def prepare_data(root_path):
param = get_params(root_path,
os.path.join(root_path, 'text_task_resnet'),
test_state=None)
test_dir_loc = os.path.join(root_path, "history")
dump_dir_loc = param['dump_dir_loc']
vocab_file = param['vocab_file']
vocab_stats_file = param['vocab_stats_file']
vocab_freq_cutoff = param['vocab_freq_cutoff']
test_data_file = param['test_data_file']
max_utter = param['max_utter']
max_len = param['max_len']
max_images = param['max_images']
preparedata = PrepareData(max_utter,
max_len,
max_images,
start_symbol_index,
end_symbol_index,
unk_symbol_index,
pad_symbol_index,
"text",
cutoff=vocab_freq_cutoff)
if os.path.isfile(vocab_file):
print 'found existing vocab file in ' + str(
vocab_file) + ', ... reading from there'
preparedata.prepare_data(test_dir_loc,
vocab_file,
vocab_stats_file,
os.path.join(dump_dir_loc, "test_smallest"),
test_data_file,
isTrain=False,
isTest=True)
def get_dialog_dict(param, is_test=False):
train_dir_loc = param['train_dir_loc']
valid_dir_loc = param['valid_dir_loc']
test_dir_loc = param['test_dir_loc']
dump_dir_loc = param['dump_dir_loc']
vocab_file = param['vocab_file']
vocab_stats_file = param['vocab_stats_file']
vocab_freq_cutoff = param['vocab_freq_cutoff']
train_data_file = param['train_data_file']
valid_data_file = param['valid_data_file']
test_data_file = param['test_data_file']
max_utter = param['max_utter']
max_len = param['max_len']
max_images = param['max_images']
max_negs = param['max_negs']
if 'test_state' in param:
test_state = param['test_state']
else:
test_state = None
preparedata = PrepareData(max_utter, max_len, max_images, max_negs, start_symbol_index, end_symbol_index,
unk_symbol_index, pad_symbol_index, "image", cutoff=vocab_freq_cutoff)
if os.path.isfile(vocab_file):
print('found existing vocab file in ' + str(vocab_file) + ', ... reading from there')
if not is_test:
preparedata.prepare_data(train_dir_loc, vocab_file, vocab_stats_file, os.path.join(dump_dir_loc, "train"),
train_data_file, True, False, None)
preparedata.prepare_data(valid_dir_loc, vocab_file, vocab_stats_file, os.path.join(dump_dir_loc, "valid"),
valid_data_file, False, False, None)
if test_state is not None:
preparedata.prepare_data(test_dir_loc, vocab_file, vocab_stats_file,
os.path.join(dump_dir_loc + "/test_data_file_state/", "test_" + test_state),
test_data_file, False, True, test_state)
else:
preparedata.prepare_data(test_dir_loc, vocab_file, vocab_stats_file, os.path.join(dump_dir_loc, "test"),
test_data_file, False, True, test_state)
def get_dialog_dict_for_test(param):
test_dir_loc = param['test_dir_loc']
dump_dir_loc = param['dump_dir_loc']
vocab_file = param['vocab_file']
vocab_stats_file = param['vocab_stats_file']
response_vocab_file = param['response_vocab_file']
vocab_freq_cutoff = param['vocab_freq_cutoff']
test_data_file = param['test_data_file']
max_utter = param['max_utter']
max_len = param['max_len']
stopwords = param['stopwords']
stopwords_histogram = param['stopwords_histogram']
max_mem_size = param['memory_size']
max_target_size = param['gold_target_size']
ques_type_id = param['ques_type_id']
ques_type_name = param['ques_type_name']
vocab_max_len = param['vocab_max_len']
wikidata_dir = param['wikidata_dir']
lucene_dir = param['lucene_dir']
transe_dir = param['transe_dir']
glove_dir = param['glove_dir']
preparedata = PrepareData(max_utter, max_len, start_symbol_index, end_symbol_index, unk_symbol_index,
pad_symbol_index, kb_pad_idx, nkb, stopwords, stopwords_histogram, lucene_dir, transe_dir,
wikidata_dir, glove_dir, max_mem_size, max_target_size, vocab_max_len, True,
cutoff=vocab_freq_cutoff)
if os.path.isfile(vocab_file):
print 'found existing vocab file in ' + str(vocab_file) + ', ... reading from there'
print 'to delete later ', os.path.join(dump_dir_loc, "train")
preparedata.prepare_data(test_dir_loc, vocab_file, vocab_stats_file, response_vocab_file,
os.path.join(dump_dir_loc, "test"), test_data_file, ques_type_id, ques_type_name)
def read_data(root_path):
ImageUrlToIndex = pkl.load(
open(path.join(root_path, 'data', 'Img_Fea_Dic.pkl')))
ImageFea = np.load(open(path.join(root_path, 'data', 'Img_Fea.npy')))
param = get_params(sys.argv[1])
train_dir_loc = param['train_dir_loc']
valid_dir_loc = param['valid_dir_loc']
test_dir_loc = param['test_dir_loc'].replace('test', 'test_smallest')
dump_dir_loc = param['dump_dir_loc']
vocab_file = param['vocab_file']
vocab_stats_file = param['vocab_stats_file']
vocab_freq_cutoff = param['vocab_freq_cutoff']
train_data_file = param['train_data_file']
valid_data_file = param['valid_data_file']
test_data_file = param['test_data_file'].replace('test', 'test_smallest')
# print test_data_file
# sys.exit(1)
max_utter = param['max_utter']
max_len = param['max_len']
max_images = param['max_images']
preparedata = PrepareData(max_utter,
max_len,
max_images,
start_symbol_index,
end_symbol_index,
unk_symbol_index,
pad_symbol_index,
"text",
cutoff=vocab_freq_cutoff)
if os.path.isfile(vocab_file):
print 'found existing vocab file in ' + str(
vocab_file) + ', ... reading from there'
preparedata.prepare_data(test_dir_loc, vocab_file, vocab_stats_file,
os.path.join(dump_dir_loc, "test_smallest"),
test_data_file)
train_dir_loc = param['train_dir_loc']
valid_dir_loc = param['valid_dir_loc']
test_dir_loc = param['test_dir_loc'].replace('test', 'test_smallest')
dump_dir_loc = param['dump_dir_loc']
vocab_file = param['vocab_file']
vocab_stats_file = param['vocab_stats_file']
vocab_freq_cutoff = param['vocab_freq_cutoff']
train_data_file = param['train_data_file']
valid_data_file = param['valid_data_file']
test_data_file = param['test_data_file'].replace('test', 'test_smallest')
#print test_data_file
#sys.exit(1)
max_utter = param['max_utter']
max_len = param['max_len']
max_images = param['max_images']
preparedata = PrepareData(max_utter,
max_len,
max_images,
start_symbol_index,
end_symbol_index,
unk_symbol_index,
pad_symbol_index,
"text",
cutoff=vocab_freq_cutoff)
if os.path.isfile(vocab_file):
print 'found existing vocab file in ' + str(
vocab_file) + ', ... reading from there'
preparedata.prepare_data(test_dir_loc, vocab_file, vocab_stats_file,
os.path.join(dump_dir_loc, "test_smallest"),
test_data_file)
