def run(args):
# 1. get the corpus
corpus: Corpus = WNUT_17()
# 2. what tag do we want to predict?
tag_type = 'ner'
# 3. make the tag dictionary from the corpus
tag_dictionary = corpus.make_tag_dictionary(tag_type=tag_type)
# initialize embeddings
embedding_types: List[TokenEmbeddings] = [
WordEmbeddings('crawl'),
WordEmbeddings('twitter'),
FlairEmbeddings('news-forward'),
FlairEmbeddings('news-backward'),
]
embeddings: StackedEmbeddings = StackedEmbeddings(embeddings=embedding_types)
# initialize sequence tagger
from flair.models import SequenceTagger
tagger: SequenceTagger = SequenceTagger(hidden_size=256,
embeddings=embeddings,
tag_dictionary=tag_dictionary,
tag_type=tag_type)
# initialize trainer
from flair.trainers import ModelTrainer
trainer: ModelTrainer = ModelTrainer(tagger, corpus, use_tensorboard=True)
trainer.train(
args.job_dir,
train_with_dev=True,
max_epochs=args.epochs
)
upload_results(args)
flair.set_seed(3)
tagger = TARSSequenceTagger2.load("resources/v3/sequence-2/final-model.pt")
label_name_map = {
"person": "Person",
"location": "Location",
"creative-work": "Creative Work",
"product": "Product",
"corporation": "Corporation",
"group": "Group"
}
print(label_name_map)
corpus = WNUT_17(tag_to_bioes=None,
tag_to_bio2="ner",
label_name_map=label_name_map)
corpus = corpus.downsample(0.1)
tag_type = "ner"
tag_dictionary = corpus.make_label_dictionary(tag_type)
tagger.add_and_switch_to_new_task("sequence-3-train",
tag_dictionary=tag_dictionary,
tag_type=tag_type)
trainer = ModelTrainer(tagger, corpus, optimizer=torch.optim.AdamW)
trainer.train(
base_path='resources/v3/sequence-3',
learning_rate=5.0e-5,
mini_batch_size=32,
mini_batch_chunk_size=None,
from FlairParamOptimizer import search_strategies, search_spaces, orchestrator
import FlairParamOptimizer.parameter_listings.parameters_for_user_input as param
from flair.embeddings import WordEmbeddings
from flair.datasets import WNUT_17
corpus = WNUT_17()
search_space = search_spaces.SequenceTaggerSearchSpace()
search_strategy = search_strategies.RandomSearch()
search_space.add_tag_type("ner")
search_space.add_budget(param.Budget.TIME_IN_H, 24)
search_space.add_evaluation_metric(param.EvaluationMetric.MICRO_F1_SCORE)
search_space.add_optimization_value(param.OptimizationValue.DEV_SCORE)
search_space.add_max_epochs_per_training_run(25)
search_space.add_parameter(param.SequenceTagger.HIDDEN_SIZE,
options=[128, 256, 512])
search_space.add_parameter(param.SequenceTagger.DROPOUT,
options=[0, 0.1, 0.2, 0.3])
search_space.add_parameter(param.SequenceTagger.WORD_DROPOUT,
options=[0, 0.01, 0.05, 0.1])
search_space.add_parameter(param.SequenceTagger.RNN_LAYERS,
options=[2, 3, 4, 5, 6])
search_space.add_parameter(param.SequenceTagger.USE_RNN, options=[True, False])
search_space.add_parameter(param.SequenceTagger.USE_CRF, options=[True, False])
search_space.add_parameter(param.SequenceTagger.REPROJECT_EMBEDDINGS,
options=[True, False])
search_space.add_parameter(
from flair.data import Corpus
from flair.datasets import WNUT_17
from flair.embeddings import FlairEmbeddings, WordEmbeddings, StackedEmbeddings
# 1. get the corpus
corpus: Corpus = WNUT_17()
print(corpus)
# 2. what tag do we want to predict?
tag_type = 'ner'
# 3. make the tag dictionary from the corpus
tag_dictionary = corpus.make_tag_dictionary(tag_type=tag_type)
print(tag_dictionary)
# 4. initialize embeddings
embedding_types = [
WordEmbeddings('glove'),
FlairEmbeddings('news-forward'),
FlairEmbeddings('news-backward'),
]
embeddings: StackedEmbeddings = StackedEmbeddings(embeddings=embedding_types)
# 5. initialize sequence tagger
from flair.models import SequenceTagger
tagger: SequenceTagger = SequenceTagger(hidden_size=256,
embeddings=embeddings,
tag_dictionary=tag_dictionary,
tag_type=tag_type,
from flair.data import MultiCorpus
from flair.datasets import WNUT_17
flair.set_seed(3)
label_name_map = {
"person": "Person",
"location": "Location",
"creative-work": "Creative Work",
"product": "Product",
"corporation": "Corporation",
"group": "Group"
}
print(label_name_map)
corpus = WNUT_17(tag_to_bioes=None, tag_to_bio2="ner")
corpus = corpus.downsample(0.1)
tag_type = "ner"
label_dictionary = corpus.make_label_dictionary(tag_type)
print(label_dictionary)
tagger = TARSSequenceTagger2(tag_dictionary=label_dictionary,
tag_type=tag_type,
task_name="TEST_NER")
trainer = ModelTrainer(tagger, corpus, optimizer=torch.optim.AdamW)
from torch.optim.lr_scheduler import OneCycleLR
trainer.train(
base_path='resources/v3/wnut17-simple',