decoder = Decoder(14, 512, num_layers=3).cuda()
learning_rate = 0.001
teacher_forcing_ratio = 0.3 # value doesn't matter; immediately overwritten
# create encoder outputs
# given some input sequence - input
# load data
data = []
dat = open(sys.argv[1])
for l in dat:
p = l.split()
targ_prog = p[0]
targ_prog_f = e.eval_init(targ_prog)
in_data = [int(p[1][2 * b:2 * b + 2], base=16) for b in range(256)]
data.append((targ_prog, in_data, targ_prog_f))
dat.close()
# shuffled lists for data sampling
in_sample = range(256)
data_sample = range(len(data))
# test accuracy on full data
TEST_ACC_SAMP = 30
start = time.time()
for i in range(N):
test.sess_open(randos[i], progs[i])
opened = time.time() - start
for i in order:
assert test.cand_query(randos[i], progs[i]) == (100000.)
done = time.time() - start
print "open after "
print opened
print "done after "
print done
print test.read_odom()
order = random.sample(range(N), N)
queries = []
for i in range(N):
queries += [test.eval_init(progs[i])]
for i in order:
assert queries[i](progs[i]) == (100000.)
for i in range(4):
print queries[order[i]](progs[i])
