"""
service = Service(service_name)
api.req_options.auto_parse_form_urlencoded = True
api.add_route('/{}'.format(service_name), service)
# api.router_options
path = os.path.abspath(
os.path.join(os.path.dirname(__file__),
"web_service/visualizer/displacy"))
api.add_static_route('/{}'.format(service_name), path)
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--name',
help="the name of the service you want to upload",
type=str,
required=True,
action=check_size(1, 30))
args = parser.parse_args()
app = application = falcon.API(middleware=[MultipartMiddleware()])
if not is_valid_input(args.name):
logger.error('ERROR: Invalid argument input for the server.')
sys.exit(0)
# init and load service
set_server_properties(app, args.name)
# run server:
port = 8080
server = make_server('0.0.0.0', port, app)
print('starting the server at port {0}'.format(port))
server.serve_forever()
# load existing model
mlp_clf.load(args.model)
results = mlp_clf.get_outputs(test_set['X'])
return results
if __name__ == "__main__":
# parse the command line arguments
parser = argparse.ArgumentParser()
parser.add_argument('--max_num_of_senses_to_search',
default=3,
type=int,
action=check_size(0, 100),
help='maximum number of senses that are tests')
parser.add_argument('--input_inference_examples_file',
type=validate_existing_filepath,
default='data/input_inference_examples.csv',
help='input_data_file')
parser.add_argument(
'--model',
default='data/mcs_model.h5',
type=validate_existing_filepath,
help='path to the file where the trained model has been stored')
parser.add_argument(
'--word_embedding_model_file',
type=validate_existing_filepath,
default='pretrained_models/GoogleNews-vectors-negative300.bin',
help='path to the word embedding\'s model')
import logging
import sys
from nlp_architect.models.np2vec import NP2vec
from nlp_architect.utils.io import check_size, validate_existing_filepath
logger = logging.getLogger(__name__)
if __name__ == "__main__":
arg_parser = argparse.ArgumentParser()
arg_parser.add_argument(
'--corpus',
default='train.txt',
type=str,
action=check_size(min_size=1),
help='path to the corpus. By default, '
'it is the training set of CONLL2000 shared task dataset.')
arg_parser.add_argument(
'--corpus_format',
default='conll2000',
type=str,
choices=[
'json',
'txt',
'conll2000'],
help='format of the input marked corpus; txt, conll2000 and json formats are supported. '
'For json format, the file should contain an iterable of sentences. '
'Each sentence is a list of terms (unicode strings) that will be used for training.')
arg_parser.add_argument(
'--mark_char',
help=
'boolean indicating whether the model to load has been stored in binary format.',
action='store_true')
arg_parser.add_argument(
'--word_ngrams',
default=0,
type=int,
choices=[0, 1],
help=
'If 0, the model to load stores word information. If 1, the model to load stores '
'subword (ngrams) information; note that subword information is relevant only to '
'fasttext models.')
