feature_len = js['feature']
reference = js["nodes_reference"]
base_nodes = map(int, js["nodes_base"].split(','))
stack_nodes = map(int, js["nodes_stack"].split(','))
final_nodes = [] #map(int, js["nodes_final"].split(','))
num_att = 4
num_output = 3
renetFile = None
if 'retrain' in js:
renetFile = HOME + 'NNs/' + js['retrain'] + '.p'
if mode == 'train':
tr = DataSet(tr_data, js['block'], feature_len)
# tr.set_t_scale(t_scale)
tr.set_num_output(num_output)
te = DataSet(te_data, js['block'], feature_len)
# te.set_t_scale(t_scale)
te.set_num_output(num_output)
else:
if mode == 'te':
te = DataSet(te_data, js['block'], feature_len)
else:
te = DataSet(tr_data, js['block'], feature_len)
te.set_num_output(num_output)
sz_in = te.sz
loop = js['loop']
print "input shape", sz_in, "LR", lr, 'feature', feature_len
netFile = HOME + 'NNs/' + js['net'] + '/fc'
batch_size = js['batch_size']
feature_len = js['feature']
lr = js['lr']
stack = js['stack']
num_output = js["num_output"]
step = js["step"]
stage = js["stage"]
renetFile = None
if 'retrain' in js:
renetFile = HOME + 'NNs/' + js['retrain'] + '/fc'
tr = DataSet(tr_data, batch_size, feature_len + stack)
te = DataSet(te_data, batch_size, feature_len + stack)
att = te.sz[1]
iterations = 10000
loop = 40
print "input attribute", att, "LR", lr, 'feature', feature_len
inputs = {}
inputs[0] = tf.placeholder(tf.float32, [None, feature_len * att])
output = tf.placeholder(tf.float32, [None, num_output])
for a in range(stack):
inputs[a + 1] = tf.placeholder(tf.float32, [None, att])
input_dic = {}
js = Utils.load_json_file(config_file, False)
te_data = []
for key in js:
if key.startswith(data_type):
te_data.append(HOME + js[key])
netFile = HOME + 'NNs/' + js['netTest'] + '/fc'
batch_size = js['batch_size']
feature_len = js['feature']
stack = js['stack']
num_output = js["num_output"]
te = DataSet(te_data, batch_size, feature_len + stack)
att = te.sz[1]
print "input shape", att, 'feature', feature_len
inputs = {}
inputs[0] = tf.placeholder(tf.float32, [None, feature_len * att])
output = tf.placeholder(tf.float32, [None, num_output])
for a in range(stack):
inputs[a + 1] = tf.placeholder(tf.float32, [None, att])
input_dic = {}
for a in range(stack + 1):
input_dic['input{}'.format(a)] = inputs[a]
netFile = HOME + 'NNs/' + js['net'] + '/p1'
batch_size = int(js['batch_size'])
feature_len = int(js['feature'])
lr = float(js['lr'])
step = js['step']
num_output = 3
nodes1 = map(int, js["nodes1"].split(','))
nodes2 = map(int, js["nodes2"].split(','))
nodes3 = map(int, js["nodes3"].split(','))
renetFile = None
if 'retrain' in js:
renetFile = HOME + 'NNs/' + js['retrain'] + '/p1'
tr = DataSet(tr_data, batch_size, feature_len + 1)
te = DataSet(te_data, batch_size, feature_len + 1)
att = te.sz[1]
iterations = 100000
loop = 2
if step == 0:
loop = 200
print "input shape", att, "LR", lr, 'feature', feature_len
net = P1Net(nodes1, nodes2, nodes3, att, num_output, feature_len, lr)
net.step = step
with tf.Session() as sess:
sess.run(net.init)
if renetFile:
if __name__ == '__main__':
config_file = "rnn_config.json"
#if len(sys.argv)>1:
# config_file = sys.argv[1]
test = None
if len(sys.argv) > 1:
test = sys.argv[1]
cfg = Config(config_file)
if test is None:
tr = DataSet(cfg.tr_data, cfg)
te = DataSet(cfg.te_data, cfg, sub_sample=0.15)
tr0 = DataSet([cfg.tr_data[0]], cfg, sub_sample=0.15)
cfg.att = te.sz[1]
else:
if test == 'te':
te = DataSet([cfg.te_data[0]], cfg)
else:
