def main():
arch = cfg.arch
model = build_network()
if torch.cuda.device_count() > 1:
model = nn.DataParallel(model)
print("now we are using %d gpus" %torch.cuda.device_count())
model.cuda()
print model
# load the model
print("=> Loading Network %s" % cfg.resume)
checkpoint = torch.load(cfg.resume)
model.load_state_dict(checkpoint['state_dict'])
print("=> loaded checkpoint '{}' (epoch {})"
.format(cfg.resume, checkpoint['epoch']))
cudnn.benchmark = False
test_loader = data_loader( BatchSize=cfg.batch_size,
NumWorkers = cfg.num_workers).test_loader
print("test data_loader are ready!")
# test mode
model.eval()
# test an image
# load the image transformer
centre_crop = trn.Compose([
trn.Resize((224,224)),
trn.CenterCrop(224),
trn.ToTensor(),
trn.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])
# fully connected layers
FC = []
for iter, test_data in enumerate(test_loader):
test_inputs, test_labels, u,v = test_data # not use landmarks while testing
test_inputs, test_labels, u,v = torch.autograd.Variable(test_inputs.cuda(), volatile=True).float(), torch.autograd.Variable(test_labels.cuda(),volatile=True).float(), torch.autograd.Variable(u.cuda(),volatile=True), torch.autograd.Variable(v.cuda(),volatile=True)
model_FC = model(test_inputs, u,v)
if iter % 100 ==0:
print(model_FC.size())
print(model_FC)
sio.savemat('FC.mat', {'FC':FC})
FC.append(model_FC.data.cpu().numpy())
sio.savemat('FC.mat', {'FC':FC})
print("Fully Connected Layers are saved as FC.mat ")
import keras
import numpy as np
import keras.backend as K
from keras.optimizers import Adam
from keras.callbacks import Callback
from model.seq2seq import Seq2Seq
from data.data_loader import data_loader
import utils.config as config
data = data_loader('./data/Sogo2008.dat', config)
chars = data.get_vocab('./data/vocab.json')
def gen_titles(s, topk=3):
xid = np.array([data.str2id(s)] * topk)
yid = np.array([[2]] * topk)
scores = [0] * topk
# print(xid, yid, scores)
for i in range(50):
proba = model.predict([xid, yid])[:, i, 3:]
# print(proba.shape)
log_proba = np.log(proba + 1e-6)
arg_topk = log_proba.argsort(axis=1)[:, -topk:]
_yid = []
_socres = []
if i == 0:
for j in range(topk):
# print(yid, arg_topk)
_yid.append(list(yid[j]) + [arg_topk[0][j] + 3])
# plan A, big change and no solution
from data.data_loader import data_loader
from preprocessing.up2down import up2down
from sklearn.preprocessing import scale
import numpy as np
prices = data_loader('../data/')
# ------- Config -------
u2d_error = 0.02
u2d_split_length = 10
# ----------------------
for price in prices:
adj_close = price.iloc[:, 5].values
adj_close = scale(adj_close)
u2d_adj_close_index = up2down(adj_close, u2d_error)
barrel = []
for index_ in range(len(u2d_adj_close_index) - u2d_split_length):
feature_point_index = u2d_adj_close_index[index_:index_ +
u2d_split_length]
feature_point_values = [0] * u2d_split_length
for i_ in range(u2d_split_length):
feature_point_values[i_] = adj_close[feature_point_index[i_]]
barrel.append((feature_point_index, feature_point_values))
barrel_length = len(barrel)
nn_distance = np.zeros([barrel_length, barrel_length], dtype=np.float)
for i_ in range(barrel_length - 1):
def run():
data = data_loader(DATA_SOURCE)
result = evaluator(data)
print(result)
#! /usr/bin/python # _*_ coding:utf-8 _*_ from data.data_loader import data_loader from sklearn.naive_bayes import MultinomialNB, BernoulliNB from model_op import Save_Model, result_metrics, Save_Weight from sklearn import svm from sklearn.svm import SVC from sklearn.ensemble import AdaBoostClassifier from sklearn.ensemble import RandomForestClassifier from sklearn.tree import DecisionTreeClassifier from sklearn import datasets, metrics filename = r'E:\zflPro\data\getfeature-resnet-1-dsn-flatten0.mat' # filename = r'getfeature-resnet-1-dsn-flatten0.mat' TrainSample, Trainlabel, TestSample, Testlabel = data_loader(filename) model_dir = r'E:\zflPro\MLmodel' model_TDlist = ["SVM", "RF", "ADB"] C_range = range(1, 20, 1) gama_max = 0.001 max_range = range(500, 1000, 10) nesti_ran = range(500) depth_range = range(30) rate_range = 10 njob_range = range(20) # svm.SVC def svmtrain(TrainSample, Trainlabel, Testlabel):