arg_parser.add_argument('--topn',
default=500,
type=int,
action=check_size(min_size=1),
help='maximal number of expanded terms to return')
arg_parser.add_argument('--grouping',
action='store_true',
default=False,
help='grouping mode')
args = arg_parser.parse_args()
se = SetExpand(np2vec_model_file=args.np2vec_model_file,
binary=args.binary,
word_ngrams=args.word_ngrams,
grouping=args.grouping)
enter_seed_str = 'Enter the seed (comma-separated seed terms):'
logger.info(enter_seed_str)
for seed_str in sys.stdin:
num_iterations = int(n_train * n_epochs * 1.0 / batch_size)
results_dir = os.path.abspath(args.results_dir)
adding_dataset = Adding(seq_len=seq_len, n_train=n_train, n_test=n_val)
model = TCNForAdding(seq_len, n_features, hidden_sizes, kernel_size=kernel_size,
dropout=dropout)
model.build_train_graph(args.lr, max_gradient_norm=args.grad_clip_value)
model.run(adding_dataset, num_iterations=num_iterations, log_interval=args.log_interval,
result_dir=results_dir)
PARSER = argparse.ArgumentParser()
PARSER.add_argument('--seq_len', type=int, action=check_size(0, 1000),
help="Number of time points in each input sequence",
default=200)
PARSER.add_argument('--log_interval', type=int, default=100, action=check_size(0, 10000),
help="frequency, in number of iterations, after which loss is evaluated")
PARSER.add_argument('--results_dir', type=validate_parent_exists,
help="Directory to write results to", default=os.path.expanduser('~/results'))
PARSER.add_argument('--dropout', type=float, default=0.0, action=check_size(0, 1),
help='dropout applied to layers, between 0 and 1 (default: 0.0)')
PARSER.add_argument('--ksize', type=int, default=6, action=check_size(0, 10),
help='kernel size (default: 6)')
PARSER.add_argument('--levels', type=int, default=7, action=check_size(0, 10),
help='# of levels (default: 7)')
PARSER.add_argument('--lr', type=float, default=2e-3, action=check_size(0, 1),
help='initial learning rate (default: 2e-3)')
PARSER.add_argument('--nhid', type=int, default=27, action=check_size(0, 1000),
from nlp_architect.models.crossling_emb import WordTranslator
from nlp_architect.utils.io import validate_existing_directory, validate_parent_exists, check_size
if __name__ == "__main__":
print("\t\t" + 40 * "=")
print("\t\t= Unsupervised Crosslingual Embeddings =")
print("\t\t" + 40 * "=")
# Parsing arguments for model parameters
parser = argparse.ArgumentParser()
parser.add_argument("--emb_dim",
type=int,
default=300,
help="Embedding Dimensions",
action=check_size(1, 1024))
parser.add_argument("--vocab_size",
type=int,
default=200000,
help="Vocabulary Size",
action=check_size(1, 1000000))
parser.add_argument("--lr",
type=float,
default=0.1,
help="Learning Rate",
action=check_size(0.00001, 2.0))
parser.add_argument("--beta",
type=float,
default=0.001,
help="Beta for W orthogornaliztion",
action=check_size(0.0000001, 5.0))
verbose=1, validation_data=(X_test, Y_test), callbacks=[es])
lstm_acc = model_hist.history['val_acc'][-1]
print("LSTM model accuracy ", lstm_acc)
# This minimizes, so the maximize we have to take the inverse :)
return 1 - lstm_acc
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--file_path', type=validate_existing_filepath, default='./',
help='file_path where the files to parse are located')
parser.add_argument('--data_type', type=str, default='amazon',
choices=['amazon'])
parser.add_argument('--output_file', type=validate_parent_exists, default='./opt_trials.pkl',
help='file_path where the output of the trials will be located')
parser.add_argument('--new_trials', type=int, default=20, action=check_size(1, 20000))
args_in = parser.parse_args()
# Check inputs
if args_in.file_path:
validate_existing_filepath(args_in.file_path)
if args_in.output_file:
validate_parent_exists(args_in.output_file)
if args_in.data_type == 'amazon':
data_in = Amazon_Reviews(args_in.file_path)
try:
if args_in.output_file.endswith('.pkl'):
with open(args_in.output_file, 'rb') as read_f:
trials_to_keep = pickle.load(read_f)
parser.add_argument(
'--use_oov',
default=False,
action='store_true',
help='use OOV test set')
parser.add_argument(
'--lr',
type=float,
default=0.001,
help='learning rate')
parser.add_argument(
'--grad_clip_norm',
type=float,
default=40.0,
help='Clip gradients such that norm is below this value.',
action=check_size(0,100))
parser.add_argument(
'--eps',
type=float,
default=1e-8,
help='epsilon used to avoid divide by zero in softmax renormalization.',
action=check_size(1e-100,1e-2))
parser.add_argument(
'--save_log',
action='store_true',
default=False,
help='Save evaluation results to log file.')
