2017, 23:36:48 UTC | Add genetics dataset - data generation only PiperOrigin-RevId: 162553677 | 20 July 2017, 03:02:05 UTC |
78acdb4 | T2T Team | 19 July 2017, 22:13:13 UTC | Fix a bug in text_encoder. "self._UNESCAPE_REGEX -> _UNESCAPE_REGEX" PiperOrigin-RevId: 162542600 | 20 July 2017, 03:01:59 UTC |
80dba89 | Niki Parmar | 19 July 2017, 04:52:38 UTC | Add celeba dataset, add to problems PiperOrigin-RevId: 162444538 | 20 July 2017, 03:01:57 UTC |
47d556a | Lukasz Kaiser | 19 July 2017, 17:36:47 UTC | Merge pull request #167 from rsepassi/push v1.1.0 | 19 July 2017, 17:36:47 UTC |
f703629 | T2T Team | 19 July 2017, 00:47:09 UTC | Change summary generation to work better in multi-model case. PiperOrigin-RevId: 162429483 | 19 July 2017, 00:52:04 UTC |
d3502cb | Ryan Sepassi | 18 July 2017, 23:58:57 UTC | v1.1.0 PiperOrigin-RevId: 162424323 | 19 July 2017, 00:52:01 UTC |
4f67b7b | Lukasz Kaiser | 18 July 2017, 23:56:29 UTC | Move algorithmic and WMT problems to Problem class, correct summaries. PiperOrigin-RevId: 162424062 | 19 July 2017, 00:51:58 UTC |
ecacf77 | Ryan Sepassi | 18 July 2017, 19:49:33 UTC | Internal merge #160 PiperOrigin-RevId: 162389811 | 19 July 2017, 00:51:40 UTC |
e1ecf01 | Noam Shazeer | 17 July 2017, 20:56:31 UTC | Create a "long_answer" model for attacking the wikipedia title->article dataset. PiperOrigin-RevId: 162264770 | 19 July 2017, 00:51:36 UTC |
5db92b5 | Ryan Sepassi | 17 July 2017, 19:43:36 UTC | Add t2t_usr_dir functionality to t2t-datagen PiperOrigin-RevId: 162253918 | 19 July 2017, 00:51:34 UTC |
930cb33 | Ryan Sepassi | 17 July 2017, 19:40:41 UTC | Create top-level docs dir for GitHub Pages PiperOrigin-RevId: 162253587 | 19 July 2017, 00:51:32 UTC |
04b4f47 | Ryan Sepassi | 17 July 2017, 18:27:14 UTC | Update documentation for adding new Problems PiperOrigin-RevId: 162242293 | 19 July 2017, 00:51:30 UTC |
963730e | Lukasz Kaiser | 18 July 2017, 19:25:14 UTC | Merge pull request #165 from chiphuyen/master Add 'num_shards" to generate_data | 18 July 2017, 19:25:14 UTC |
0d250b3 | Noé Casas | 18 July 2017, 18:25:19 UTC | Support --t2t_usr_dir also in t2t-datagen (#160) * Refactor user directory loading functionality and use it also from t2t-datagen * Move flag declaration to the binary files | 18 July 2017, 18:25:19 UTC |
c3a59b4 | Huyen Nguyen | 18 July 2017, 18:22:44 UTC | add arguments num_shards to generate_data | 18 July 2017, 18:22:44 UTC |
8c850c9 | Huyen Nguyen | 18 July 2017, 18:20:55 UTC | add arguments num_shards to generate_data | 18 July 2017, 18:20:55 UTC |
9709825 | Huyen Nguyen | 18 July 2017, 18:10:43 UTC | more num_shards | 18 July 2017, 18:10:43 UTC |
042c44c | Huyen Nguyen | 18 July 2017, 17:58:22 UTC | take in account FLAGS.num_shards | 18 July 2017, 17:58:22 UTC |
c91989c | Lukasz Kaiser | 18 July 2017, 03:07:15 UTC | Merge pull request #162 from vthorsteinsson/ice Bug fixes in inference and data generation; faster token unescaping | 18 July 2017, 03:07:15 UTC |
1368b00 | Lukasz Kaiser | 18 July 2017, 02:51:35 UTC | Merge pull request #158 from stefan-it/mken-problem Add new problem: Macedonian to English (SETimes corpus) | 18 July 2017, 02:51:35 UTC |
e2ed8ed | vthorsteinsson | 17 July 2017, 17:04:31 UTC | Bug fixes in generator_utils and trainer_utils | 17 July 2017, 17:04:31 UTC |
