github.com/stanfordnlp/stanza.git - Unnamed repository; edit this file 'description' to name the repository.

Age	Commit message (Expand)	Author
2022-11-13	Potentially use a MHA layer in the taggertagger_mha	John Bauer
2022-11-13	Add a scale to the AddSinusoidalEncoding	John Bauer
2022-11-13	Add AdamW with amsgrad as an option	John Bauer
2022-11-13	Reload the current best tagger for continued training as an option	John Bauer
2022-11-13	Add a separate --second_lr flag for when the initial optimizer might work bet...	John Bauer
2022-11-13	Add adamw as an optimizer option	John Bauer
2022-11-13	Make the second optimizer a parameter	John Bauer
2022-11-13	Add madgrad as a possible optional dependency	John Bauer
2022-11-13	Add a weight decay option for the first and second optimizers used in the tagger	John Bauer
2022-11-12	Hopefully fix for the last time the printing of charmodel	John Bauer
2022-11-12	Create output directory if not exists. Add a skeletal argparse for the --hel...	John Bauer
2022-11-12	Add Adadelta as an optimizer option to the tagger	John Bauer
2022-11-12	Different name in the tagger than in the parser... dangers of cut & paste code	John Bauer
2022-11-12	Log norms in the tagger - this gets very noisy. Only do it when desperate to...	John Bauer
2022-11-11	Add a tiny bit of doc to the positional encoding	John Bauer
2022-11-11	Add a simple test of running POS training for a few iterations	John Bauer
2022-11-11	Turn the split dataset name for vlsp22 into a -x-y format. Makes it easier t...	John Bauer
2022-11-11	Add notes on a relu vs gelu experiment	John Bauer
2022-11-11	Split the retagging operation into chunks. The tqdm is no longer as smooth, ...	John Bauer
2022-11-09	Addition of extra nonlinearities for experiments (#1149)	Hung Bui
2022-11-09	BrokenPipeError is more appropriate - fits the errors that come out of subpro...	John Bauer
2022-11-08	Slightly update a score	John Bauer
2022-11-08	Add a flag to use or not use .xz	John Bauer
2022-11-07	Add a variant of multihead attention where there's one key or one key per label	John Bauer
2022-11-07	Change convert_pretrain to use argparse so it has a nice --help method	John Bauer
2022-11-07	More notes on TOP_DOWN experiments	John Bauer
2022-11-07	Skip blank lines	John Bauer
2022-11-07	Change CompoundUnary to use a constructor similar to the OpenConstituent cons...	John Bauer
2022-11-07	Update some experiment numbers	John Bauer
2022-11-07	Simplify the operation of unary transitions in the event we are using TOP_DOW...	John Bauer
2022-11-07	Log the open nodes used in a model	John Bauer
2022-11-06	Add a flag to remove all sentences which don't fit in a bert tokenizer when p...	John Bauer
2022-11-06	Add the ability to quiet the logging	John Bauer
2022-11-06	Update error to ValueError (more appropriate) and log what the unexpected typ...	John Bauer
2022-11-06	Update log line & allow list of str instead of list of tuples	John Bauer
2022-11-05	Add some doc on the transition schemes	John Bauer
2022-11-05	Discard Devanagari text from the VI wikipedia	John Bauer
2022-11-05	A couple comments on how the NER training is organized	John Bauer
2022-11-05	Allow unknown compound transitions composed of known transitions in the dev o...	John Bauer
2022-11-05	Also chuck some sentences with long words	John Bauer
2022-11-05	throw out long JA sentences as well when tokenizing Wikipedia	John Bauer
2022-11-05	Specifically exclude one sentence from VI Wikipedia which makes Bert sad	John Bauer
2022-11-04	Refactor the tokenization method from tokenize_wiki.py Reuse it to add an op...	John Bauer
2022-11-04	AddSinulsoidalEncoding as a module	John Bauer
2022-11-04	The tokenization script was changed to account for length and emdash, so it i...	John Bauer
2022-11-03	Throw out wiki docs of length 2 as well when building a silver dataset	John Bauer
2022-11-03	Add min_len and max_len args to tokenize_wiki.py. Skip one line wiki docs, s...	John Bauer
2022-11-02	Fix format error in log line	John Bauer
2022-11-01	slice in a more generic manner when copying model. makes it easier to make f...	John Bauer
2022-11-01	Set this option in the partitioned test so that it still tests this code path...	John Bauer