parser.add_argument(
'--log_file',
type=str,
default='memn2n_dialgoue_results.txt',
import argparse
import tensorflow as tf
from nlp_architect.utils.io import validate_existing_directory, check_size, validate_parent_exists
# Parse the command line arguments
parser = argparse.ArgumentParser()
parser.add_argument('--data_path',
default='data',
type=validate_existing_directory,
help='enter path for training data')
parser.add_argument('--gpu_id',
default="0",
type=str,
help='enter gpu id',
action=check_size(0, 8))
parser.add_argument('--max_para_req',
default=300,
type=int,
help='enter the max length of paragraph',
action=check_size(30, 300))
parser.add_argument('--epochs',
default=15,
type=int,
help='enter the number of epochs',
action=check_size(1, 30))
parser.add_argument('--select_device',
default='GPU',
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument('--gold_standard_file',
default='data/goldStd.csv',
type=validate_existing_filepath,
help='path to gold standard file')
parser.add_argument(
'--word_embedding_model_file',
type=validate_existing_filepath,
default='pretrained_models/GoogleNews-vectors-negative300.bin',
help='path to the word embedding\'s model')
parser.add_argument('--training_to_validation_size_ratio',
default=0.8,
type=float,
action=check_size(0, 1),
help='ratio between training and validation size')
parser.add_argument(
'--data_set_file',
default='data/data_set.pkl',
type=validate_parent_exists,
help='path the file where the train, valid and test sets will be stored'
)
args = parser.parse_args()
# training set
X_train = []
y_train = []
# validation set
X_valid = []
hidden_sizes,
kernel_size=kernel_size,
dropout=dropout)
model.build_train_graph(args.lr, max_gradient_norm=args.grad_clip_value)
model.run(adding_dataset,
num_iterations=num_iterations,
log_interval=args.log_interval,
result_dir=results_dir)
PARSER = argparse.ArgumentParser()
PARSER.add_argument('--seq_len',
type=int,
action=check_size(0, 1000),
help="Number of time points in each input sequence",
default=200)
PARSER.add_argument(
'--log_interval',
type=int,
default=100,
action=check_size(0, 10000),
help="frequency, in number of iterations, after which loss is evaluated")
PARSER.add_argument('--results_dir',
type=validate_parent_exists,
help="Directory to write results to",
default=os.path.expanduser('~/results'))
PARSER.add_argument(
'--dropout',
type=float,
from ngraph.frontends.neon import Saver
import ngraph.transformers as ngt
from nlp_architect.models.kvmemn2n import KVMemN2N
from nlp_architect.data.wikimovies import WIKIMOVIES
from nlp_architect.utils.io import validate_parent_exists, check_size
from examples.kvmemn2n.interactive_util import interactive_loop
# parse the command line arguments
parser = NgraphArgparser(__doc__)
parser.add_argument(
'--emb_size',
type=int,
default='50',
help='Size of the word-embedding used in the model. (default 50)',
action=check_size(1, 20000))
parser.add_argument('--nhops',
type=int,
default='3',
help='Number of memory hops in the network',
action=check_size(1, 20))
parser.add_argument('--lr',
type=float,
default=0.01,
help='learning rate',
action=check_size(0, 5))
parser.add_argument('--subset',
type=str,
default='wiki-entities',
choices=['full', 'wiki-entities'],
help='wikiMovies dataset to use for training examples.')