d8d379c | vthorsteinsson | 16 July 2017, 16:04:12 UTC | Merge decoding of tokens using regex | 16 July 2017, 16:04:12 UTC |
fb808c3 | vthorsteinsson | 16 July 2017, 16:01:29 UTC | Merge remote-tracking branch 'upstream/master' | 16 July 2017, 16:01:29 UTC |
48997b5 | Stefan Schweter | 15 July 2017, 23:07:19 UTC | Register new problem: setimes_mken_tokens_32k for Macedonian to English translation (with SETimes corpus) | 15 July 2017, 23:07:19 UTC |
f34ab16 | Stefan Schweter | 15 July 2017, 23:05:46 UTC | Add id for Macedonian tokens | 15 July 2017, 23:05:46 UTC |
f7f3346 | Stefan Schweter | 15 July 2017, 23:04:47 UTC | Add data file urls for Macedonian-English | 15 July 2017, 23:04:47 UTC |
0c66117 | Lukasz Kaiser | 14 July 2017, 23:58:31 UTC | Merge pull request #156 from rsepassi/push v1.0.14 | 14 July 2017, 23:58:30 UTC |
c8b7000 | Ryan Sepassi | 14 July 2017, 23:37:02 UTC | Version bump 1.0.14 PiperOrigin-RevId: 162025739 | 14 July 2017, 23:40:37 UTC |
a80273f | T2T Team | 14 July 2017, 23:35:58 UTC | Added checks for making sure the key and value depths are divisible by the number of attention heads when doing multihead attention. PiperOrigin-RevId: 162025615 | 14 July 2017, 23:40:33 UTC |
1b1d7ed | Ryan Sepassi | 14 July 2017, 23:35:49 UTC | Port WMT en-de tokens 8k/32k to new Problem registry PiperOrigin-RevId: 162025600 | 14 July 2017, 23:40:31 UTC |
4b4c800 | Lukasz Kaiser | 14 July 2017, 22:58:38 UTC | Removing obsolete problems, merging. PiperOrigin-RevId: 162020990 | 14 July 2017, 23:40:28 UTC |
c4407b8 | Noam Shazeer | 14 July 2017, 20:36:26 UTC | Some fixes/configs/comments for language modeling plans. PiperOrigin-RevId: 162000426 | 14 July 2017, 23:40:20 UTC |
ca7b045 | Ryan Sepassi | 14 July 2017, 00:47:51 UTC | Problem base class and registry PiperOrigin-RevId: 161892788 | 14 July 2017, 23:40:19 UTC |
1e18474 | Ashish Vaswani | 13 July 2017, 22:57:21 UTC | internal. PiperOrigin-RevId: 161878428 | 14 July 2017, 23:40:15 UTC |
56d65f0 | Ryan Sepassi | 12 July 2017, 23:00:29 UTC | Working open-source distributed training PiperOrigin-RevId: 161731856 | 14 July 2017, 23:40:14 UTC |
29f2e2e | Ryan Sepassi | 12 July 2017, 19:39:04 UTC | Update instructions for user components to be compatible with Py3 PiperOrigin-RevId: 161701617 | 14 July 2017, 23:40:12 UTC |
a2b1c60 | vthorsteinsson | 14 July 2017, 17:32:57 UTC | Used regex in _unescape_token() | 14 July 2017, 17:32:57 UTC |
43bfb9f | Lukasz Kaiser | 14 July 2017, 17:05:34 UTC | Merge pull request #154 from vthorsteinsson/iceparse Source/target pair text files; Icelandic parsing support; fixes | 14 July 2017, 17:05:34 UTC |
5a72e5c | vthorsteinsson | 14 July 2017, 14:22:38 UTC | Adapted to upstream tokenizer change | 14 July 2017, 14:22:38 UTC |
4617c01 | Lukasz Kaiser | 14 July 2017, 01:26:00 UTC | Merge pull request #153 from rshin/shakeshake Add Shake-Shake model | 14 July 2017, 01:26:00 UTC |
2adf3ae | Richard Shin | 14 July 2017, 00:49:45 UTC | Add shakeshake_type hparam: batch, image, equal | 14 July 2017, 00:49:45 UTC |
7bf4936 | vthorsteinsson | 14 July 2017, 00:40:04 UTC | Standardized EOS token | 14 July 2017, 00:40:04 UTC |
27c6185 | vthorsteinsson | 14 July 2017, 00:14:13 UTC | Cleanup in text_encoder.py | 14 July 2017, 00:14:13 UTC |
