Rooba
/
spaCy
mirror of https://github.com/explosion/spaCy.git
1
1
Fork 0
Code Issues Packages Projects Releases Wiki Activity
8,880 Commits 70 Branches 177 Tags 298 MiB
Commit Graph

325 Commits

Author SHA1 Message Date
Ines Montani a609a1ca29
Merge pull request #2152 from explosion/feature/tidy-up-dependencies 
💫 Tidy up dependencies
 2018-03-29 14:35:09 +02:00
Matthew Honnibal b5098079d8 Fix error on urllib 2018-03-29 00:08:16 +02:00
Ines Montani 98e9cda677
Merge pull request #2158 from explosion/feature/fix-multiple-vectors (resolves #1660) 
💫 Fix loading of multiple vector models
 2018-03-28 23:08:24 +02:00
Matthew Honnibal 17c3e7efa2 Add message noting vectors 2018-03-28 16:33:43 +02:00
ines 7fbc9e5874 Replace requests with urllib 2018-03-28 12:46:07 +02:00
ines ac88c72c9a Fix ftfy workaround and remove old import 2018-03-28 12:14:28 +02:00
Matthew Honnibal 070b6c6495 Remove dependency on ftfy 2018-03-28 12:07:02 +02:00
Matthew Honnibal b7136cb094 Support zipped vector files in init-model 2018-03-27 21:01:18 +00:00
Matthew Honnibal 1f7229f40f Revert "Merge branch 'develop' of https://github.com/explosion/spaCy into develop" 
This reverts commit c9ba3d3c2d, reversing
changes made to 92c26a35d4.
 2018-03-27 19:23:02 +02:00
Matthew Honnibal f57bfbccdc Fix non-projective label filtering 2018-03-27 13:41:33 +02:00
Matthew Honnibal 8bbd26579c Support GPU in UD training script 2018-03-27 09:53:35 +00:00
Matthew Honnibal 406548b976 Support .gz and .tar.gz files in spacy init-model 2018-03-24 17:18:32 +01:00
Matthew Honnibal 85717f570c Merge branch 'master' of https://github.com/explosion/spaCy 2018-03-23 20:30:42 +01:00
Matthew Honnibal 8902754f0b Fix vector loading for ud_train 2018-03-23 20:30:00 +01:00
Xiaoquan Kong a71b99d7ff bugfix for global-variable-change-in-runtime related issue (#2135) 
* Bugfix: setting pollution from spacy/cli/ud_train.py to whole package

* Add contributor agreement of howl-anderson
 2018-03-23 11:36:38 +01:00
Matthew Honnibal 044397e269 Support .gz and .tar.gz files in spacy init-model 2018-03-21 14:33:23 +01:00
Matthew Honnibal bede11b67c
Improve label management in parser and NER (#2108) 
This patch does a few smallish things that tighten up the training workflow a little, and allow memory use during training to be reduced by letting the GoldCorpus stream data properly.

Previously, the parser and entity recognizer read and saved labels as lists, with extra labels noted separately. Lists were used becaue ordering is very important, to ensure that the label-to-class mapping is stable.

We now manage labels as nested dictionaries, first keyed by the action, and then keyed by the label. Values are frequencies. The trick is, how do we save new labels? We need to make sure we iterate over these in the same order they're added. Otherwise, we'll get different class IDs, and the model's predictions won't make sense.

To allow stable sorting, we map the new labels to negative values. If we have two new labels, they'll be noted as having "frequency" -1 and -2. The next new label will then have "frequency" -3. When we sort by (frequency, label), we then get a stable sort.

Storing frequencies then allows us to make the next nice improvement. Previously we had to iterate over the whole training set, to pre-process it for the deprojectivisation. This led to storing the whole training set in memory. This was most of the required memory during training.

To prevent this, we now store the frequencies as we stream in the data, and deprojectivize as we go. Once we've built the frequencies, we can then apply a frequency cut-off when we decide how many classes to make.

Finally, to allow proper data streaming, we also have to have some way of shuffling the iterator. This is awkward if the training files have multiple documents in them. To solve this, the GoldCorpus class now writes the training data to disk in msgpack files, one per document. We can then shuffle the data by shuffling the paths.