create_data_dict,
)
from .matchlstm_ansptr import MatchLSTMAnswerPointer
from nlp_architect.utils.io import validate_existing_directory, check_size, validate_parent_exists
# Parse the command line arguments
parser = argparse.ArgumentParser()
parser.add_argument(
"--data_path",
default="data",
type=validate_existing_directory,
help="enter path for training data",
)
parser.add_argument("--gpu_id", default="0", type=str, help="enter gpu id", action=check_size(0, 8))
parser.add_argument(
"--max_para_req",
default=300,
type=int,
help="enter the max length of paragraph",
action=check_size(30, 300),
)
parser.add_argument(
"--epochs", default=15, type=int, help="enter the number of epochs", action=check_size(1, 30)
)
parser.add_argument(
"--select_device",
return target_word_vec1
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument('--gold_standard_file', default='data/goldStd.csv',
type=validate_existing_filepath,
help='path to gold standard file')
parser.add_argument('--word_embedding_model_file',
type=validate_existing_filepath,
default='pretrained_models/GoogleNews-vectors-negative300.bin',
help='path to the word embedding\'s model')
parser.add_argument('--training_to_validation_size_ratio', default=0.8, type=float,
action=check_size(0, 1), help='ratio between training and validation size')
parser.add_argument('--data_set_file', default='data/data_set.pkl',
type=validate_parent_exists,
help='path the file where the train, valid and test sets will be stored')
args = parser.parse_args()
# training set
X_train = []
y_train = []
# validation set
X_valid = []
y_valid = []
# 1. read GS file
[target_word_vec, definition_vec, hypernym_vec, label_vec] = \
print("Ensembling with weights: ")
for na in norm_accuracies:
print(na)
ensembled_predictions = simple_ensembler([lstm_predictions, one_hot_cnn_predictions],
norm_accuracies)
final_preds = np.argmax(ensembled_predictions, axis=1)
# Get the final accuracy
print(classification_report(np.argmax(Y_test, axis=1), final_preds,
target_names=data.labels_0.columns.values))
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--file_path', type=str, default='./',
help='file_path where the files to parse are located')
parser.add_argument('--data_type', type=str, default='amazon',
choices=['amazon'],
help='dataset source')
parser.add_argument('--epochs', type=int, default=10,
help='Number of epochs for both models', action=check_size(1, 20000))
args_in = parser.parse_args()
# Check file path
if args_in.file_path:
validate_existing_filepath(args_in.file_path)
if args_in.data_type == 'amazon':
data_in = Amazon_Reviews(args_in.file_path)
ensemble_models(data_in, args_in)
def clean_group(phrase_group):
"""
Returns the shortest element in a group of phrases
Args:
phrase_group (String): a group of phrases separated by ';'
Returns:
The shortest phrase in the group (String)
"""
text = [x.lstrip() for x in phrase_group.split(';')]
return min(text, key=len)
if __name__ == '__main__':
parser = argparse.ArgumentParser(prog='trend_analysis.py')
parser.add_argument('target_topics', metavar='target_topics', type=validate_existing_filepath,
help='a path to a csv topic-list extracted from the '
'target corpus')
parser.add_argument('ref_topics', metavar='ref_topics', type=validate_existing_filepath,
help='a path to a csv topic-list extracted from the '
'reference corpus')
parser.add_argument('--top_n', type=int, action=check_size(0, 100000), default=10000,
help='compare only top N topics (default: 10000)')
parser.add_argument('--top_vectors', type=int, action=check_size(0, 100000), default=500,
help='include only top N vectors in the scatter graph (default: 500)')
args = parser.parse_args()
analyze(args.target_topics, args.ref_topics, args.target_topics,
args.ref_topics, args.top_n, args.top_vectors)
np_list.append(np)
logger.info("np_list=%s", str(np_list))
return se.similarity(np_list, seed, args.similarity)
if __name__ == "__main__":
parser = argparse.ArgumentParser(prog='expand_server.py')
parser.add_argument('model_path',
metavar='model_path',
type=validate_existing_filepath,
help='a path to the w2v model file')
parser.add_argument('--host',
type=str,
default='localhost',
help='set port for the server',
action=check_size(1, 20))
parser.add_argument('--port',
type=int,
default=1234,
help='set port for the server',
action=check_size(0, 65535))
parser.add_argument('--grouping',
action='store_true',
default=False,
help='grouping mode')
parser.add_argument('--similarity',
default=0.5,
type=float,
action=check_size(0, 1),
help='similarity threshold')
parser.add_argument(
Args:
api (:obj:`falcon.api`): the Falcon API
service_name (str): the name of the service to init and load
"""
service = Service(service_name)
api.req_options.auto_parse_form_urlencoded = True
api.add_route('/{}'.format(service_name), service)
# api.router_options
path = os.path.abspath(os.path.join(os.path.dirname(__file__),
"web_service/visualizer/displacy"))
api.add_static_route('/{}'.format(service_name), path)
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--name', help="the name of the service you want to upload", type=str,
required=True, action=check_size(1, 30))
args = parser.parse_args()
app = application = falcon.API(middleware=[MultipartMiddleware()])
if not is_valid_input(args.name):
logger.error('ERROR: Invalid argument input for the server.')
