def setUp(self):
# Need this for Resource Auth
App.get_config()['SECRET_KEY'] = 'mysecrettestkey'
self.db = DB
# do once per test
load_data('%s/data/fixtures.json' % TEST_DIR)
def train(self):
train_data, test_data, n_labels = models.load_data(self.args)
n_epochs = self.args['training']['n_epochs']
batch_size = self.args['training']['batch_size']
d_steps = self.args['training']['d_steps']
log_interval = self.args['training']['log_interval']
self.writer = torch.utils.tensorboard.SummaryWriter(
log_dir=self.log_dir + '/run/')
self.fixed_z = self.z_dist.sample((64, ))
self.fixed_y = self.y_dist.sample((64, ))
self.z_start = self.z_dist.sample((32, ))
self.z_end = self.z_dist.sample((32, ))
self.z_inter_list = metrics.slerp(self.z_start, self.z_end, 14)
for eidx in trange(n_epochs, leave=True, desc='Epoch'):
for iidx, real_batch in enumerate(tqdm(train_data)):
noise_batch = self.z_dist.sample((batch_size, ))
loss_d = self._train_d(real_batch=real_batch,
noise_batch=noise_batch)
self.writer.add_scalar('Loss/%d/d' % eidx,
loss_d,
global_step=iidx)
if iidx % d_steps == 0:
# train G
if len(real_batch) == 2:
label_batch = real_batch[1]
else:
label_batch = None
loss_g = self._train_g(noise_batch=noise_batch,
label_batch=label_batch)
self.writer.add_scalar('Loss/%d/g' % eidx,
loss_g,
global_step=iidx)
if eidx % log_interval == 0:
self._log(eidx, test_data)
def load_and_report():
# fetching the current time as name
name = names_of_fig()
# load Pricing test data
all_contracts, p_sorted = load_data()
# List contracts
print(f"Showing all_contracts..........\n{all_contracts.tail(10)}\n")
print(f"Showing p_sorted......\n{p_sorted.tail(10)}\n")
# output from potential_pair
ret, list_sect = potential_pairs(all_contracts, p_sorted)
print(f"showing list_sect.......\n{list_sect}\n")
print(f"showing ret........\n{ret.tail(10)}\n")
# show the results of in sample testing
ret.iloc[0] = 1
ret.index = all_contracts.index
plt.figure(figsize=(15, 7))
plt.xlabel('Trade Date')
plt.grid(True)
plt.plot(ret)
plt.legend(list(ret.columns))
plt.show()
plt.savefig(os.path.join("charts/sample test", "sample test " + name))
# calculate the performance
perf = ret.calc_stats()
perf.display()
perf.to_csv(sep=',', path="train_perfer.csv")
# plot the maxinum drawndown of each pair
ffn_ret = ffn.to_drawdown_series(ret)
plt.figure(figsize=(15, 7))
plt.grid(True)
plt.plot(ffn_ret)
plt.legend(list(ffn_ret.columns))
plt.show()
plt.savefig(os.path.join("charts/ffn drawdown", "ffn max drawdown " + name))
# In sample back testing of portfolio
port = ret.mean(axis=1)
plt.figure(figsize=(15, 7))
plt.grid(True)
plt.plot(port)
plt.show()
#plt.legend(list(port.columns))
plt.savefig(os.path.join("charts/testing of portfolio", "portfolio test " + name))
perf = port.calc_stats()
print(f"\n\nPrinting perf stats.......\n{perf.stats}\n")
# In sample back testing of portfolio maxinum drawndown
ffn_port = ffn.to_drawdown_series(port)
plt.figure(figsize=(15, 7))
plt.grid(True)
plt.plot(ffn_port)
#plt.legend(list(ffn_port.columns))
plt.savefig(os.path.join("charts/back testing of portfolio maxinum drawndown", "maxinum drawndown " + name))
####################
##sample back testing######
#####################
test_ret, testing_data = sample_backtest(list_sect)
print(f"\n\n\nShowing sample back testing- testing data tail.......\n\n{testing_data.tail(3)}\n")
test_ret.iloc[0] = 1
print(f"\n\nShowing sample backtesting test_ret tail.......\n\n{test_ret.tail(3)}")
print(f"\n\nshowing test_ret shape......\n\n{test_ret.shape})")
print(f"\n\n\nshowing test_ret index........\n\n{test_ret.index}")
