import sys
from params import Params
from dataset import qa
import keras.backend as K
import units
from loss import *
from models.match import keras as models
from params import Params
params = Params()
config_file = 'config/qalocal.ini' # define dataset in the config
params.parse_config(config_file)
reader = qa.setup(params)
qdnn = models.setup(params)
model = qdnn.getModel()
from loss import *
model.compile(loss=rank_hinge_loss({'margin': 0.2}),
optimizer=units.getOptimizer(name=params.optimizer,
lr=params.lr),
metrics=['accuracy'])
model.summary()
# generators = [reader.getTrain(iterable=False) for i in range(params.epochs)]
# [q,a,score] = reader.getPointWiseSamples()
# model.fit(x = [q,a,a],y = [q,a,q],epochs = 10,batch_size =params.batch_size)
# def gen():
# while True:
def config_adapter(conf):
conf.text1_maxlen = conf.max_sequence_length
conf.embed = conf.lookup_table
conf.vocab_size = len(conf.alphabet)
conf.hidden_sizes = [20, 1]
conf.num_layers = 2
conf.embed_size = conf.embedding_size
conf.bin_num =1
conf.target_mode = "ranking"
if __name__ == "__main__":
from params import Params
params = Params()
config_file = 'config/qalocal.ini' # define dataset in the config
params.parse_config(config_file)
params.network_type = "anmm.ANMM"
from dataset import qa
reader = qa.setup(params)
from models.match import keras as models
model = models.setup(params)