sys.exit(0)
# init and load service
set_server_properties(app, args.name)
# run server:
port = 8080
server = make_server('0.0.0.0', port, app)
print('starting the server at port {0}'.format(port))
server.serve_forever()
return target_word_vec1
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument('--gold_standard_file', default='data/goldStd.csv',
type=validate_existing_filepath,
help='path to gold standard file')
parser.add_argument('--word_embedding_model_file',
type=validate_existing_filepath,
default='pretrained_models/GoogleNews-vectors-negative300.bin',
help='path to the word embedding\'s model')
parser.add_argument('--training_to_validation_size_ratio', default=0.8, type=float,
action=check_size(0, 1), help='ratio between training and validation size')
parser.add_argument('--data_set_file', default='data/data_set.pkl',
type=validate_parent_exists,
help='path the file where the train, valid and test sets will be stored')
args = parser.parse_args()
# training set
X_train = []
y_train = []
# validation set
X_valid = []
y_valid = []
# 1. read GS file
[target_word_vec, definition_vec, hypernym_vec, label_vec] = \
from nlp_architect.utils.io import validate, validate_existing_directory, \
validate_existing_filepath, validate_parent_exists, check_size
"""
Training script for reading comprehension model
"""
# parse the command line arguments
parser = NgraphArgparser(__doc__)
parser.add_argument('--data_path', help='enter path for training data',
type=str)
parser.add_argument('--gpu_id', default="0", help='enter gpu id',
type=str,action=check_size(0,10))
parser.add_argument('--max_para_req', default=100, help='enter the max length of paragraph',
type=int, action=check_size(30,300))
parser.add_argument('--batch_size_squad',default=16, help='enter the batch size',
type=int, action=check_size(1,256))
parser.set_defaults()
args = parser.parse_args()
os.environ["CUDA_VISIBLE_DEVICES"] = args.gpu_id
hidden_size = 150
gradient_clip_value = 15
embed_size = 300
import os
import numpy as np
from nlp_architect.utils.mrc_utils import (
create_squad_training, max_values_squad, get_data_array_squad, create_data_dict)
from nlp_architect.models.matchlstm_ansptr import MatchLSTM_AnswerPointer
import argparse
import tensorflow as tf
from nlp_architect.utils.io import validate_existing_directory, check_size, validate_parent_exists
# Parse the command line arguments
parser = argparse.ArgumentParser()
parser.add_argument('--data_path', default='data', type=validate_existing_directory,
help='enter path for training data')
parser.add_argument('--gpu_id', default="0", type=str,
help='enter gpu id', action=check_size(0, 8))
parser.add_argument('--max_para_req', default=300, type=int,
help='enter the max length of paragraph', action=check_size(30, 300))
parser.add_argument('--epochs', default=15, type=int,
help='enter the number of epochs', action=check_size(1, 30))
parser.add_argument('--select_device', default='GPU', type=str,
help='enter the device to execute on', action=check_size(3, 9))
parser.add_argument('--train_set_size', default=None, type=int,
help='enter the size of the training set', action=check_size(200, 90000))
parser.add_argument('--hidden_size', default=150, type=int,
help='enter the number of hidden units', action=check_size(30, 300))
help='boolean indicating whether the model to load has been stored in binary '
'format.',
action='store_true')
arg_parser.add_argument(
'--word_ngrams',
default=0,
type=int,
choices=[0, 1],
help='If 0, the model to load stores word information. If 1, the model to load stores '
'subword (ngrams) information; note that subword information is relevant only to '
'fasttext models.')
arg_parser.add_argument(
'--mark_char',
default='_',
type=str,
action=check_size(1, 2),
help='special character that marks word separator and NP suffix.')
arg_parser.add_argument(
'--np',
default='Intel Corp.',
type=str,
action=check_size(min=1),
help='NP to print its word vector.')
