Paul O'Leary McCann 8f66176b2d Fix loss? ... This rewrites the loss to not use the Thinc crossentropy code at all. The main difference here is that the negative predictions are being masked out (= marginalized over), but negative gradient is still being reflected. I'm still not sure this is exactly right but models seem to train reliably now.		2021-07-05 18:17:10 +09:00
..
__init__.py	Native coref component (#7243)	2021-03-03 13:50:14 +01:00
coref_util.py	Minor fix in crossing spans code	2021-07-03 18:41:46 +09:00
coref.py	Fix loss?	2021-07-05 18:17:10 +09:00
entity_linker.py	make EntityLinker robust for nO=None (#7930)	2021-05-06 18:14:47 +10:00
multi_task.py	Fixing pretrain (#7342)	2021-03-09 14:01:13 +11:00
parser.py	consistently use registry as callable	2021-03-02 17:56:28 +01:00
tagger.py	consistently use registry as callable	2021-03-02 17:56:28 +01:00
textcat.py	consistently use registry as callable	2021-03-02 17:56:28 +01:00
tok2vec.py	Replace negative rows with 0 in StaticVectors (#7674)	2021-04-22 18:04:15 +10:00