explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2024-12-24 17:06:29 +03:00
Code Issues Packages Projects Releases Wiki Activity

Fix GPU training and evaluation

Browse Source
This commit is contained in:
Matthew Honnibal 2017-05-18 08:30:33 -05:00
parent a438cef8c5
commit ca70b08661
1 changed files with 21 additions and 16 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

37
spacy/cli/train.py
View File

@ -65,7 +65,7 @@ def train_config(config):
def train_model(Language, train_data, dev_data, output_path, n_iter, **cfg):
print("Itn.\tDep. Loss\tUAS\tNER F.\tTag %\tToken %")
nlp = Language(pipeline=['token_vectors', 'tags']) #, 'dependencies'])
nlp = Language(pipeline=['token_vectors', 'tags', 'dependencies'])
dropout = util.env_opt('dropout', 0.0)
# TODO: Get spaCy using Thinc's trainer and optimizer
with nlp.begin_training(train_data, **cfg) as (trainer, optimizer):
@ -75,7 +75,7 @@ def train_model(Language, train_data, dev_data, output_path, n_iter, **cfg):
for i, (docs, golds) in enumerate(epoch):
state = nlp.update(docs, golds, drop=dropout, sgd=optimizer)
losses['dep_loss'] += state.get('parser_loss', 0.0)
losses['tag_loss'] += state.get('tagger_loss', 0.0)
losses['tag_loss'] += state.get('tag_loss', 0.0)
to_render.insert(0, nlp(docs[-1].text))
to_render[0].user_data['title'] = "Batch %d" % i
with Path('/tmp/entities.html').open('w') as file_:
@ -98,30 +98,35 @@ def train_model(Language, train_data, dev_data, output_path, n_iter, **cfg):
def evaluate(Language, gold_tuples, path):
with (path / 'model.bin').open('rb') as file_:
nlp = dill.load(file_)
scorer = Scorer()
for raw_text, sents in gold_tuples:
sents = merge_sents(sents)
for annot_tuples, brackets in sents:
if raw_text is None:
tokens = Doc(nlp.vocab, words=annot_tuples[1])
state = None
for proc in nlp.pipeline:
state = proc(tokens, state=state)
# TODO:
# 1. This code is duplicate with spacy.train.Trainer.evaluate
# 2. There's currently a semantic difference between pipe and
# not pipe! It matters whether we batch the inputs. Must fix!
all_docs = []
all_golds = []
for raw_text, paragraph_tuples in dev_sents:
if gold_preproc:
raw_text = None
else:
tokens = nlp(raw_text)
gold = GoldParse.from_annot_tuples(tokens, annot_tuples)
scorer.score(tokens, gold)
paragraph_tuples = merge_sents(paragraph_tuples)
docs = self.make_docs(raw_text, paragraph_tuples)
golds = self.make_golds(docs, paragraph_tuples)
all_docs.extend(docs)
all_golds.extend(golds)
scorer = Scorer()
for doc, gold in zip(self.nlp.pipe(all_docs), all_golds):
scorer.score(doc, gold)
return scorer
def print_progress(itn, losses, dev_scores):
# TODO: Fix!
scores = {}
for col in ['dep_loss', 'uas', 'tags_acc', 'token_acc', 'ents_f']:
for col in ['dep_loss', 'tag_loss', 'uas', 'tags_acc', 'token_acc', 'ents_f']:
scores[col] = 0.0
scores.update(losses)
scores.update(dev_scores)
tpl = '{:d}\t{dep_loss:.3f}\t{uas:.3f}\t{ents_f:.3f}\t{tags_acc:.3f}\t{token_acc:.3f}'
tpl = '{:d}\t{dep_loss:.3f}\t{tag_loss:.3f}\t{uas:.3f}\t{ents_f:.3f}\t{tags_acc:.3f}\t{token_acc:.3f}'
print(tpl.format(itn, **scores))