This is a squash merge, as I made a lot of very small commits. Individual commit messages below.

* Simplify label management for TransitionSystem and its subclasses

* Fix serialization for new label handling format in parser

* Simplify and improve GoldCorpus class. Reduce memory use, write to temp dir

* Set actions in transition system

* Require thinc 6.11.1.dev4

* Fix error in parser init

* Add unicode declaration

* Fix unicode declaration

* Update textcat test

* Try to get model training on less memory

* Print json loc for now

* Try rapidjson to reduce memory use

* Remove rapidjson requirement

* Try rapidjson for reduced mem usage

* Handle None heads when projectivising

* Stream json docs

* Fix train script

* Handle projectivity in GoldParse

* Fix projectivity handling

* Add minibatch_by_words util from ud_train

* Minibatch by number of words in spacy.cli.train

* Move minibatch_by_words util to spacy.util

* Fix label handling

* More hacking at label management in parser

* Fix encoding in msgpack serialization in GoldParse

* Adjust batch sizes in parser training

* Fix minibatch_by_words

* Add merge_subtokens function to pipeline.pyx

* Register merge_subtokens factory

* Restore use of msgpack tmp directory

* Use minibatch-by-words in train

* Handle retokenization in scorer

* Change back-off approach for missing labels. Use 'dep' label

* Update NER for new label management

* Set NER tags for over-segmented words

* Fix label alignment in gold

* Fix label back-off for infrequent labels

* Fix int type in labels dict key

* Fix int type in labels dict key

* Update feature definition for 8 feature set

* Update ud-train script for new label stuff

* Fix json streamer

* Print the line number if conll eval fails

* Update children and sentence boundaries after deprojectivisation

* Export set_children_from_heads from doc.pxd

* Render parses during UD training

* Remove print statement

* Require thinc 6.11.1.dev6. Try adding wheel as install_requires

* Set different dev version, to flush pip cache

* Update thinc version

* Update GoldCorpus docs

* Remove print statements

* Fix formatting and links [ci skip]
 2018-03-19 02:58:08 +01:00
Matthew Honnibal d7ce6527fb Use increasing batch sizes in ud-train 2018-03-14 20:15:28 +01:00
Matthew Honnibal 5dddb30e5b Fix ud-train script 2018-03-11 01:26:45 +01:00
Matthew Honnibal 2cab4d6517 Remove use of attr module in ud_train 2018-03-11 00:59:39 +01:00
Matthew Honnibal 754ea1b2f7 Link in spaCy CoNLL commands 2018-03-10 23:42:15 +01:00
Matthew Honnibal 3478ea76d1 Add ud_train and ud_evaluate CLI commands 2018-03-10 23:41:55 +01:00
Matthew Honnibal b59765ca9f Stream gold during spacy train 2018-03-10 22:32:45 +01:00
Matthew Honnibal 86405e4ad1 Fix CLI for multitask objectives 2018-02-18 10:59:11 +01:00
Matthew Honnibal a34749b2bf Add multitask objectives options to train CLI 2018-02-17 22:03:54 +01:00
Matthew Honnibal 262d0a3148 Fix overwriting of lexical attributes when loading vectors during training 2018-02-17 18:11:11 +01:00
Johannes Dollinger bf94c13382 Don't fix random seeds on import 2018-02-13 12:42:23 +01:00
Ali Zarezade 9df9da34a3
Fix init_model issue 
Fixing issue #1928
 2018-02-03 17:21:34 +03:30
ines 3c1fb9d02d Make validate command fail more gracefully if version not found 
Mostly relevant during develoment when working with .dev versions
 2018-01-31 22:06:28 +01:00
Adam Binford 1a2c2f7d7f Fixed auto linking after download and added simple test to check 2018-01-29 14:25:21 -05:00
