def re_evalute(CONST):
threshold = float(CONST.THRESHOLD)
from util.utils import load_pickle as load
gru = load(CONST.DATA_DIR + "GRU" + "predictions")
lstm = load(CONST.DATA_DIR + "LSTM" + "predictions")
biRNN = load(CONST.DATA_DIR + "BiRNN" + "predictions")
print("########### GRU ###############")
find_best_slot1("gru", gru["predictions"], gru["y"])
print("########### LSTM ##############")
find_best_slot1("lstm", lstm["predictions"], lstm["y"])
print("########### BiRNN LSTM ##############")
find_best_slot1("birnn", biRNN["predictions"], biRNN["y"])
def make_loss_graph(CONST):
from util.utils import load_pickle as load
gru = load(CONST.DATA_DIR + 'GRU')
lstm = load(CONST.DATA_DIR + 'LSTM')
birnn = load(CONST.DATA_DIR + 'BiRNN')
cutoff = 105
gru = gru[:cutoff]
lstm = lstm[:cutoff]
birnn= birnn[:cutoff]
import matplotlib.pyplot as plt
x = [i for i in range(cutoff)]
plt.plot(np.array(x), np.array(gru))
plt.plot(np.array(x), np.array(lstm))
plt.plot(np.array(x), np.array(birnn))
plt.legend(['GRU', 'LSTM', 'BiRNN LSTM'], loc='upper right')
plt.savefig("losses_all.png")
print("all losses plot saved")
from util.heatmap import avg_distance_and_heatmaps
avg_distance_and_heatmaps(alphas, sentences,
CONST.SENTENCE_PATH + "hursh/")
if __name__ == "__main__":
# Set and Overload Arguments
CONST.parse_argument(argparse.ArgumentParser())
# Set Time of Experiment
now = datetime.datetime.now()
time_stamp = "_".join([
str(a) for a in [now.month, now.day, now.hour, now.minute, now.second]
])
data = load(CONST.DATA_DIR + CONST.DATA_FILE)
"""
x_train = data["x_train"]
x_dev = data["x_dev"]
x_test = data["x_test"]
y_train = data["y_train"]
y_dev = data["y_dev"]
y_test = data["y_test"]
l_train = data["l_train"]
l_dev = data["l_dev"]
l_test = data["l_test"]
train_sentences = data["train_sentences"]
dev_sentences = data["dev_sentences"]
test_sentences = data["test_sentences"]
embeddings = data["embeddings"]
aspects = data["aspects"]
def re_evalute_slot3(CONST):
from util.utils import load_pickle as load
birnn = load(CONST.DATA_DIR + "BiRNNslot3" + "_predictions")
evaluate_multiclass(birnn["predictions"], birnn["y"], True)
