explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2024-11-15 06:09:01 +03:00
Code Issues Packages Projects Releases Wiki Activity
10,950 Commits 68 Branches 175 Tags 279 MiB
73b1f651d4
Commit Graph

6494 Commits

Author SHA1 Message Date
Matthw Honnibal
73b1f651d4 Rescale gradients for mlm 2019-10-24 17:35:37 +02:00
Matthw Honnibal
7d81d17ce5 Restore LayerNorm for mish 2019-10-24 17:35:10 +02:00
Matthw Honnibal
134056684f Fix number characters 2019-10-24 17:34:16 +02:00
Matthw Honnibal
754a36d981 Add bos_eos function for tok2vec 2019-10-24 17:33:47 +02:00
Matthw Honnibal
9ca109597d Fix parser model for depth 1 2019-10-23 05:14:00 +02:00
Matthw Honnibal
9e32d8271c Avoid failing if cupy random can'tbe set 2019-10-23 05:13:28 +02:00
Matthw Honnibal
f8bf5b7fe5 Fx gpu_id arg in pretrain 2019-10-23 04:41:40 +02:00
Matthw Honnibal
95648dcdd7 Pass parser settings better 2019-10-23 04:41:20 +02:00
Matthw Honnibal
8892ce98aa Improve settings in _ml 2019-10-23 04:40:10 +02:00
Matthw Honnibal
6c8785a238 Add option for GPU ID to pretrain 2019-10-22 22:44:24 +02:00
Matthw Honnibal
1dce86c555 Pass settings better from parser 2019-10-22 03:26:43 +02:00
Matthw Honnibal
ab7f85dfa2 Use Mish layer if pieces==1 in CNN 2019-10-22 03:26:27 +02:00
Matthw Honnibal
7ef3bcdc1c Support cnn_maxout_pieces arg in pretrain 2019-10-22 03:25:30 +02:00
Matthw Honnibal
5a272d9029 Add method to decode predicted characters 2019-10-21 03:56:15 +02:00
Matthw Honnibal
f2808f78a7 Fix parser_maxout_pieces for depth=0 2019-10-21 01:25:03 +02:00
Matthw Honnibal
b4e0040d10 Pass Tok2Vec settings a bit better 2019-10-21 01:11:47 +02:00
Matthw Honnibal
fef50277d7 Support parser depth=0 2019-10-21 01:11:30 +02:00
Matthw Honnibal
eba89f08bd Use chars loss in ClozeMultitask 2019-10-20 17:47:15 +02:00
Matthw Honnibal
77af446d04 Move characters_loss function, add window option 2019-10-20 17:47:00 +02:00
Matthw Honnibal
5a601ef46a Add cnn_window option to pretrain 2019-10-20 17:46:34 +02:00
Matthw Honnibal
3a67aa857e Clarify parser model CPU/GPU code 
The previous version worked with previous thinc, but only
because some thinc ops happened to have gpu/cpu compatible
implementations. It's better to call the right Ops instance.
 2019-10-20 17:15:17 +02:00
Matthw Honnibal
ee56c6a4e1 Implement character-based pretraining objective 2019-10-19 11:42:38 +02:00
Matthw Honnibal
36de9bf72a Add more spacy pretrain options 2019-10-18 17:24:13 +02:00
Matthw Honnibal
f3e2aaea1e Fix GPU selection in spacy train 2019-10-18 17:23:55 +02:00
Matthw Honnibal
49c0adc706 Add character-based bilstm tok2vec 2019-10-18 17:23:37 +02:00
Matthw Honnibal
727ede6599 Make character copy non-blocking 2019-10-18 17:23:19 +02:00
Matthw Honnibal
4da1c1c211 Try to make cuda call non-blocking 2019-10-18 17:22:16 +02:00
Matthw Honnibal
b2e8f37965 Make cuda streams non-blocking by default 2019-10-18 17:21:57 +02:00
Matthw Honnibal
ca0759b325 Pass config better in nn_parser 2019-10-17 21:10:56 +02:00
Matthw Honnibal
e737750a02 Fix bilstm_depth default in pretrain command 2019-10-17 21:10:08 +02:00
Matthw Honnibal
6aa1c53b1b Call resume_training for base model in train CLI 2019-10-17 21:09:41 +02:00
Matthw Honnibal
3f26c50a4d Refactor some of tok2vec 2019-10-17 17:58:00 +02:00
Matthw Honnibal
e63f28079a Try 3 NER features 2019-10-07 16:51:03 +02:00
Matthw Honnibal
2d55ccdd27 Support option of three NER features 2019-10-07 16:50:44 +02:00
Matthw Honnibal
c8857181f8 Fix get labels for textcat 2019-10-07 16:50:15 +02:00
Matthw Honnibal
a6a2ff217f Fix char_embed for gpu 2019-10-07 16:49:32 +02:00
Matthw Honnibal
f4040a98f0 Fix passing of cats in gold.pyx 2019-10-07 16:49:00 +02:00
Matthw Honnibal
a132da1558 Fix gold-preproc training mode 2019-10-07 02:07:03 +02:00
Matthw Honnibal
63ff233ba2 Enable GPU in pytorch n use_gpu functon 2019-10-06 19:24:21 +02:00
Matthw Honnibal
9dbaea1ab4 Use cosine loss in Cloze multitask 2019-10-06 19:23:46 +02:00
Matthw Honnibal
157d3d769b Support bilstm_depth arg in spacy pretrain 2019-10-06 19:22:26 +02:00
Matthw Honnibal
615ebe584f Add option to ignore zero vectors in get_cossim_loss 2019-10-06 19:20:54 +02:00
adrianeboyd
cbc2cee2c8 Improve URL_PATTERN and handling in tokenizer (#4374) 
* Move prefix and suffix detection for URL_PATTERN

Move prefix and suffix detection for `URL_PATTERN` into the tokenizer.
Remove associated lookahead and lookbehind from `URL_PATTERN`.

Fix tokenization for Hungarian given new modified handling of prefixes
and suffixes.

* Match a wider range of URI schemes
 2019-10-05 13:00:09 +02:00
Ines Montani
fec9433044 Make PhraseMatcher.vocab consistent with Matcher.vocab (closes #4373) 2019-10-04 12:18:41 +02:00
Matthew Honnibal
37ef874d8b Set version to v2.2.1 2019-10-03 14:50:39 +02:00
Sofie Van Landeghem
4e7259c6cf Bugfix initializing DocBin with attributes (#4368) 
* docbin init fix + documentation fix + unit tests

* newline

* try with zlib instead of gzip (python 2 incompatibilities)
 2019-10-03 14:48:45 +02:00
Ben Taylor
1db79a33cb most_similar() return the k most similar vectors (#4364) 
* most_similar return n-most similar vectors

* updated most_similar comment

* add bintay contributor agreement

* sign bintay contributor agreement

* fix most_similar documentation typo

* fixed error in prune_vectors

* updated prune_vectors test
 2019-10-03 14:09:44 +02:00
Matthew Honnibal
2eb31012e7 Set version to v2.2.0 2019-10-02 14:40:06 +02:00
Matthew Honnibal
796072e560 Set version to v2.2.0.dev19 2019-10-02 12:51:29 +02:00
Sofie Van Landeghem
9d3ce7cba2 Ensure training doesn't crash with empty batches (#4360) 
* unit test for previously resolved unflatten issue

* prevent batch of empty docs to cause problems
 2019-10-02 12:50:47 +02:00