Matthew Honnibal 7ca49c2061
Merge branch 'master' into feature-improve-model-download 2018-01-10 18:21:55 +01:00
Søren Lind Kristiansen 10dab8eef8 Remove dummy variable from function calls 2018-01-05 09:37:05 +01:00
Søren Lind Kristiansen 7f0ab145e9 Don't pass CLI command name as dummy argument 2018-01-04 21:33:47 +01:00
ines 2c656f90fb Exit with 1 if incompatible models found (see #1714) 2018-01-03 21:20:35 +01:00
ines dacfaa2ca4 Ensure that download command exits properly (resolves #1714) 2018-01-03 21:03:36 +01:00
Søren Lind Kristiansen a9ff6eadc9 Prefix dummy argument names with underscore 2018-01-03 20:48:12 +01:00
ines 1081e08efb Fix formatting 2018-01-03 20:14:50 +01:00
ines d8109964d6 Use --no-deps on model install 
In general, it's nice for models to specify spaCy as a dependency. However, this tends to cause problems in conda environments, as pip will re-install spaCy and its dependencies (especially Thinc)
 2018-01-03 17:40:37 +01:00
ines 319d754309 Fix overwriting of existing symlinks 
Check for is_symlink() to also overwrite invalid and outdated symlinks. Also show better error message if link path exists but is not symlink (i.e. file or directory).
 2018-01-03 17:39:36 +01:00
ines 8ba0dfd017 Make message on failed linking more clear 2018-01-03 17:38:09 +01:00
Søren Lind Kristiansen d6327e8495 Fix handling case when vectors not specified 2018-01-03 12:20:49 +01:00
Søren Lind Kristiansen bcc51d7d8b Fix shifted positional arguments 2018-01-03 12:19:47 +01:00
Søren Lind Kristiansen 5a9d377580 Remove abbreviation for positional plac argument 2017-12-11 11:08:29 +01:00
Isaac Sijaranamual 20ae0c459a Fixes "Error saving model" #1622 2017-12-10 23:07:13 +01:00
Isaac Sijaranamual e188b61960 Make cli/train.py not eat exception 2017-12-10 22:53:08 +01:00
ines 5eaa61c2b8 Fix formatting 2017-12-07 10:23:09 +01:00
ines 24e80c51b8 Document init-model command 2017-12-07 10:14:37 +01:00
Matthew Honnibal c91f451b0f Fix imports and CLI in init-model 2017-12-07 10:03:07 +01:00
ines 82e80ff928 Rename model command to init_model and fix formatting 2017-12-07 09:59:23 +01:00
Ines Montani 2feeb428d6
Merge pull request #1646 from GreenRiverRUS/master 
Added model command to create models from raw data
 2017-12-07 08:54:26 +00:00
Thomas Werkmeister 94eac75b7c
fix setup.py spacy req string for packaging 
Requirement should be `spacy>=2.0.2` instead of `spacy2.0.2`
 2017-12-03 04:16:28 -06:00
Vadim Mazaev 495eacf470 Merge branch 'model_command' 2017-11-30 12:30:26 +03:00
Vadim Mazaev c332ffdde1 Added model command to create model from raw data: 
words counts, brown clusters and vectors
 2017-11-27 01:21:47 +03:00
Matthew Honnibal 2acc907d55 Improve profiling 2017-11-23 12:33:03 +00:00
Matthew Honnibal 8d692771f6 Improve profiling 2017-11-15 13:51:25 +01:00
ines 4c5d2c80d5 Re-add python -m to commands, too brittle :( (see #1536) 2017-11-10 02:30:55 +01:00
Matthew Honnibal de45702bbe Strip dev suffixes from version for compatibility check 2017-11-08 18:40:21 +01:00
Matthew Honnibal a2f980de4e Exclude .devN versioning from compatibility check 2017-11-08 18:03:52 +01:00
ines a4662a31a9 Move model package templates to cli.package and update docs 2017-11-07 12:15:35 +01:00
Matthew Honnibal c2bbf076a4 Add document length cap for training 2017-11-03 01:54:54 +01:00
Matthew Honnibal eca41f0cf6 Fix filename conversion for conllu 2017-11-01 21:26:49 +01:00
Matthew Honnibal e237472cdc Fix tag and filename conversion for conllu 2017-11-01 21:25:33 +01:00