args = arg_parser.parse_args()
np2vec_model = NP2vec.load(
args.np2vec_model_file,
binary=args.binary,
word_ngrams=args.word_ngrams)
target_names=data.labels_0.columns.values))
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--file_path",
type=str,
default="./",
help="file_path where the files to parse are located")
parser.add_argument("--data_type",
type=str,
default="amazon",
choices=["amazon"],
help="dataset source")
parser.add_argument(
"--epochs",
type=int,
default=10,
help="Number of epochs for both models",
action=check_size(1, 20000),
)
args_in = parser.parse_args()
# Check file path
if args_in.file_path:
validate_existing_filepath(args_in.file_path)
if args_in.data_type == "amazon":
data_in = Amazon_Reviews(args_in.file_path)
ensemble_models(data_in, args_in)
12 1:puppet 1:sequences 1:were done by <NULL> Paska . 1:puppet 1:sequences
1:were Roman
13 1:done 1:by 1:Roman Paska . <NULL> for the 1:done 1:by 1:Roman Music
***** Misc *****
Notice that you don't get anything from the third line of data.txt, as there
are no recognized entities. The final ngram in each output line is preceded
by a tab in the actual output, though it's hard to discern above.
'''
parser = argparse.ArgumentParser(
description='Generates windowed examples for wikipedia files. By default,' +
' creates pairs of window<TAB>entity when used with entities.'
)
parser.add_argument('data_dir', type=str, help='name of root directory for files')
parser.add_argument('-n', type=int, help='Max number of examples to process.',
action=check_size(1, 100000000))
parser.add_argument('-e', '--entities', type=str,
help='entities file (each line specifies ngrams to always chunk together)')
parser.add_argument('-a', '--all_windows', action='store_true',
help='if set, keeps all windows (not just ones entities). defaults to ' +
' True if entities file not present, False if it is present.')
parser.add_argument('-m', '--movie_in_all', action='store_true',
help='if set, prepends movie to every line in example')
parser.add_argument('-i', '--inverse', action='store_true',
help='if set, also write "inversed" version of each fact to the kb')
parser.add_argument('-r', '--replace_centroids', action='store_true',
help='specifies whether to remove the center words of windows from ' +
'their windows (defaults false, if true replaces word with <NULL>)')
parser.add_argument('-dm', '--dontmerge', action='store_true',
help='default behavior merges lines from the same example--set this flag ' +
'to disable and only consider windows from the same line in the file')
metavar="target_topics",
type=validate_existing_filepath,
help="a path to a csv topic-list extracted from the "
"target corpus",
)
parser.add_argument(
"ref_topics",
metavar="ref_topics",
type=validate_existing_filepath,
help="a path to a csv topic-list extracted from the "
"reference corpus",
)
parser.add_argument(
"--top_n",
type=int,
action=check_size(0, 100000),
default=10000,
help="compare only top N topics (default: 10000)",
)
parser.add_argument(
"--top_vectors",
type=int,
action=check_size(0, 100000),
default=500,
help="include only top N vectors in the scatter graph (default: 500)",
)
args = parser.parse_args()
analyze(
args.target_topics,
args.ref_topics,
args.target_topics,
if __name__ == '__main__':
# read input args and validate
parser = argparse.ArgumentParser()
parser.add_argument(
'--input_file',
type=validate_existing_filepath,
required=True,
help='Input texts file path (samples to pass for inference)')
parser.add_argument('--model_name',
default='chunker_model',
type=str,
required=True,
help='Model name (used for saving the model)')
parser.add_argument('-b',
type=int,
action=check_size(1, 9999),
default=1,
help='inference batch size')
args = parser.parse_args()
model_path = path.join(path.dirname(path.realpath(__file__)),
'{}.h5'.format(str(args.model_name)))
settings_path = path.join(path.dirname(path.realpath(__file__)),
'{}.params'.format(str(args.model_name)))
validate_existing_filepath(model_path)
validate_existing_filepath(settings_path)
# load model and parameters
model = SequenceChunker()
model.load(model_path)