te = DataSet([cfg.tr_data[0]], cfg)
cfg.att = te.sz[1]
iterations = 10000
loop = cfg.loop
print "input attribute", cfg.att, "LR", cfg.lr, \
'feature', cfg.feature_len, 'add', cfg.add_len
batch_size = int(js['batch_size'])
feature_len = int(js['feature'])
lr = float(js['lr'])
step = js['step']
nAddition = 10
num_output = 3
nodes1 = map(int, js["nodes1"].split(','))
nodes2 = map(int, js["nodes2"].split(','))
nodes3 = map(int, js["nodes3"].split(','))
renetFile = None
if 'retrain' in js:
renetFile = HOME + 'NNs/' + js['retrain'] + '/p1'
tr = DataSet(tr_data, batch_size, feature_len, nadd=nAddition)
te = DataSet(te_data, batch_size, feature_len, nadd=nAddition)
att = te.sz[1]
iterations = 100000
loop = 100
# if step==0:
# loop=200
print "input shape", att, "LR", lr, 'feature', feature_len
net = P1Net(nodes1, nodes2, nodes3, att, num_output, feature_len, lr,
nAddition)
net.step = step
with tf.Session() as sess:
sess.run(net.init)
return pathname + '_' + basename + '_avg.p'
if __name__ == '__main__':
config_file = "rnn_config_4.json"
test = None
if len(sys.argv) > 1:
test = sys.argv[1]
cfg = Config(config_file)
avg_file = avg_file_name(cfg.netFile)
if test is None:
tr = DataSet(cfg.tr_data, cfg)
get_avg_file(tr, avg_file)
te = DataSet(cfg.te_data, cfg, sub_sample=0.15)
tr0 = DataSet([cfg.tr_data[0]], cfg, sub_sample=0.1)
cfg.att = te.sz[1]
tr.avg_correction(avg_file)
tr0.avg_correction(avg_file)
else:
if test == 'te':
te = DataSet([cfg.te_data[0]], cfg)
else:
te = DataSet([cfg.tr_data[0]], cfg)
cfg.att = te.sz[1]
te.avg_correction(avg_file)
parser.add_argument('-t', '--test', help='test', required=False)
args = parser.parse_args()
config_file = "config.json" if args.config is None else args.config
mode = 'train' if args.test is None else args.test
iterations = 10000
js = Utils.load_json_file(config_file)
dtype = torch.float
device = torch.device("cpu")
# device = torch.device("cuda:0") # Uncomment this to run on GPU
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)
batch_size = js['batch_size']
feature_len = js['feature']
lr = js['lr']
#stack = js['stack']
num_output = js["num_output"]
step = js["step"]
#stage = js["stage"]
t_scale = js['t_scale']
#net_type = js['net_type']
renetFile = None
if 'retrain' in js:
renetFile = HOME + 'NNs/' + js['retrain'] + '/fc'
tr = DataSet(tr_data, batch_size, feature_len)
tr0 = DataSet([tr_data[0]], batch_size, feature_len)
te = DataSet(te_data, batch_size, feature_len)
tr0.set_t_scale(t_scale)
tr.set_t_scale(t_scale)
te.set_t_scale(t_scale)
tr0.set_num_output(num_output)
tr.set_num_output(num_output)
te.set_num_output(num_output)
att = te.sz[1]
iterations = 10000
loop = js["loop"]
print "input attribute", att, "LR", lr, 'feature', feature_len
inputs = {}
if key.startswith('te'):
te_data.append(HOME + js['te'])
netFile = HOME + 'NNs/' + js['net'] + '/fc'
batch_size = js['batch_size']
feature_len = js['feature']
base = js['base']
step = js["step"]
loop = 1
num_output = js['num_output']
t_scale = js['t_scale']
renetFile = HOME + 'NNs/' + js['netTest'] + '/fc'
te = DataSet(te_data, batch_size, feature_len * base)
te.set_t_scale(t_scale)
te.set_num_output(num_output)
att = te.sz[1]
print "input attribute", att, 'feature', feature_len
inputs = {}
input_dic = {}
output = tf.placeholder(tf.float32, [None, num_output])
for a in range(base):
inputs[a] = tf.placeholder(tf.float32, [None, att * feature_len])
input_dic['input{}'.format(a)] = inputs[a]
net = PyraNet(input_dic)