ines affd3404ab Remove old model command (now "vocab") 2017-11-01 13:14:03 +01:00
ines 37e62ab0e2 Update vector meta in meta.json 2017-11-01 01:25:09 +01:00
Matthew Honnibal c390f2d745 Make it easier to pass explicit no-pruning to vocab 2017-10-31 20:14:47 +01:00
Matthew Honnibal 3659a807b0 Remove vector pruning arg from train CLI 2017-10-31 19:21:05 +01:00
Matthew Honnibal 59203a2e8a Move vector pruning command into spacy vocab cli tool 2017-10-31 19:10:01 +01:00
ines 803e41bc66 Merge branch 'develop' of https://github.com/explosion/spaCy into develop 2017-10-30 18:39:51 +01:00
ines abf8aa05d3 Populate --create-meta defaults from file if available 
If meta.json is found in directory and user chooses to overwrite it, show existing data as defaults.
 2017-10-30 18:39:38 +01:00
ines ce98fa7934 Fix formatting 2017-10-30 18:38:55 +01:00
ines 98c35d2585 Fix spacy vocab command 2017-10-30 18:38:41 +01:00
Matthew Honnibal e98451b5f7 Add -prune-vectors argument to spacy.cly.train 2017-10-30 18:00:10 +01:00
Explosion Bot 05a1dd570e Fix vocab script 2017-10-30 16:19:22 +01:00
Explosion Bot b46bdce8d2 Add missing import 2017-10-30 16:18:10 +01:00
Explosion Bot 0fc1209421 Wire up new vocab command 2017-10-30 16:14:50 +01:00
Matthew Honnibal 64e4ff7c4b Merge 'tidy-up' changes into branch. Resolve conflicts 2017-10-28 13:16:06 +02:00
ines d941fc3667 Tidy up CLI 2017-10-27 14:38:39 +02:00
Matthew Honnibal 531142a933 Merge remote-tracking branch 'origin/develop' into feature/better-parser 2017-10-27 12:34:48 +00:00
Matthew Honnibal b9616419e1 Add try/except around bz2 import 2017-10-27 01:18:05 +00:00
ines 11e3f19764 Fix vectors data added after training (see #1457) 2017-10-25 16:08:26 +02:00
ines 057954695b Read pipeline and vector data off model in --generate-meta 2017-10-25 16:03:26 +02:00
ines 273e638183 Add vector data to model meta after training (see #1457) 2017-10-25 16:03:05 +02:00
ines 95f6174516 Remove tensorizer from model pipeline example in spacy package 2017-10-24 16:00:56 +02:00
ines 24512420b1 Show error if data_path does not exist or is None (see #1102) 2017-10-19 00:53:49 +02:00
Matthew Honnibal dc01acd821 Escape encoding in validate function 2017-10-12 22:23:21 +02:00
ines fff1028391 Add validate CLI command 2017-10-12 20:05:06 +02:00
Matthew Honnibal a955843684 Increase default number of epochs 2017-10-12 13:13:01 +02:00
Matthew Honnibal acba2e1051 Fix metadata in training 2017-10-11 08:55:52 +02:00
Matthew Honnibal 74c2c6a58c Add default name and lang to meta 2017-10-11 08:49:12 +02:00
Matthew Honnibal 5156074df1 Make loading code more consistent in train command 2017-10-10 12:51:20 -05:00
Matthew Honnibal 97c9b5db8b Patch spacy.train for new pipeline management 2017-10-09 23:41:16 -05:00
Matthew Honnibal a635240398 Add conll_ner2json converter 2017-10-09 22:03:26 -05:00
Matthew Honnibal 735d18654d Add NER converter for CoNLL 2003 data 2017-10-09 20:06:28 -05:00
Matthew Honnibal 808d8740d6 Remove print statement 2017-10-09 08:45:20 -05:00
Matthew Honnibal 0f41b25f60 Add speed benchmarks to metadata 2017-10-09 08:05:37 -05:00
Matthew Honnibal be4f0b6460 Update defaults 2017-10-08 02:08:12 -05:00
Matthew Honnibal 9d66a915da Update training defaults 2017-10-07 21:02:38 -05:00
Matthew Honnibal 09442d25ec Merge remote-tracking branch 'origin/develop' into feature/parser-history-model 2017-10-07 07:05:04 -05:00
Matthew Honnibal f4c9a98166 Fix spacy evaluate command on non-GPU 2017-10-06 13:17:47 -05:00
Matthew Honnibal c6cd81f192 Wrap try/except around model saving 2017-10-05 08:14:24 -05:00