word_length = model.max_word_len
with open(settings_path, 'rb') as fp:
args.lr, num_iterations=num_iterations, log_interval=n_per_epoch,
result_dir=results_dir, ckpt=None)
else:
sequences = model.run_inference(args.ckpt, num_samples=args.num_samples,
sos=ptb_dict.sos_symbol, eos=ptb_dict.eos_symbol)
for seq in sequences:
sentence = []
for idx in seq:
while idx == ptb_dict.sos_symbol:
continue
sentence.append(ptb_dict.idx2word[idx])
print(" ".join(sentence) + "\n")
PARSER = argparse.ArgumentParser()
PARSER.add_argument('--seq_len', type=int, action=check_size(0, 1000),
help="Number of time points in each input sequence",
default=60)
PARSER.add_argument('--results_dir', type=validate_parent_exists,
help="Directory to write results to",
default=os.path.expanduser('~/results'))
PARSER.add_argument('--dropout', type=float, default=0.45, action=check_size(0, 1),
help='dropout applied to layers, value in [0, 1] (default: 0.45)')
PARSER.add_argument('--ksize', type=int, default=3, action=check_size(0, 10),
help='kernel size (default: 3)')
PARSER.add_argument('--levels', type=int, default=4, action=check_size(0, 10),
help='# of levels (default: 4)')
PARSER.add_argument('--lr', type=float, default=4, action=check_size(0, 100),
help='initial learning rate (default: 4)')
PARSER.add_argument('--nhid', type=int, default=600, action=check_size(0, 1000),
help='number of hidden units per layer (default: 600)')
***** Misc *****
Notice that you don't get anything from the third line of data.txt, as there
are no recognized entities. The final ngram in each output line is preceded
by a tab in the actual output, though it's hard to discern above.
'''
parser = argparse.ArgumentParser(
description='Generates windowed examples for wikipedia files. By default,'
+ ' creates pairs of window<TAB>entity when used with entities.')
parser.add_argument('data_dir',
type=str,
help='name of root directory for files')
parser.add_argument('-n',
type=int,
help='Max number of examples to process.',
action=check_size(1, 100000000))
parser.add_argument(
'-e',
'--entities',
type=str,
help='entities file (each line specifies ngrams to always chunk together)')
parser.add_argument(
'-a',
'--all_windows',
action='store_true',
help='if set, keeps all windows (not just ones entities). defaults to ' +
' True if entities file not present, False if it is present.')
parser.add_argument('-m',
'--movie_in_all',
action='store_true',
help='if set, prepends movie to every line in example')
'--corpus',
help='path to the input corpus. Compressed files (gz) are also supported. By default, '
'it is a subset of English Wikipedia. '
'get subset of English wikipedia from '
'https://github.com/NervanaSystems/nlp-architect/raw/'
'master/datasets/wikipedia/enwiki-20171201_subset.txt.gz')
arg_parser.add_argument(
'--marked_corpus',
default='enwiki-20171201_subset_marked.txt',
type=validate_parent_exists,
help='path to the marked corpus corpus.')
arg_parser.add_argument(
'--mark_char',
default='_',
type=str,
action=check_size(1, 2),
help='special character that marks NP\'s in the corpus (word separator and NP suffix). '
'Default value is _.')
arg_parser.add_argument(
'--grouping',
action='store_true',
default=False,
help='perform noun-phrase grouping')
arg_parser.add_argument(
'--chunker', type=str,
choices=['spacy', 'nlp_arch'],
default='spacy',
help='chunker to use for detecting noun phrases. \'spacy\' for using spacy built-in '
'chunker or \'nlp_arch\' for NLP Architect NP Extractor')
args = arg_parser.parse_args()
"""
Training script for reading comprehension model
"""
# parse the command line arguments
parser = NgraphArgparser(__doc__)
parser.add_argument('--data_path',
help='enter path for training data',
type=str)
parser.add_argument('--gpu_id',
default="0",
help='enter gpu id',
type=str,
action=check_size(0, 10))
parser.add_argument('--max_para_req',
default=100,
help='enter the max length of paragraph',
type=int,
action=check_size(30, 300))
parser.add_argument('--batch_size_squad',
default=16,
help='enter the batch size',
type=int,
action=check_size(1, 256))
parser.set_defaults()