6d4e7b4 | vthorsteinsson | 13 July 2017, 23:57:17 UTC | Upstream merge | 13 July 2017, 23:57:17 UTC |
9dc2826 | vthorsteinsson | 13 July 2017, 23:51:40 UTC | Iceparser adaptations | 13 July 2017, 23:51:40 UTC |
8287094 | Richard Shin | 13 July 2017, 23:06:21 UTC | Reword comments | 13 July 2017, 23:06:21 UTC |
9f02a51 | Richard Shin | 13 July 2017, 23:04:30 UTC | Fix bugs, add more explanation | 13 July 2017, 23:04:30 UTC |
f44f51f | Richard Shin | 13 July 2017, 20:53:11 UTC | Add shake-shake for CIFAR-10 | 13 July 2017, 21:09:53 UTC |
28e0e4e | Lukasz Kaiser | 13 July 2017, 20:44:30 UTC | Merge pull request #141 from ReDeiPirati/yellowfin_optimizer Integrate YellowFin Optimizer(with test) in T2T | 13 July 2017, 20:44:30 UTC |
7566c4d | Lukasz Kaiser | 13 July 2017, 20:38:00 UTC | Merge pull request #151 from cshanbo/wmt_zhen_translate add wmt_zhen_token_32k | 13 July 2017, 20:38:00 UTC |
9a1f888 | Lukasz Kaiser | 13 July 2017, 20:27:55 UTC | Merge pull request #139 from deasuke/unicode_fix replace unicode in python3 | 13 July 2017, 20:27:55 UTC |
7d3c10b | ReDeiPirati | 12 July 2017, 10:55:53 UTC | Integrate Yellow Optimizer(with test) in T2T(Unfortunately there is a bug in TF which do not ensure a smooth running). | 12 July 2017, 10:55:53 UTC |
89ddffe | Deasuke | 12 July 2017, 07:13:50 UTC | use decode instead of unicode in PY3 | 12 July 2017, 07:13:50 UTC |
b1ad6f0 | Deasuke | 12 July 2017, 06:34:41 UTC | replace unicode with six.text_type | 12 July 2017, 06:34:41 UTC |
235392c | Shanbo Cheng | 12 July 2017, 05:20:41 UTC | add wmt_zhen_token_32k | 12 July 2017, 05:20:41 UTC |
006ecb5 | Lukasz Kaiser | 12 July 2017, 01:34:42 UTC | Merge pull request #136 from rsepassi/push v1.0.13 | 12 July 2017, 01:34:42 UTC |
64defb7 | Ryan Sepassi | 12 July 2017, 01:27:00 UTC | Fix T2T open-source distributed training PiperOrigin-RevId: 161612824 | 12 July 2017, 01:27:28 UTC |
9efa4e2 | Ryan Sepassi | 12 July 2017, 01:23:35 UTC | Fix T2T open-source distributed training PiperOrigin-RevId: 161612632 | 12 July 2017, 01:27:25 UTC |
72f0874 | Lukasz Kaiser | 12 July 2017, 00:40:07 UTC | internal merge PiperOrigin-RevId: 161608262 | 12 July 2017, 01:27:24 UTC |
912daf7 | Noam Shazeer | 11 July 2017, 04:16:38 UTC | Changes to data generators and SubwordTextEncoder ... should not break anything. PiperOrigin-RevId: 161483309 | 12 July 2017, 01:27:22 UTC |
09a6084 | Niki Parmar | 10 July 2017, 22:37:56 UTC | Remove redundant copies of res_fn. Use from common_layers PiperOrigin-RevId: 161451356 | 12 July 2017, 01:27:20 UTC |
afd1565 | Lukasz Kaiser | 10 July 2017, 22:12:28 UTC | merge from github PiperOrigin-RevId: 161447896 | 12 July 2017, 01:27:18 UTC |
d827bb2 | Lukasz Kaiser | 11 July 2017, 23:21:42 UTC | Merge pull request #133 from EndingCredits/master Add new set network layer type and example model | 11 July 2017, 23:21:42 UTC |
bcf9a8b | Ending_Credits | 11 July 2017, 14:17:10 UTC | Added example model | 11 July 2017, 14:17:10 UTC |
3e4b686 | vthorsteinsson | 11 July 2017, 13:27:42 UTC | Merge remote-tracking branch 'upstream/master' | 11 July 2017, 13:27:42 UTC |
30887b8 | vthorsteinsson | 11 July 2017, 13:19:16 UTC | Target string displayed; smaller fixes | 11 July 2017, 13:19:16 UTC |