Rooba
/
spaCy
mirror of https://github.com/explosion/spaCy.git
1
1
Fork 0
Code Issues Packages Projects Releases Wiki Activity

* Add more options to bin/parser/train

This commit is contained in:
Matthew Honnibal 2015-06-05 23:49:26 +02:00
parent 1fee7ade61
commit 1736fc5a67
1 changed files with 17 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
bin/parser/train.py
View File

@ -141,8 +141,11 @@ def train(Language, gold_tuples, model_dir, n_iter=15, feat_set=u'basic',
nlp.vocab.strings.dump(path.join(model_dir, 'vocab', 'strings.txt')) nlp.vocab.strings.dump(path.join(model_dir, 'vocab', 'strings.txt'))
def evaluate(Language, gold_tuples, model_dir, gold_preproc=False, verbose=False): def evaluate(Language, gold_tuples, model_dir, gold_preproc=False, verbose=False,
beam_width=None):
nlp = Language(data_dir=model_dir) nlp = Language(data_dir=model_dir)
if beam_width is not None:
nlp.parser.cfg.beam_width = beam_width
scorer = Scorer() scorer = Scorer()
for raw_text, sents in gold_tuples: for raw_text, sents in gold_tuples:
if gold_preproc: if gold_preproc:
@ -179,9 +182,10 @@ def write_parses(Language, dev_loc, model_dir, out_loc):
@plac.annotations( @plac.annotations(
train_loc=("Location of training file or directory"), train_loc=("Location of training file or directory"),
dev_loc=("Location of development file or directory"), dev_loc=("Location of development file or directory"),
model_dir=("Location of output model directory",),
eval_only=("Skip training, and only evaluate", "flag", "e", bool),
corruption_level=("Amount of noise to add to training data", "option", "c", float), corruption_level=("Amount of noise to add to training data", "option", "c", float),
gold_preproc=("Use gold-standard sentence boundaries in training?", "flag", "g", bool), gold_preproc=("Use gold-standard sentence boundaries in training?", "flag", "g", bool),
model_dir=("Location of output model directory",),
out_loc=("Out location", "option", "o", str), out_loc=("Out location", "option", "o", str),
n_sents=("Number of training sentences", "option", "n", int), n_sents=("Number of training sentences", "option", "n", int),
n_iter=("Number of training iterations", "option", "i", int), n_iter=("Number of training iterations", "option", "i", int),
@ -190,17 +194,20 @@ def write_parses(Language, dev_loc, model_dir, out_loc):
debug=("Debug mode", "flag", "d", bool) debug=("Debug mode", "flag", "d", bool)
) )
def main(train_loc, dev_loc, model_dir, n_sents=0, n_iter=15, out_loc="", verbose=False, def main(train_loc, dev_loc, model_dir, n_sents=0, n_iter=15, out_loc="", verbose=False,
debug=False, corruption_level=0.0, gold_preproc=False, beam_width=1): debug=False, corruption_level=0.0, gold_preproc=False, beam_width=1,
gold_train = list(read_json_file(train_loc)) eval_only=False):
train(English, gold_train, model_dir, if not eval_only:
feat_set='basic' if not debug else 'debug', gold_train = list(read_json_file(train_loc))
gold_preproc=gold_preproc, n_sents=n_sents, train(English, gold_train, model_dir,
corruption_level=corruption_level, n_iter=n_iter, feat_set='basic' if not debug else 'debug',
beam_width=beam_width) gold_preproc=gold_preproc, n_sents=n_sents,
corruption_level=corruption_level, n_iter=n_iter,
beam_width=beam_width)
if out_loc: if out_loc:
write_parses(English, dev_loc, model_dir, out_loc) write_parses(English, dev_loc, model_dir, out_loc)
scorer = evaluate(English, list(read_json_file(dev_loc)), scorer = evaluate(English, list(read_json_file(dev_loc)),
model_dir, gold_preproc=gold_preproc, verbose=verbose) model_dir, gold_preproc=gold_preproc, verbose=verbose,
beam_width=beam_width)
print 'TOK', 100-scorer.token_acc print 'TOK', 100-scorer.token_acc
print 'POS', scorer.tags_acc print 'POS', scorer.tags_acc
print 'UAS', scorer.uas print 'UAS', scorer.uas