def generate(self, input_variable, batch_size):
input_variable = input_variable.view(batch_size, -1)
encoder_hidden = self.encoder.init_hidden(batch_size)
encoder_outputs, encoder_hidden = self.encoder(input_variable,
encoder_hidden)
decoder_hidden = encoder_hidden
decoder_inputs = [(Variable(torch.LongTensor([[SOS_token]])),
decoder_hidden)] # SOS
beam = BeamSearch(self.vocab_size2, self.beam_size, decoder_hidden)
# loop beam search
for di in xrange(self.target_length):
decoder_outputs = []
for decoder_input, decoder_hidden in decoder_inputs:
decoder_output, decoder_hidden, _ = self.decoder(
decoder_input, decoder_hidden,
encoder_outputs) # SOS + Predict
decoder_outputs.append((decoder_output, decoder_hidden))
decoder_inputs = beam.beam_search(decoder_outputs)
return beam.generate(self.generate_num)
for i in range(100):
sat = f.generate_sat(n, m, k)
# print("\nThe generated 3-sat problem is: ", sat)
# Hill Climbing
# print("\nHill Climbing")
state, boolval = hc.hill_climbing(f, sat, initial_state, bits_changed, m)
# printS(state,boolval)
if boolval:
pf[0] += 1
# pf.append(boolval)
#Beam Search
# print("\nBeam Search - 3")
state, boolval = bs.beam_search(f, sat, initial_state, bits_changed, m,
beam_widths[0])
# printS(state,boolval)
if boolval:
pf[1] += 1
# pf.append(boolval)
# print("\nBeam Search - 4")
state, boolval = bs.beam_search(f, sat, initial_state, bits_changed, m,
beam_widths[1])
# printS(state,boolval)
if boolval:
pf[2] += 1
# pf.append(boolval)
# Variable Neighbourhood Descent
# print("\nVariable Neighbourhood Descent")
state, boolval = vn.var_neigh(f, sat, initial_state, var_functions, m)
