def sample(model, song, config):
"""
# This function takes the desired output length and input characters as arguments, returning the produced sentence
:param config: Dict of settings
:param model: nn.Module
:param song: String
:param limit: Int
:return: String (new generated song)
"""
char_to_idx, idx_to_char = char_mapping()
model.eval()
i = 0
while song[-1] != '%' and i < config["LIMIT_LEN"]:
char = predict(model, song, config, char_to_idx, idx_to_char)
song += char
i += 1
return song
print('Number of development samples:' + str(len(dev_set)))
print("Number of node: " + str(len_node) + ", while max allowed is " + str(options.max_node_num))
print("Number of parent node: " + str(len_in_node) + ", truncated to " + str(options.max_in_node_num))
print("Number of child node: " + str(len_out_node) + ", truncated to " + str(options.max_out_node_num))
print("The entity size: " + str(entity_size) + ", truncated to " + str(options.max_entity_size))
# Build dictionary and mapping of words, characters, edges
words, chars, edges = collect_data(train_set)
print('Number of words:' + str(len(words)))
print('Number of characters:' + str(len(chars)))
print('Number of edges:' + str(len(edges)))
dict_word, word_to_id, id_to_word = word_mapping(words)
dict_char, char_to_id, id_to_char = char_mapping(chars)
dict_edge, edge_to_id, id_to_edge = edge_mapping(edges)
options.word_to_id = word_to_id
options.char_to_id = char_to_id
options.edge_to_id = edge_to_id
if options.binary_classification:
options.relation_num = 2
else:
options.relation_num = 6
train_set = get_dataset_from_instances(train_set, word_to_id, char_to_id, edge_to_id, options)
dev_set = get_dataset_from_instances(dev_set, word_to_id, char_to_id, edge_to_id, options)
# Build dataloader of training set and development set
import torch
from torch.nn import CrossEntropyLoss
from models import LSTMSimple, VanillaRNN
from utils import read_songs_from, char_mapping, encode_songs, get_device, negative_log_likelihood
from datetime import datetime
from train import fit
from plotting import save_loss_graph
def load_data(file):
songs = read_songs_from('data/' + file)
songs_encoded = encode_songs(songs, char_to_idx)
return songs, songs_encoded
char_to_idx, idx_to_char = char_mapping()
train, train_encoded = load_data('train.txt')
val, val_encoded = load_data('val.txt')
test, test_encoded = load_data('test.txt')
config = {
"EPOCHS": 15,
"CHUNK_SIZE": 100,
"VOCAB_SIZE": len(char_to_idx.keys()),
"LR": 0.001, # Default in Adam 0.001,
"WEIGHT_DECAY": 0, # Default in Adam 0
"HIDDEN": 100,
# For songs sampling
"TEMPERATURE": 1,
# mapping of words frenquency decreasing
dico_words_train = word_mapping(train_sentences,
parameters["lower"])[0]
dico_words, word_to_id, id_to_word = augment_with_pretrained(
dico_words_train.copy(), wordmodel,
list(
itertools.chain.from_iterable([[w[0] for w in s]
for s in dt_sentences]))
if not parameters['all_emb'] else None)
else:
dico_words, word_to_id, id_to_word = word_mapping(
train_sentences, parameters["lower"])
dico_words_train = dico_words
# Create a dictionary and a mapping for words / POS tags / tags
dico_chars, char_to_id, id_to_char = char_mapping(train_sentences)
dico_tags, tag_to_id, id_to_tag = tag_mapping(train_sentences)
dico_pts, pt_to_id, id_to_pt = pt_mapping(train_sentences + dev_sentences)
if not os.path.exists(os.path.join(models_path, model_name)):
os.makedirs(os.path.join(models_path, model_name))
save_mappings(os.path.join(models_path, model_name,
'mappings.pkl'), word_to_id, char_to_id,
tag_to_id, pt_to_id, dico_words, id_to_tag)
else:
word_to_id, char_to_id, tag_to_id, pt_to_id, dico_words, id_to_tag = reload_mappings(
os.path.join(models_path, model_name, 'mappings.pkl'))
dico_words_train = dico_words
id_to_word = {v: k for k, v in word_to_id.items()}
# Index sentences
m3 = 0
