def run_algorithms(list, algorithm):
# Creating a sort object
sort = Algorithms(list)
if algorithm == 'insertionsort' or algorithm == 'insertsort':
return sort.InsertSort()
elif algorithm == 'selectionsort' or algorithm == 'selectsort':
return sort.SelectSort()
elif algorithm == 'countingsort' or algorithm == 'countsort':
return sort.CountSort()
elif algorithm == 'shellsort':
return sort.ShellSort()
elif algorithm == 'quicksort':
return sort.QuickSort(list, 0, len(list))
elif algorithm == 'mergesort':
return sort.MergeSort(list)
elif algorithm == 'heapsort':
return sort.HeapSort(list)
elif algorithm == 'radixsort':
return sort.RadixSort()
else:
tools.error(algorithm)
sys.exit(2)
def define_optimizer(self, opts):
# Get graph:
graph = tf.get_default_graph()
# Define optimizer:
if opts.optimizer == 'adam':
optimizer = tf.train.AdamOptimizer(opts.initial_learning_rate)
elif opts.optimizer == 'sgd':
optimizer = tf.train.GradientDescentOptimizer(
opts.initial_learning_rate)
elif opts.optimizer == 'momentum':
optimizer = tf.train.MomentumOptimizer(opts.initial_learning_rate,
opts.momentum)
elif opts.optimizer == 'rmsprop':
optimizer = tf.train.RMSPropOptimizer(opts.initial_learning_rate,
momentum=opts.momentum)
else:
tools.error('Optimizer not recognized.')
# Operation to compute the gradients:
L_comb = graph.get_tensor_by_name('L_comb:0')
gradients = optimizer.compute_gradients(L_comb)
# Operation to apply the gradietns:
optimizer.apply_gradients(gradients, name='apply_grads_adam')
return gradients
def create_protocol(request):
import hashlib
if request.method == 'POST':
protocol_form = CreateProtocolForm(request.POST)
if protocol_form.is_valid():
try:
cd = protocol_form.cleaned_data
if ProtocolList.objects.filter(
name=cd['name'], user_id=request.user.id).exists():
return error('Duplicate record!')
protocol = ProtocolList(name=cd['name'],
user_id=request.user.id)
protocol.save()
softwares = request.POST.getlist('software', '')
parameters = request.POST.getlist('parameter', '')
steps = []
for index, software in enumerate(softwares):
if parameters[index]:
m = hashlib.md5()
m.update(software + ' ' + parameters[index].strip())
steps.append(
Protocol(software=software,
parameter=parameters[index],
parent=protocol.id,
hash=m.hexdigest(),
user_id=request.user.id))
Protocol.objects.bulk_create(steps)
return success('Your protocol have been created!')
except Exception, e:
return error(e)
else:
return error(str(protocol_form.errors))
def status(ctx):
"""Display some informations and statistics"""
title("Current status")
section("Settings")
click.echo("GEOZONES_HOME: {0}".format(ctx.obj["home"]))
section("Levels")
for level in ctx.obj["levels"]:
click.echo("{id}: {label}".format(**level.__dict__))
section("downloads")
urls = (level.urls for level in ctx.obj["levels"] if level.urls)
urls = set([url for lst in urls for url in lst])
for url in urls:
filename = basename(url).strip()
click.echo("{0} ... ".format(filename), nl=False)
if os.path.exists(os.path.join(DL_DIR, filename)):
success("present")
else:
error("absent")
section("coverage")
zones = DB()
total = 0
properties = ("population", "area", "wikipedia")
totals = dict((prop, 0) for prop in properties)
def countprop(name):
results = zones.aggregate(
[
{"$match": {name: {"$exists": True}, "level": {"$in": [l.id for l in ctx.obj["levels"]]}}},
{"$group": {"_id": "$level", "value": {"$sum": 1}}},
]
)
return dict((r["_id"], r["value"]) for r in results)
def display_prop(name, count, total):
click.echo("\t{0}: ".format(name), nl=False)
if count == 0:
func = error
elif count == total:
func = success
else:
func = warning
func("{0}/{1}".format(count, total))
counts = dict((p, countprop(p)) for p in properties)
for level in ctx.obj["levels"]:
count = zones.count({"level": level.id})
total += count
click.echo("{0}: {1}".format(level.id, count))
for prop in properties:
prop_count = counts[prop].get(level.id, 0)
totals[prop] += prop_count
display_prop(prop, prop_count, count)
click.secho("TOTAL: {0}".format(total), bold=True)
for prop in properties:
prop_total = totals[prop]
display_prop(prop, prop_total, total)
def process_dataset(self, workdir, db, url, extractor):
'''Extract territories from a given file for a given level with a given extractor function'''
loaded = 0
filename = join(workdir, basename(url))
with fiona.open('/', vfs='zip://{0}'.format(filename), encoding='utf8') as collection:
info('Extracting {0} elements from {1} ({2} {3})'.format(
len(collection), basename(filename), collection.driver, to_string(collection.crs)
))
for polygon in collection:
try:
zone = extractor(polygon)
if not zone:
continue
zone['keys'] = dict((k, v) for k, v in zone.get('keys', {}).items() if v is not None)
geom = shape(polygon['geometry'])
if extractor.simplify:
geom = geom.simplify(extractor.simplify)
if geom.geom_type == 'Polygon':
geom = MultiPolygon([geom])
elif geom.geom_type != 'MultiPolygon':
warning('Unsupported geometry type "{0}" for "{1}"'.format(geom.geom_type, zone['name']))
continue
zoneid = '/'.join((self.id, zone['code']))
zone.update(_id=zoneid, level=self.id, geom=geom.__geo_interface__)
db.find_one_and_replace({'_id': zoneid}, zone, upsert=True)
loaded += 1
except Exception as e:
error('Error extracting polygon {0}: {1}', polygon['properties'], str(e))
info('Loaded {0} zones for level {1} from file {2}'.format(loaded, self.id, filename))
return loaded
def add_job(request):
if request.method == 'POST':
job_form = SingleJobForm(request.POST)
if job_form.is_valid():
cd = job_form.cleaned_data
try:
if cd['parameter'].find(';') == -1:
cd['parameter'] += ';'
protocol = ProtocolList.objects.get(id=cd['protocol'])
if protocol.check_owner(
request.user.id) or request.user.is_superuser:
job = Queue(
protocol_id=cd['protocol'],
parameter=cd['parameter'],
run_dir=get_config('env', 'workspace'),
user_id=request.user.id,
input_file=cd['input_files'],
)
if check_disk_quota_lock(request.user.id):
job.save()
return success('Successfully added job into queue.')
