def load_json(file, new_root_dir=None, decompression=False):
"""Load a JSON file using json_tricks"""
if decompression:
with open(file, 'rb') as f:
my_object = load(f, decompression=decompression)
else:
with open(file, 'r') as f:
my_object = load(f, decompression=decompression)
if new_root_dir:
my_object.root_dir = new_root_dir
return my_object
def test_file_numpy():
path = join(mkdtemp(), 'pytest-np.json')
with open(path, 'wb+') as fh:
dump(deepcopy(npdata), fh, compression=9)
with open(path, 'rb') as fh:
data2 = load(fh, decompression=True)
_numpy_equality(data2)
def import_keras_json(filename): #import architecture and weights from keras network print 'Loading Model Architecture from Keras .json output...' with open(filename+"_arch.json") as datafile: arch_dict=load(datafile,preserve_order=True) return arch_dict
def test_file_numpy(): path = join(mkdtemp(), 'pytest-np.json') with open(path, 'wb+') as fh: dump(npdata, fh, compression=9) with open(path, 'rb') as fh: data2 = load(fh, decompression=True) _numpy_equality(data2)
def load_song(path, **kwargs):
'''
Small helper function to load song from path
:param path: (string) | path to load
:param kwargs: (arr) | to pass arguments to the Slice sub class
:return:
'''
y, sr = load(path=path, **kwargs)
return (y, sr)
def load_json_from_temp_folder(temp_out_folder, expected):
# load the results from the temporary folder
out_dict = {}
for exp in expected:
fpath = os.path.join(temp_out_folder, exp + '.json')
if os.path.isfile(fpath):
out_dict[exp] = json.load(open(fpath))
os.unlink(fpath) # remove file created in the temporary folder
else:
raise Exception(u'Missing output {0:s} file'.format(exp))
return out_dict
def main(argv):
inputfile = ''
outputfile = ''
tweets = []
labels = []
tknzr = TweetTokenizer()
try:
opts, args = getopt.getopt(argv, "hi:o:", ["ifile=", "ofile="])
except getopt.GetoptError:
print('vectorize.py -i <inputfile.json> -o <outputfile>')
sys.exit(2)
for opt, arg in opts:
if opt == '-h':
print('vectorize.py -i <inputfile.json> -o <outputfile>')
sys.exit()
elif opt in ("-i", "--ifile"):
inputfile = arg
elif opt in ("-o", "--ofile"):
outputfile = arg
data = load(open(inputfile, 'r'))
# pre process data all uniques and with accepted klasses
for pre_data in data:
if (pre_data['klass'] != 'NONE'):
tweets.append(tknzr.tokenize(pre_data['text']))
labels.append(pre_data['klass'])
# load http://crscardellino.me/SBWCE/ trained model
model = gensim.models.KeyedVectors.load_word2vec_format(
'SBW-vectors-300-min5.bin', binary=True)
shape = (len(tweets), MAX_NB_WORDS, 1)
tweets_tensor = np.zeros(shape, dtype=np.int32)
for i in range(len(tweets)):
#vectorizing each word in the tweet with a vector shape = (300,)
for f in range(len(tweets[i])):
word = tweets[i][f]
if f >= MAX_NB_WORDS:
continue
#if is not in the vocabulary
if word in model.wv.vocab:
tweets_tensor[i][f] = model.wv.index2word.index(word)
else:
#if it is a mention vectorize a name, for example @michael123 -> would be Carlos
if word[0] == '@':
tweets_tensor[i][f] = model.wv.index2word.index(name())
#if not append the unknown token
else:
tweets_tensor[i][f] = model.wv.index2word.index('unk')
#End of sentence token
if (f < MAX_NB_WORDS):
tweets_tensor[i][f] = model.wv.index2word.index('eos')
labels_array = np.array(list(
map(lambda label: label_to_value(label), labels)),
dtype=np.int32)
print(tweets_tensor)
print(labels_array)
img_x, img_y, img_z = 28,28,1
image_dim=(img_z,img_x,img_y)
X_train = X_train.reshape(X_train.shape[0], img_z, img_x, img_y)
X_test = X_test.reshape(X_test.shape[0], img_z, img_x, img_y)
X_train = X_train.astype('float32')
X_test = X_test.astype('float32')
X_train /= 255
X_test /= 255
samples_train = X_train.shape[0]
samples_test = X_test.shape[0]
#import architecture and weights from keras network
print 'Loading Model Architecture from Keras .json output...'
in_filename='mnist_CNN_v2'
with open(in_filename+"_arch.json") as datafile:
arch_dict=load(datafile,preserve_order=True)
print 'Building the Network...'
model_dict={}
conn_dict={}
presentation_time=0.05
# image=X_train[0]
# model_dict['image']=image
model = nengo.Network()
with model:
#rebuild the keras model
for key, value in arch_dict.items():
if key == 'input':
# This code is mostly by Sze Meng Tan, re-implementing the algorithms
# described in his 1986 PhD thesis "Aperture-synthesis mapping and
# parameter estimation"
import numpy
from scipy.optimize import leastsq, brent
import numpy as np
from json_tricks.np import load
from crocodile.synthesis import *
# Load kernel cache
KERNEL_CACHE = {}
with open('gridder.json', 'r') as f:
for key, val in load(f):
KERNEL_CACHE[tuple(key[0])] = val
def trap(vec, dx):
# Perform trapezoidal integration
return dx * (numpy.sum(vec) - 0.5 * (vec[0] + vec[-1]))
def func_to_min(h, x0, M, R):
N = len(h)
nu = (np.arange(M, dtype=float) + 0.5) / (2 * M)
x = x0 * np.arange(N+1, dtype=float)/N
C = calc_gridder_as_C(h, x0, nu, R)
dnu = nu[1] - nu[0]
dx = x[1] - x[0]
h_ext = np.concatenate(([1.0], h))
loss = np.zeros((len(h_ext), 2, M), dtype=float)
def from_json(input_str):
try: # file given
return json.load(open(input_str), preserve_order=False)
except IOError: # string given
return json.loads(input_str, 'r', preserve_order=False)
def load_music_data(attrstr):
attrfile = IO.get_abspath_from_relpath_in_tomato(
'music_data', attrstr + '.json')
return json.load(open(attrfile))
