def predict_sentence(sentence, nn_model, w2v_model, index_to_token, diversity=0.5):
input_sequence = tokenize(sentence + ' ' + EOS_SYMBOL)
tokens_sequence = _predict_sequence(input_sequence, nn_model, w2v_model, index_to_token, diversity)
predicted_sentence = ' '.join(tokens_sequence)
return predicted_sentence
def get_tokenized_dialog_lines(iterable_dialog_lines):
for line in iterable_dialog_lines:
tokenized_dialog_line = tokenize(line)
tokenized_dialog_line.append(EOS_SYMBOL)
yield tokenized_dialog_line
def get_tokenized_dialog_lines(iterable_dialog_lines):
for line in iterable_dialog_lines:
tokenized_dialog_line = tokenize(line)
tokenized_dialog_line.append(EOS_SYMBOL)
yield tokenized_dialog_line
eeg_topo_data[i, :, 4, 0] = data[:, 4]
eeg_topo_data[i, :, 6 ,0] = data[:, 5]
eeg_topo_data[i, :, 8, 3] = data[:, 6]
eeg_topo_data[i, :, 8, 5] = data[:, 7]
eeg_topo_data[i, :, 6, 8] = data[:, 8]
eeg_topo_data[i, :, 4, 8] = data[:, 9]
eeg_topo_data[i, :, 3, 7] = data[:, 10]
eeg_topo_data[i, :, 2, 6] = data[:, 11]
eeg_topo_data[i, :, 2, 8] = data[:, 12]
eeg_topo_data[i, :, 1, 5] = data[:, 13]
ground_truth_labels = torch.zeros(ground_truth.shape[0])
labels = []
for i, data in enumerate(ground_truth):
ground_truth_labels[i], name = tokenize(data)
labels.append(name)
print(sorted(Counter(list(labels)).items()))
# create dataset, split dataset to train_set and test_set
dataset = Dataset(eeg_topo_data, ecg_data, gsr_data, ground_truth_labels)
test_split = 0.3
num_data = len(dataset)
num_test = int(num_data*test_split)
num_train = num_data - num_test
train_set, test_set = torch.utils.data.random_split(dataset, [num_train, num_test])
train_data = torch.utils.data.DataLoader(train_set, batch_size=128, shuffle=True, num_workers=2)