else:
return error(
'You have exceed the disk quota limit! Please delete some files!'
)
else:
return error('You are not owner of the protocol.')
except Exception, e:
return error(e)
return error(str(job_form.errors))
def manage_reference(request):
if request.method == 'POST':
reference_form = CreateReferenceForm(request.POST)
if reference_form.is_valid():
cd = reference_form.cleaned_data
if References.objects.filter(user_id=request.user.id,
name=cd['name']).exists():
return error('Duplicate record!')
ref = References(
name=cd['name'],
path=cd['path'],
user_id=request.user.id,
)
ref.save()
return success(ref.id)
else:
return error(str(reference_form.errors))
else:
if request.user.is_superuser:
reference_list = References.objects.all()
else:
reference_list = References.objects.filter(
user_id=request.user.id).all()
return render(request, 'ui/manage_reference.html',
{'references': reference_list})
def joint_path(self, p1_flat, p2_flat):
# Parameters
if self.load_torch:
if self.dirmodel == None:
tools.error('Missing dirmodel.')
dirbase = self.dirmodel + 'tramo3/'
[pj_dense_weight,
pj_dense_bias] = read_weights_dense(dirbase + 'l1_dense.txt',
varname='pj_dense')
[pj_bn_mean, pj_bn_variance, pj_bn_offset, pj_bn_scale] = \
read_weights_bn(dirbase+'l2_bn.txt', varname='pj_bn')
else:
if self.xavier_init:
pj_dense_weight = weight_variable_xavier(
[768, 256], 768, 256, name='pj_dense_weight')
pj_dense_bias = bias_variable([256], name='pj_dense_bias')
# pj_bn_mean = bias_variable([256], name='pj_bn_mean')
# pj_bn_variance = weight_variable_xavier([256], 256, 256, name='pj_bn_variance')
# pj_bn_offset = bias_variable([256], name='pj_bn_offset')
# pj_bn_scale = weight_variable_xavier([256], 256, 256, name='pj_bn_scale')
else:
pj_dense_weight = weight_variable([768, 256],
name='pj_dense_weight')
pj_dense_bias = bias_variable([256], name='pj_dense_bias')
# pj_bn_mean = weight_variable([256], name='pj_bn_mean')
# pj_bn_variance = weight_variable([256], name='pj_bn_variance')
# pj_bn_offset = weight_variable([256], name='pj_bn_offset')
# pj_bn_scale = weight_variable([256], name='pj_bn_scale')
pj_bn_mean, \
pj_bn_variance, \
pj_bn_offset, \
pj_bn_scale = self.bn_variables(256, basename='pj_bn_scale')
# Probability of keeping units in the dropout layer:
keep_prob = tf.placeholder(tf.float32, name='keep_prob')
# Model
pj_join = tf.concat([p1_flat, p2_flat], 1, name='pj_join')
pj_dense = tf.add(tf.matmul(pj_join, pj_dense_weight),
pj_dense_bias,
name='pj_dense')
pj_bn = tf.nn.batch_normalization(pj_dense,
pj_bn_mean,
pj_bn_variance,
pj_bn_offset,
pj_bn_scale,
BN_EPS,
name='pj_bn')
pj_relu1 = tf.nn.relu(pj_bn, name='pj_relu1')
pj_dropout = tf.nn.dropout(pj_relu1, keep_prob, name='pj_dropout')
return pj_dropout
def process_dataset(self, workdir, db, url, extractor):
'''Extract territories from a given file for a given level with a given extractor function'''
loaded = 0
filename = join(workdir, basename(url))
# Identify the shapefile to avoid multiple file error on GDAL 2
with ZipFile(filename) as z:
candidates = [n for n in z.namelist() if n.endswith('.shp')]
if len(candidates) != 1:
raise ValueError(
'Unable to find a unique shpaefile into {0}'.format(
filename))
shp = candidates[0]
with fiona.open('/{0}'.format(shp),
vfs='zip://{0}'.format(filename),
encoding='utf8') as collection:
info('Extracting {0} elements from {1} ({2} {3})'.format(
len(collection), basename(filename), collection.driver,
to_string(collection.crs)))
for polygon in collection:
try:
zone = extractor(polygon)
if not zone:
continue
zone['keys'] = dict(
(k, v) for k, v in zone.get('keys', {}).items()
if v is not None)
geom = shape(polygon['geometry'])
if extractor.simplify:
geom = geom.simplify(extractor.simplify)
if geom.geom_type == 'Polygon':
geom = MultiPolygon([geom])
elif geom.geom_type != 'MultiPolygon':
warning(
'Unsupported geometry type "{0}" for "{1}"'.format(
geom.geom_type, zone['name']))
continue
zoneid = '/'.join((self.id, zone['code']))
zone.update(_id=zoneid,
level=self.id,
geom=geom.__geo_interface__)
db.find_one_and_replace({'_id': zoneid}, zone, upsert=True)
loaded += 1
except Exception as e:
error('Error extracting polygon {0}: {1}',
polygon['properties'], str(e))
info('Loaded {0} zones for level {1} from file {2}'.format(
loaded, self.id, filename))
return loaded
def body_path(self):
# Full image path:
x_b = tf.placeholder(tf.float32, shape=[None, 128, 128, 3], name='x_b')
# Check we are not applying any correction if loading the Torch model:
if self.load_torch and (self.correct_block2 or self.correct_avgpool):
tools.error(
'Not posible to load Torch model and apply any correction.')
# Parameters
if self.load_torch:
if self.dirmodel == None:
tools.error('Missing dirmodel.')
var_dict = self.load_weights_body()
else:
var_dict = self.random_weights_body()
# Model
block_configurations = [[2, 1, 1, 0], [2, 1, 1, 0], [2, 1, 1, 0]]
# Loop adding all blocks:
ini = 0
x = x_b
for block_idx in range(3):
var_dict_block = tools.get_subdictionary(ini, 8, var_dict)
strideW = block_configurations[block_idx][0]
strideH = block_configurations[block_idx][1]
padW = block_configurations[block_idx][2]
padH = block_configurations[block_idx][3]
if (not self.correct_block2) and block_idx == 1:
x = add_block_v2(x, var_dict_block, strideW, strideH, padW,
padH, 2, block_idx + 1)
else:
x = add_block(x, var_dict_block, strideW, strideH, padW, padH,
2, block_idx + 1)
ini = ini + 8
# average pooling
if self.correct_avgpool:
p2_avgpool = tf.layers.average_pooling2d(x,
pool_size=16,
strides=1,
padding="VALID",
name='p2_avgpool')
else:
p2_avgpool = tf.layers.average_pooling2d(x,
pool_size=3,
strides=16,
padding="VALID",
name='p2_avgpool')
# flattening
body_path_end = tf.reshape(p2_avgpool, [-1, 128], name='body_path_end')
return body_path_end
def delete_reference(request):
if request.method == 'GET':
if 'ref' in request.GET:
ref = References.objects.get(id=request.GET['ref'])
if ref.check_owner(request.user.id) or request.user.is_superuser:
ref.delete()
return success('Your reference has been deleted.')
else:
return error('You are not owner of the reference.')
else:
return error('Missing parameter.')
else:
return error('Error Method.')
def evaluate(xvals, coeff, data, variance, spec):
"""
Evaluate polynomial at xvals, repairing values where variance==0
with data/spec.
Parameters:
-----------
xvals: 1D float ndarray
X values for data.
coeff: 1D float ndarray
The polynomial coefficients (highest to lowest power).
data: 1D float ndarray
The data vector to fit.
variance: 1D float ndarray
The variance vector.
spec: 1D float ndarray
The spectrum vector.
deg: Integer
Degree of the polynomial.
Returns:
--------
fiteval: 1D float ndarray
The estimated spectrum
Example:
--------
See example in polyfunc.fit()
"""
# Check inputs:
nx = np.size(xvals)
nd = np.size(data)
nv = np.size(variance)
ns = np.size(spec)
if nx != nd or nx != nv or nx != ns:
tools.error("The length of data ({:d}), variance ({:d}), spec ({:d}), "
"and/or xvals ({:d}) are incompatible.".format(nd, nv, ns, nx))
# Find values with zero variance:
zerov = (variance == 0.0)
# Evaluate coefficients:
fiteval = np.polyval(coeff, xvals)
# Use actual data where varv is 0:
fiteval[zerov] = data[zerov] / spec[zerov]
return fiteval
def delete_step(request):
if request.method == 'GET':
if 'id' in request.GET:
try:
step = Protocol.objects.get(id=int(request.GET['id']))
if step.check_owner(
request.user.id) or request.user.is_superuser:
step.delete()
return success('Your step has been deleted.')
else:
return error('You are not owner of the step.')
except Exception, e:
return error(e)
else:
return error('Unknown parameter.')
def terminate_job(request):
if request.method == 'POST':
terminate_form = JobManipulateForm(request.POST)
if terminate_form.is_valid() or request.user.is_superuser:
cd = terminate_form.cleaned_data
try:
job = Queue.objects.get(id=cd['job'])
if job.check_owner(request.user.id):
job.terminate_job()
return success('Your job will be terminated soon.')
else:
return error('Your are not the owner of the job.')
except Exception, e:
return error(e)
else:
return error(str(terminate_form.errors))
def import_learning(request):
if request.session['learning']:
if request.session['learning']['a'] != 'no records':
learn = Prediction(
step_hash=request.session['learning']['hash'],
type=request.session['learning']['type'],
a=request.session['learning']['a'],
b=request.session['learning']['b'],
r=request.session['learning']['r'],
)
learn.save()
return success('Imported.')
else:
return error('Can not import records!')
else:
return error('Error')
def on_all_m_click(self, do=0):
# 0 = get current memory or 1 = plus memory or 2 = minus memory or 3 = clear memory
if do == 3:
self.memory = 0.0
elif not self.exp.is_error():
self.update_expression()
if do == 0:
self.history.out()
r = str(self.memory)
if r.endswith('.0'):
r = r.replace('.0', '')
self.exp.put_data_on_exp(r)
self.update_monitor()
elif do in (1, 2):
res = get_result(self.exp.exp, self.ui.radio_bt_1.isChecked())
if error(res):
self.history.out()
self.last_invalid_exp = self.exp.exp
self.exp.set_exp(res)
self.update_monitor()
else:
try:
exp_float = float(res)
self.memory += exp_float if do == 1 else -exp_float
except:
self.history.out()
self.last_invalid_exp = self.exp.exp
self.exp.set_exp(ERRORS[2])
self.update_monitor()
def fetch_learning(request):
import urllib2, json
query_string = request.GET['hash'] + ',' + request.GET['type'] + ',' + str(
get_config('env', 'cpu')) + ',' + str(get_config(
'env', 'memory')) + ',' + str(os_to_int())
api_bus = get_config('ml', 'api') + '/Index/share/q/' + query_string
try:
req = urllib2.Request(api_bus)
res_data = urllib2.urlopen(req)
res = json.loads(res_data.read())
session_dict = {
'hash': request.GET['hash'],
'type': request.GET['type'],
'a': res['a'],
'b': res['b'],
'r': res['r'],
}
request.session['learning'] = session_dict
template = loader.get_template('ui/fetch_learning.html')
context = RequestContext(request, {
'step': res,
})
return success(template.render(context))
except Exception, e:
return error(api_bus)
def save_score(self):
"""Saves the current score"""
nick = nick_selection.get_nick()
with open(self.SCORES_FILE, 'a') as fh:
fh.write(nick+"\t"+str(self.score))
# Send to remote server
params = self.REMOTE_SCORES_PARAMS
params["do"] = "add"
params["nick"] = nick
params["score"] = self.score
try:
r = requests.get(self.REMOTE_SCORES_URL, params=params)
if r.status_code != 200 or 'ack' not in r.text:
tools.error("Unable to post high score.")
except requests.exceptions.ConnectionError as e:
tools.error("Network error when submitting high score:"+str(e))
def delete_protocol(request):
if request.method == 'GET':
if 'id' in request.GET:
protocol_parent = ProtocolList.objects.get(
id=int(request.GET['id']))
if protocol_parent.check_owner(
request.user.id) or request.user.is_superuser:
protocol_parent.delete()
steps = Protocol.objects.filter(parent=int(request.GET['id']))
steps.delete()
return success('Your protocol has been deleted.')
else:
return error('You are not owner of the protocol.')
else:
return error('Unknown parameter.')
else:
return error('Method error.')
def get_learning_result(request):
if request.method == 'GET':
learning_form = QueryLearningForm(request.GET)
if learning_form.is_valid():
cd = learning_form.cleaned_data
try:
train = Prediction.objects.get(step_hash=cd['stephash'],
type=cd['type'])
template = loader.get_template('ui/get_learning_result.html')
context = RequestContext(request, {
'hit': train,
})
return success(template.render(context))
except Exception, e:
return error(e)
else:
return error(str(learning_form.errors))
def __init__(self, name, dict, ast_params):
self.ast_params = ast_params
self.attributes = []
self.clone_by_ref = False
self.cloner_prologue = ""
# Is the class concrete? (Default is false.)
self.concrete = False
self.declAttribute = ""
self.derived = []
self.desc = ""
self.hide = False
self.includes = {}
self.inline = {}
self.name = name
self.super = ""
self.super_non_nodes = []
for key in dict:
# Catch duplicate keys.
if isinstance(key, tuple):
(realkey, value) = key
error('duplicate key: ' + name + "::" + realkey)
if key not in [
'attributes',
'concrete',
'declAttribute',
'default',
'desc',
'hide',
'inline',
'printer',
'super',
'cloner_prologue',
'clone_by_ref',
]:
warning('unknown Node attribute: ' + name + "::" + key)
self.__dict__[key] = dict[key]
# If we have only one super-class, it has been parsed as a single
# value, but we want a list.
if (not isinstance(self.super, list)):
self.super = [self.super]
self.attributes = map(self.attribute_of_dict, self.attributes)
def __init__(self, name, dict, ast_params):
self.ast_params = ast_params
self.attributes = []
self.clone_by_ref = False
self.cloner_prologue = ""
# Is the class concrete? (Default is false.)
self.concrete = False
self.declAttribute = ""
self.derived = []
self.desc = ""
self.hide = False
self.includes = {}
self.inline = {}
self.name = name
self.super = ""
self.super_non_nodes = []
for key in dict:
# Catch duplicate keys.
if isinstance(key, tuple):
(realkey, value) = key
error("duplicate key: " + name + "::" + realkey)
if key not in [
"attributes",
"concrete",
"declAttribute",
"default",
"desc",
"hide",
"inline",
"printer",
"super",
"cloner_prologue",
"clone_by_ref",
]:
warning("unknown Node attribute: " + name + "::" + key)
self.__dict__[key] = dict[key]
# If we have only one super-class, it has been parsed as a single
# value, but we want a list.
if not isinstance(self.super, list):
self.super = [self.super]
self.attributes = map(self.attribute_of_dict, self.attributes)
def update_parameter(request):
if request.method == 'GET':
from urllib import unquote
update_parameter_form = StepManipulateForm(request.GET)
if update_parameter_form.is_valid():
cd = update_parameter_form.cleaned_data
step = Protocol.objects.get(id=cd['id'])
if (step.check_owner(request.user.id)
or request.user.is_superuser) and step.check_parent(
cd['parent']):
step.update_parameter(unquote(cd['parameter']))
step.save()
return success('Your step has been updated.')
else:
return error('Your are not owner of the step.')
else:
return error(str(update_parameter_form.errors))
else:
return error('Method error')
def load_batch_with_labels(self, opts):
# Initialize arrays:
im_prep_batch = np.zeros([opts.batch_size, 128, 128, 3],
dtype=np.float32)
true_labels = np.zeros((opts.batch_size, NDIM_DISC), dtype=np.float32)
# Fill the batches:
for idx_in_batch in range(opts.batch_size):
# Corresponding image index:
im_idx = self.curr_batch * opts.batch_size + idx_in_batch
if im_idx >= self.n_images_per_epoch:
tools.error('Image index over number of images per epoch')
# Load images (full and body):
im_full = tools.load_images_onepath(
self.annotations[self.indexes[im_idx]], opts)
# Add one dimension (for batch)
im_prep_batch[idx_in_batch, :, :, :] = im_full
# Build the batch with the true labels:
# Discrete:
for cat_idx in range(NCAT_IMAGENET):
if tools.category_in_annotation(
self.annotations[self.indexes[im_idx]], cat_idx):
true_labels[idx_in_batch, cat_idx] = 1
# Update batch index:
self.curr_batch = self.curr_batch + 1
# If we have completed a whole epoch, prepare a new one and restart the batch index:
if self.curr_batch == self.n_batches_per_epoch:
self.curr_batch = 0
self.prepare_epoch(opts)
# inputs = [im_full_prep_batch, im_body_prep_batch]
# labels = [true_labels_cont, true_labels_disc]
inputs = {'im_prep_batch': im_prep_batch}
labels = {'true_labels': true_labels}
return inputs, labels
def show_step(request):
if request.method == 'POST':
query_protocol_form = ProtocolManipulateForm(request.POST)
if query_protocol_form.is_valid():
cd = query_protocol_form.cleaned_data
if request.user.is_superuser:
step_list = Protocol.objects.filter(parent=cd['parent']).all()
else:
step_list = Protocol.objects.filter(
parent=cd['parent']).filter(user_id=request.user.id).all()
template = loader.get_template('ui/show_steps.html')
context = RequestContext(request, {
'step_list': step_list,
})
return success(template.render(context))
else:
return error(str(query_protocol_form.errors))
else:
return error('Method error.')
def show_learning_steps(request):
if request.method == 'GET':
if 'parent' in request.GET:
if request.user.is_superuser:
step_list = Protocol.objects.filter(
parent=int(request.GET['parent'])).all()
else:
step_list = Protocol.objects.filter(
parent=int(request.GET['parent'])).filter(
user_id=request.user.id).all()
template = loader.get_template('ui/show_learning_steps.html')
context = RequestContext(request, {
'step_list': step_list,
})
return success(template.render(context))
else:
return error('Wrong parameter.')
else:
return error('Method error.')
def show_job_log(request):
if request.method == 'POST':
query_job_form = JobManipulateForm(request.POST)
if query_job_form.is_valid():
cd = query_job_form.cleaned_data
log_path = os.path.join(get_config('env', 'log'), str(cd['job']))
try:
log_file = open(log_path, mode='r')
log = log_file.readlines()
log.reverse()
log = log[:100]
# log_content = [line+'<br />' for line in log]
log_content = '<br />'.join(log)
log_file.close()
return success(log_content)
except Exception, e:
return error(e)
else:
return error(str(query_job_form.errors))
def call(self, x):
prev_prediction = prev_reconstruction = x
prev_code = self.code
errors = []
for n in range(self.n_loops):
prediction = self.fn(self.decode(prev_code))
code = self.fn(self.encode(prev_reconstruction))
reconstruction = self.fn(self.decode(code))
errors.extend([
error(prev_prediction, prediction),
error(x, prediction) * 420.,
error(prev_reconstruction, reconstruction),
error(x, reconstruction) * 420.,
error(prev_code, code)
])
prev_reconstruction = reconstruction
prev_prediction = prediction
prev_code = code
y = self.out(errors)
return y
def __init__(
self, serial_port, nb_leds, corrections={"r": 1.0, "g": 1.0, "b": 1.0}, brightness=1.0, nb_steps_fading=25
):
"""Params:
* ser is an open pySerial object
* nb_leds is the total number of LEDs
* corrections is used to correct the differences between colors
intensity
* brightness is used to reduce the global brightess
* nb_steps_fading is the number of steps to use by default for
fading
"""
self.corrections = {k: float(brightness) * corrections[k] for k in corrections}
self.serial_port = serial_port
self.ser = serial.Serial(port=self.serial_port, baudrate=115200)
if not self.ser.isOpen():
try:
self.ser.open()
except Exception, e:
tools.error("Error opening serial port: " + str(e))
def handle(self):
self.data = self.rfile.readline().strip()
print "Got from {}:".format(self.client_address[0])
print self.data
# ACK
self.wfile.write(len(self.data))
# Send to the LEDs
try:
self.data = json.loads(self.data)
try:
fading_duration = self.data["fading_duration"]
except:
fading_duration = 0
self.control.send_colors(self.data["colors"],
fading=self.data["fading"],
fading_duration=fading_duration)
except:
tools.error("Invalid packet")
def build_model(opts):
if opts.modelname == 'cnn_emotic_1':
cnn_builder = cnn_emotic_1.cnn_builder_class(opts.cnn_opts[opts.modelname], opts)
else:
tools.error('modelname not recognized.')
# Define network:
if opts.net_arch == 'orig':
cnn_builder.define_network()
elif opts.net_arch == 'fullpath':
cnn_builder.define_fullpath()
elif opts.net_arch == 'bodypath':
cnn_builder.define_bodypath()
else:
tools.error('Network architecture not recognized.')
# Define loss:
if opts.loss_type == 'orig' or opts.loss_type == 'onlycont' or opts.loss_type == 'onlydisc':
cnn_builder.define_loss_orig(opts)
elif opts.loss_type == 'simple1':
cnn_builder.define_loss_simple1()
elif opts.loss_type == 'fullpath':
cnn_builder.define_loss_fullpath()
elif opts.loss_type == 'bodypath':
cnn_builder.define_loss_bodypath()
else:
tools.error('Loss type not recognized.')
# Define optimizer:
gradients = cnn_builder.define_optimizer(opts)
return gradients
def on_equal_click(self):
self.update_expression()
if not self.exp.is_error():
self.history.out()
res = get_result(self.exp.exp, self.ui.radio_bt_1.isChecked())
if self.make_power_enabled:
self.on_power_click()
if error(res):
self.last_invalid_exp = self.exp.exp
else:
self.history.history.append(self.exp.exp)
self.exp.set_exp(res)
self.update_monitor()
def show_job_folder(request):
import time, base64
if request.method == 'POST':
query_job_form = JobManipulateForm(request.POST)
if query_job_form.is_valid():
cd = query_job_form.cleaned_data
try:
job = Queue.objects.get(id=cd['job'])
if job.check_owner(
request.user.id) or request.user.is_superuser:
result_folder = job.get_result()
user_path = os.path.join(get_config('env', 'workspace'),
str(request.user.id),
result_folder)
user_files = []
for root, dirs, files in os.walk(user_path):
for file_name in files:
file_full_path = os.path.join(root, file_name)
file_path = file_full_path.replace(user_path+'\\', '')\
.replace(user_path+'/', '').replace(user_path, '')
tmp = dict()
tmp['name'] = file_path
tmp['file_size'] = os.path.getsize(file_full_path)
tmp['file_create'] = time.ctime(
os.path.getctime(file_full_path))
tmp['trace'] = base64.b64encode(
os.path.join(result_folder, file_path))
user_files.append(tmp)
template = loader.get_template('ui/show_job_folder.html')
context = RequestContext(request, {
'user_files': user_files,
})
return success(template.render(context))
else:
return error('Your are not the owner of the job.')
except Exception, e:
return error(e)
else:
return error(str(query_job_form.errors))
def full_path(self):
# Full image path:
x_f = tf.placeholder(tf.float32, shape=[None, 224, 224, 3], name='x_f')
# Parameters
if self.load_torch:
if self.dirmodel == None:
tools.error('Missing dirmodel.')
var_dict = self.load_weights_full()
else:
var_dict = self.random_weights_full()
# Model
block_configurations = [[4, 1, 2, 0], [2, 1, 2, 0], [2, 1, 1, 0],
[1, 1, 1, 0], [2, 1, 1, 0], [1, 1, 1, 0],
[2, 1, 1, 0], [1, 1, 1, 0]]
# Loop adding all blocks:
ini = 0
x = x_f
for block_idx in range(8):
var_dict_block = tools.get_subdictionary(ini, 8, var_dict)
strideW = block_configurations[block_idx][0]
strideH = block_configurations[block_idx][1]
padW = block_configurations[block_idx][2]
padH = block_configurations[block_idx][3]
x = add_block(x, var_dict_block, strideW, strideH, padW, padH, 1,
block_idx + 1)
ini = ini + 8
# average pooling
p1_avgpool = tf.layers.average_pooling2d(x,
pool_size=4,
strides=1,
padding="VALID",
name='p1_avgpool')
# flattening
full_path_end = tf.reshape(p1_avgpool, [-1, 640], name='full_path_end')
return full_path_end
def add_step(request):
import hashlib
if request.method == 'POST':
step_form = CreateStepForm(request.POST)
if step_form.is_valid():
cd = step_form.cleaned_data
try:
protocol = ProtocolList.objects.get(id=cd['parent'])
if protocol.check_owner(
request.user.id) or request.user.is_superuser:
m = hashlib.md5()
m.update(cd['software'] + ' ' + cd['parameter'].strip())
step = Protocol(software=cd['software'],
parameter=cd['parameter'],
parent=cd['parent'],
user_id=request.user.id,
hash=m.hexdigest())
step.save()
return success('Your step have been created.')
else:
return error('You are not owner of the protocol.')
except Exception, e:
return error(str(e))
def download_upload_file(request, f):
import base64
file_path = os.path.join(get_config('env', 'workspace'),
str(request.user.id), 'uploads',
base64.b64decode(f.replace('f/', '')))
try:
response = FileResponse(open(file_path, 'rb'))
response['Content-Type'] = 'application/octet-stream'
response['Content-Disposition'] = 'attachment;filename="{0}"'.format(
os.path.basename(file_path))
response['Content-Length'] = os.path.getsize(file_path)
return response
except Exception, e:
return error(e)
def predictions_path(self, pj_dropout):
# Parameters
if self.load_torch:
if self.dirmodel == None:
tools.error('Missing dirmodel.')
dirbase = self.dirmodel + 'tramo3/'
[yc_weight, yc_bias] = read_weights_dense(dirbase + 'l6_dense.txt',
varname='yc')
[yd_weight, yd_bias] = read_weights_dense(dirbase + 'l7_dense.txt',
varname='yd')
else:
if self.xavier_init:
yc_weight = weight_variable_xavier([256, NDIM_CONT],
256,
NDIM_CONT,
name='yc_weight')
yc_bias = bias_variable([NDIM_CONT], name='yc_bias')
yd_weight = weight_variable_xavier([256, NDIM_DISC],
256,
NDIM_DISC,
name='yd_weight')
yd_bias = bias_variable([NDIM_DISC], name='yd_bias')
else:
yc_weight = weight_variable([256, NDIM_CONT], name='yc_weight')
yc_bias = bias_variable([NDIM_CONT], name='yc_bias')
yd_weight = weight_variable([256, NDIM_DISC], name='yd_weight')
yd_bias = bias_variable([NDIM_DISC], name='yd_bias')
# Model
yc = tf.add(tf.matmul(pj_dropout, yc_weight), yc_bias, name='yc')
yd = tf.add(tf.matmul(pj_dropout, yd_weight), yd_bias, name='yd')
# yc = tf.sigmoid(tf.add(tf.matmul(pj_dropout, yc_weight), yc_bias), name='yc')
# yd = tf.sigmoid(tf.add(tf.matmul(pj_dropout, yd_weight), yd_bias), name='yd')
return yc, yd
def send_instructions(self, data):
"""Sends instructions to the LEDs server to display the current
configuration
params:
data is a dict {fading: bool, colors: {}}, cf get_diff
"""
if len(data["colors"]) == 0:
return
sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
try:
sock.connect(self.server_address)
data = json.dumps(data) + "\n"
sock.sendall(data)
received = sock.recv(1024)
try:
if int(received) != len(data.strip()):
tools.error("Error while sending instructions for LEDs")
except:
tools.error("Error while sending instructions for LEDs")
finally:
sock.close()
def stdextr(data, x1, x2, variance=None, mask=None, interp=False):
"""
Standard box extraction of spectrum. Step 4 of Horne (1989).
Parameters:
-----------
data: 2D float ndarray
Sky-subtracted spectrum image of shape [nwavelength, nposition].
x1: Integer
Left X boundary of region to extract the spectrum.
x2: Integer
Right X boundary of region to extract the spectrum.
Note that: 0 <= x1 <= x2 <= nx
variance: 2D float ndarray
Variance image from processed image.
mask: 2D integer ndarray
Mask of the data image (1 = good pixel, 0 = bad pixel).
interp: Bool
If True, lineraly interpolate the data for bad pixels.
Returns:
--------
stdspec: 1D float ndarray
The extracted spectrum.
stdvar: 1D float ndarray
Variance of extracted spectrum.
Example:
--------
>>> import sys
>>> import astropy.io.fits as fits
>>> import matplotlib.pyplot as plt
>>> sys.path.append("./src/")
>>> import stdextr as se
>>> data = fits.getdata("./images/ex1.fits")
>>> spec, sstd = se.stdextr(data, 230, 270)
>>> plt.plot(spec)
"""
# Check inputs:
nwave, nx = np.shape(data)
if variance is None:
variance = np.ones((nwave, nx), np.double)
if mask is None:
mask = np.ones((nwave, nx), np.byte)
if x1 < 0 or x2 <= x1 or nx < x2:
tools.error("Invalid x1, x2 boundaries (={:d}, {:d}), the values must "
"satisfy:\n 0 <= x1 < x2 <= nx (={:d}).".format(x1, x2, nx))
if np.shape(variance) != (nwave, nx):
tools.error("Incompatible shapes between data image ({:d}, {:d}) and "
"variance image ({:d}, {:d}).".format(nwave, nx, *np.shape(variance)))
if np.shape(mask) != (nwave, nx):
tools.error("Incompatible shapes between data image ({:d}, {:d}) and "
"mask image ({:d}, {:d}).".format(nwave, nx, *np.shape(mask)))
# Interpolate over bad pixels:
if interp:
stdspec = np.zeros(nwave)
for i in np.arange(nwave):
bad = np.where(mask[i, x1:x2] == 0)
good = np.where(mask[i, x1:x2] == 1)
datav = np.copy(data[i, x1:x2])
if len(bad) != 0:
interpol = si.interp1d(datav[good], good[0], kind="linear")
datav[bad] = interpol(bad[0])
stdspec[i] = np.sum(datav)
return stdspec, np.zeros(nwave)
# Standard extraction:
stdspec = np.sum((data * mask)[:, x1:x2], axis=1)
stdvar = np.sum((variance * mask)[:, x1:x2], axis=1)
return stdspec, stdvar
def error(self, text):
error(text)
self.errors += 1
else:
print('Game over')
self.game_over_animation()
self.save_score()
return self.score
def close(self):
"""Handle the cleaning, if necessary"""
if hasattr(self.game_controller, 'close'):
self.game_controller.close()
def __exit__(self, type, value, traceback):
self.close()
if __name__ == "__main__":
if len(sys.argv) < 2:
tools.error("Usage: "+sys.argv[0]+" SERVER_IP")
sys.exit(1)
HOST = sys.argv[1]
PORT = 4242
try:
while True:
print("Let's go !")
with Game((HOST, PORT), ctrl.Controller) as game:
print("Score " + str(game.loop()))
except KeyboardInterrupt:
print("\nExit…")
sys.exit()
def status(ctx):
'''Display some informations and statistics'''
title('Current status')
section('Settings')
click.echo('GEOZONES_HOME: {0}'.format(ctx.obj['home']))
section('Levels')
for level in ctx.obj['levels']:
click.echo('{id}: {label}'.format(**level.__dict__))
section('downloads')
urls = (level.urls for level in ctx.obj['levels'] if level.urls)
urls = set([url for lst in urls for url in lst])
for url in urls:
filename = basename(url).strip()
click.echo('{0} ... '.format(filename), nl=False)
if os.path.exists(os.path.join(DL_DIR, filename)):
success('present')
else:
error('absent')
section('coverage')
zones = DB()
total = 0
properties = ('population', 'area', 'wikipedia')
totals = dict((prop, 0) for prop in properties)
def countprop(name):
results = zones.aggregate([
{'$match': {
name: {'$exists': True},
'level': {'$in': [l.id for l in ctx.obj['levels']]}
}},
{'$group': {'_id': '$level', 'value': {'$sum': 1}}}
])
return dict((r['_id'], r['value']) for r in results)
def display_prop(name, count, total):
click.echo('\t{0}: '.format(name), nl=False)
if count == 0:
func = error
elif count == total:
func = success
else:
func = warning
func('{0}/{1}'.format(count, total))
counts = dict((p, countprop(p)) for p in properties)
for level in ctx.obj['levels']:
count = zones.count({'level': level.id})
total += count
click.echo('{0}: {1}'.format(level.id, count))
for prop in properties:
prop_count = counts[prop].get(level.id, 0)
totals[prop] += prop_count
display_prop(prop, prop_count, count)
click.secho('TOTAL: {0}'.format(total), bold=True)
for prop in properties:
prop_total = totals[prop]
display_prop(prop, prop_total, total)
def fit(xvals, data, variance, spec, deg=2):
"""
Fit data/spectrum using the variances as weights, using a polynomial
function.
Parameters:
-----------
xvals: 1D float ndarray
X values for data.
data: 1D float ndarray
The data vector to fit.
variance: 1D float ndarray
The variance vector.
spec: 1D float ndarray
The spectrum vector.
deg: Integer
Degree of the polynomial.
Returns:
--------
est: 1D float ndarray
The estimated spectrum.
coeff: 1D float ndarray
The polynomial coefficients (from highest to lowest power).
Example:
--------
>>> nx = 100
>>> xvals = np.arange(nx, dtype=np.double)
>>> prof = ((-(xvals/nx*2 - 1)**2 + 1)*0.75)/nx*2
>>> sbump = 10
>>> spec = np.sin(xvals/nx * np.pi * 6) + sbump
>>> gain = 10
>>> basey = profv*specv
>>> data = basey + basey * np.random.normal(0.0, 1.0, nx)/gain
>>> variance = basey/gain
>>> badloc = np.random.randint(0, 50, 4)
>>> data[badloc] = np.random.uniform(0, 2*np.amax(data), 4)
>>> prof[badloc] = 0
>>> deg = 2
>>> est, coeffv = pf.fit(xvals, data, variance, spec, deg)
>>> plt.figure(1)
>>> plt.clf()
>>> plt.plot(data/spec, 'or')
>>> plt.plot(xvals, est, 'b')
>>> est = pf.evaluate(xvals, coeffv, data, variance, spec)
>>> plot(xvals, est, 'ob')
"""
# Check inputs:
nd = np.size(data)
nv = np.size(variance)
ns = np.size(spec)
nx = np.size(xvals)
if nx != nd or nx != nv or nx != ns:
tools.error("The length of data ({:d}), variance ({:d}), spec ({:d}), "
"and/or xvals ({:d}) are incompatible.".format(nd, nv, ns, nx))
if deg < 0 or nx <= deg:
tools.error("Invalid polynomial degree ({:d}), must be: 0 < deg < nx-1 "
"= {:d}.".format(deg, nx-1))
# Initial estimate:
est = data / spec
# Find where the variance has non-zero values:
nzerov = (variance != 0.0)
# Fit polynomial where variance != 0:
merrors = variance/spec**2
# Use all pixels for estimation
merrors = np.clip(merrors, 1e-8, np.amax(merrors))
coeff = np.polyfit(xvals, data/spec, deg, w=merrors)
est[nzerov] = np.polyval(coeff, xvals[nzerov])
return est, coeff
def run(env, start_response):
"""Main."""
if 'PATH_INFO' not in env:
env['PATH_INFO'] = '/'
path = filter(None, os.path.normpath(env['PATH_INFO']).split('/'))
if env['REQUEST_METHOD'] == 'GET':
if len(path) == 0:
path.append('_')
dataset, path = path[0], path[1:]
try:
dataset = int(dataset)
if not tools.exists(dataset):
raise ValueError
except ValueError:
ret = tools.error(404, "Invalid dataset ID.")
else:
if len(path) == 0:
path.append('json')
handler, command = path[0], path[1:]
if handler not in tools.get_handlers:
ret = tools.error(501, "Didn't understand (sorry).")
else:
ret = tools.get_handlers[handler](dataset, command)
elif env['REQUEST_METHOD'] == 'POST' or env['REQUEST_METHOD'] == 'PUT':
if len(path) == 0:
# Create new dataset
ret = tools.post_handlers['create'](env)
elif len(path) == 1 and path[0] == 'load':
# Load from RDF
ret = tools.post_handlers['load'](env)
elif len(path) == 1:
# Update existing dataset
try:
dataset = int(path[0])
if not tools.exists(dataset):
raise ValueError
except ValueError:
ret = tools.error(404, "Invalid dataset ID.")
else:
ret = tools.post_handlers['update'](dataset, env)
else:
# Handlers
dataset, handler, command = path[0], path[1], path[2:]
try:
dataset = int(path[0])
if not tools.exists(dataset):
raise ValueError
except ValueError:
ret = tools.error(404, "Invalid dataset ID.")
else:
if handler in tools.post_handlers:
ret = tools.post_handlers[handler](dataset, env, command)
else:
ret = tools.error(400, "Bad request.")
else:
# Unsupported action
ret = tools.error(501, "Didn't understand (sorry).")
status, headers, content = ret
start_response(status, headers)
return content