# plotting test_ret sample back testing
plt.plot(test_ret)
plt.legend(list(test_ret.columns))
plt.savefig(os.path.join("charts/sample Backtesting/test_ret", "test_ret " + name))
# Out sample back testing of portfolio
port = test_ret.mean(axis=1)
plt.figure(figsize=(15, 7))
plt.grid(True)
plt.plot(port)
# plt.legend(list(port.columns))
plt.savefig(os.path.join("charts/sample Backtesting/portfolio", "backtesting portfolio " + name))
perf = port.calc_stats()
print(perf.stats)
ffn_backtest_sample = ffn.to_drawdown_series(port)
plt.figure(figsize=(15, 7))
plt.grid(True)
plt.plot(ffn_backtest_sample)
# plt.legend(list(ffn_backtest_sample.columns))
plt.savefig(os.path.join("charts/sample Backtesting/ffn_drawdown_port", "drwadown_port " + name))
return None
#parser.add_argument('patch_shape', type=int,
# help='The size of the input patch window.'),
#parser.add_argument('label_patch_shape', type=int,
# help='The size of the predicted patch window.'),
#parser.add_argument('num_channels', type=int,
# help='Number of channels in the dataset.'),
#args = parser.parse_args()
#path_testset = self.testset
path_testset = '/home/local/USHERBROOKE/havm2701/data/DBFrames'
batch_size = 100
dataset = '/home/local/USHERBROOKE/havm2701/git.repos/Deep_VAMP/theano/mnist.pkl.gz'
datasets = models.load_data(dataset)
pdb.set_trace()
train_set_x, train_set_y = datasets[0]
valid_set_x, valid_set_y = datasets[1]
test_set_x, test_set_y = datasets[2]
pdb.set_trace()
#train = VAMP(start=0,stop=10000,image_resize=[128,64],toronto_prepro=True)
#valid = VAMP(start=10000,stop=12000,image_resize=[128,64],toronto_prepro=True)
#valid = valid.get_reshaped_images()
#train = train.get_reshaped_images()
dataset ={}
#train.y=np.argmax(train.y,axis=1)
#train.y = np.asarray(train.y,dtype=np.32)
#valid.y=np.argmax(valid.y,axis=1)
import numpy as np
import PIL.Image as pil
import matplotlib.pyplot as plt
import random as rng
import cv2 as cv
from models import load_data, view_image
print(__doc__)
data_dir = 'data/'
train_data, train_labels, sub = 'train_images.pkl', 'train_labels.csv', 'test_images.pkl'
# Load the data into variables and normalize data
X, y, sub = load_data(data_dir, train_data, train_labels, sub)
rng.seed(12345)
# Test some opencv bs to see if it works
def thresh_callback(val):
threshold = val
canny_output = cv.Canny(src_gray, threshold, threshold*2)
_, contours, hierarchy = cv.findContours(canny_output, cv.RETR_EXTERNAL, cv.CHAIN_APPROX_SIMPLE)
# Bounding rectangles
contours_poly = [None]*len(contours)
boundRect = [None]*len(contours)
centers = [None]*len(contours)
radius = [None]*len(contours)
def get_wheel_sets():
return load_data()
def setUp(self):
self.db = DB
# do once per test
load_data('%s/data/minimal.json' % TEST_DIR)
'''
@author: viet
Prototyping the thinning pipeline, might not use it tho
'''
from models import load_data, view_image
from models.img_processing import threshold_background
import numpy as np
import cv2 as cv
from PIL import Image
train_data, train_labels, sub_data = load_data('data/', 'train_images.pkl',
'train_labels.csv',
'test_images.pkl')
train_labels = train_labels['Category'].values # Get labels
train_data, sub_data = (train_data)[:, :, :, None], (sub_data)[:, :, :, None]
train_data, sub_data = np.transpose(train_data, (0, 3, 1, 2)), np.transpose(
sub_data, (0, 3, 1, 2))
def convert_to_3_channels(img_array):
'Literally does what the name says it does'
new_array = []
for i in range(len(img_array)):
e = img_array[i][0]
new_image = [e, e, e] # 3 channels
new_array.append(new_image)
return np.asarray(new_array)