use_locking=False, name='Adam').\
minimize(loss)
init = tf.global_variables_initializer()
saver = tf.train.Saver()
t00 = datetime.datetime.now()
with tf.Session() as sess:
sess.run(init)
if renetFile:
saver.restore(sess, renetFile)
str1 = ''
for a in range(iterations):
tr_pre_data = tr.prepare()
tr_loss, tr_median = run_data_stack_avg2(tr_pre_data, input_dic,
sess, xy, stack)
te_pre_data = te.prepare()
te_loss, te_median = run_data_stack_avg2(te_pre_data, input_dic,
sess, xy, stack)
t1 = datetime.datetime.now()
t1 = datetime.datetime.now()
str = "it: {0:.2f} {1:.2f} ".format(
a * loop / 1000.0, (t1 - t00).total_seconds() / 3600.0)
for s in range(stack + 1):
str += " {0:.3f} {1:.3f} {2:.3f} {3:.3f} ".format(
tr_loss[s], te_loss[s], tr_median[s], te_median[s])
sess.run(init)
if test is not None:
saver.restore(sess, cfg.netFile)
mul = -1
if len(sys.argv) > 2:
mul = int(sys.argv[2])
run_test(input_dic, sess, xy, te, cfg, mul)
if cfg.renetFile:
saver.restore(sess, cfg.renetFile)
str1 = ''
for a in range(iterations):
tr_pre_data = tr0.prepare()
tr_loss, tr_median = run_data(tr_pre_data, input_dic, sess, xy,
'tr', cfg)
te_pre_data = te.prepare()
te_loss, te_median = run_data(te_pre_data, input_dic, sess, xy,
'te', cfg)
t1 = datetime.datetime.now()
str = "it: {0:.3f} {1:.3f} {2:4.2e}".\
format(a*loop/1000.0, (t1 - t00).total_seconds()/3600.0, lr)
s = -1
while True:
s += len(tr_loss) / 2
str += " {0:.3f} {1:.3f} {2:.3f} {3:.3f} ".format(
tr_loss[s], te_loss[s], tr_median[s], te_median[s])
# use_locking=False, name='Adam').\
# minimize(ls[0])
init = tf.global_variables_initializer()
saver = tf.train.Saver()
t00 = datetime.datetime.now()
with tf.Session() as sess:
sess.run(init)
if cfg.renetFile:
saver.restore(sess, cfg.renetFile)
str1 = ''
for a in range(iterations):
tr_pre_data = tr0.prepare()
tr_loss, tr_median = run_data_stack_avg3(tr_pre_data, input_dic,
sess, xy, 'tr', att)
te_pre_data = te.prepare()
te_loss, te_median = run_data_stack_avg3(te_pre_data, input_dic,
sess, xy, 'te', att)
t1 = datetime.datetime.now()
str = "it: {0:.3f} {1:.3f}".format(
a * loop / 1000.0, (t1 - t00).total_seconds() / 3600.0)
s = 0
while True:
# for s in range(0, feature_len+1, 5 ):
if s > cfg.feature_len:
s = cfg.feature_len
if len(sys.argv) > 2:
mul = int(sys.argv[2])
saver.restore(sess, cfg.netFile)
run_test(input_dic, sess, xy, te, cfg, mul)
if cfg.renetFile:
saver.restore(sess, cfg.renetFile)
str1 = ''
for a in range(iterations):
t1 = datetime.datetime.now()
str = "it: {0:.3f} {1:.3f} {2:4.2e}".\
format(a*loop/1000.0, (t1 - t00).total_seconds()/3600.0, lr)
tr_pre_data = tr0.prepare()
tr_loss, tr_median = run_data_0(tr_pre_data, input_dic, sess, xy,
'tr', cfg)
te_pre_data = te.prepare()
te_loss, te_median = run_data_0(te_pre_data, input_dic, sess, xy,
'te', cfg)
s = -1
while True:
s += len(tr_loss) / 2
str += " {0:.3f} {1:.3f} {2:.3f} {3:.3f} ".format(
tr_loss[s], te_loss[s], tr_median[s], te_median[s])
if s == len(tr_loss) - 1:
break
cfg = Config(config_file)
tr = DataSet(cfg.tr_data, cfg.memory_size, cfg.feature_len)
te = DataSet(cfg.te_data, cfg.memory_size, cfg.feature_len)
tr.set_net_type(cfg.net_type)
te.set_net_type(cfg.net_type)
tr.set_t_scale(cfg.t_scale)
te.set_t_scale(cfg.t_scale)
tr.set_num_output(cfg.num_output)
te.set_num_output(cfg.num_output)
att = te.sz[1]
D_in = cfg.feature_len * att
D_out = cfg.num_output
tr_pre_data = tr.prepare(multi=1)
for b in tr_pre_data:
d = torch.from_numpy(b[0]).type(torch.FloatTensor)
t = torch.from_numpy(b[1]).type(torch.FloatTensor)
te_pre_data = te.prepare(multi=1)
for b in te_pre_data:
de = torch.from_numpy(b[0]).type(torch.FloatTensor)
te = torch.from_numpy(b[1]).type(torch.FloatTensor)
H = 200
# Create random Tensors to hold input and outputs.
#d = torch.randn(N, D_in, device=device, dtype=dtype)
#t = torch.randn(N, D_out, device=device, dtype=dtype)
xy = {}
for a in range(base + 1):
xy[a] = net.layers['output{}'.format(a)]
init = tf.global_variables_initializer()
saver = tf.train.Saver()
t00 = datetime.datetime.now()
multi = -1
with tf.Session() as sess:
saver.restore(sess, renetFile)
rst_dic = {}
truth_dic = {}
for _ in range(loop):
te_pre_data = te.prepare(rd=False, multi=multi)
for b in te_pre_data:
length = b[0].shape[1] / base
feed = {} # inputs['input0']: b[0][:, :length]}
for a in range(base):
feed[input_dic['input{}'.format(a)]] = b[0][:, length *
a:length *
(a + 1)]
result = []
for a in range(base + 1):
r = sess.run(xy[a], feed_dict=feed)
result.append(r)
result = np.array(result)
for a in range(len(b[2])):
if not b[2][a] in rst_dic: