def Load():
# parse the xmls and (re)build the models and clips index
global ModelDemos, ModelDirs
ModelDirs = {}
ModelDemos = {}
if cfg.vdhd_data is None:
return
for dir_name in os.listdir(cfg.vdhd_data):
# inside the vghd data dir,
# each dir is named as the card no,
model_dir = pj(cfg.vdhd_data, dir_name)
if os.path.isdir(model_dir):
# for each such dir,
for file_item in os.listdir(model_dir):
# if there is an xml file,
if file_item.endswith(".xml") and "cache" not in file_item:
# store the dir.
ModelDirs[dir_name] = model_dir
ModelDemos[dir_name] = []
demo_dir = pj(cfg.vghd_models, dir_name)
# if a same dir exists inside the vghd models dir,
if os.path.isdir(demo_dir):
# store the clips.
for demo_file in os.listdir(demo_dir):
# we will need to change here if we want
# to include more than demo files
if demo_file.endswith(".demo"):
ModelDemos[dir_name].append(demo_file)
def clean(rep="."):
"""Thorough cleaning of all arborescence rooting at rep.
Args:
rep: (str) root directory to start the walk
Returns:
None
"""
for name in ("build", "dist"):
pth = pj(rep, name)
if exists(pth):
rmtree(pth)
for root, dnames, fnames in walk(rep):
# do not walk directories starting with "."
for name in tuple(dnames):
if "clean.no" in listdir(pj(root, name)):
dnames.remove(name)
elif name.startswith("."):
dnames.remove(name)
elif name == "__pycache__":
rmtree(pj(root, name))
dnames.remove(name)
for name in fnames:
if not name.startswith("."):
if splitext(name)[1] in [".pyc", ".pyo"]:
remove(pj(root, name))
def docker_build(dist=None, force=False):
''' build Docker image from ./docker/<dist>/Dockerfile '''
force = prep_bool_arg(force)
def build(dist, force):
image_existing = False
if not force:
with settings(warn_only=True):
image_existing = local('docker images | grep -q package_build/{dist}'.format(dist=dist)).succeeded
if not image_existing:
print '(re)building image {}...'.format(dist)
local('docker build --tag=package_build/{dist} {path}/docker/{dist}/'.format(dist=dist, path=PATH))
if dist:
print 'building Docker image for {}...'.format(dist)
builddir = pj(PATH, 'docker', dist)
os.path.isdir(builddir) or abort('{} dir is not existing'.format(builddir))
build(dist, force)
else:
print 'building Docker images for all distributions...'
for entry in os.listdir(pj(PATH, 'docker')):
if os.path.isdir(pj(PATH, 'docker', entry)):
build(entry, force)
def my_special_function(name):
""" This function will create 42 C files. """
path = pj(EXPERIMENT_FOLDER, name)
if not os.path.exists(path):
os.makedirs(path)
c_template = TEMPLATE_ENV.get_template("dummy_template.c")
# We make the id start at 1 and finish at 42
for value in range(1, 43):
with open(pj(path, "dummy_%d.c" % value), "w") as f:
f.write(c_template.render(my_value=value))
# If you change the platform target and want to push to iotlab
# don't forget to update the nodes names
makefile_template = TEMPLATE_ENV.get_template("dummy_makefile")
with open(pj(path, "Makefile"), "w") as f:
f.write(makefile_template.render(contiki=CONTIKI_FOLDER,
target="iotlab-m3"))
config_template = TEMPLATE_ENV.get_template("dummy_iotlab.json")
res = [
{"nodes": ["m3-%d.grenoble.iot-lab.info" % num],
"firmware_path": pj(path, "dummy_%d.iotlab-m3" % num)
} for num in range(1, 43)]
with open(pj(path, "iotlab.json"), "w") as f:
f.write(json.dumps(res, sort_keys=True,
indent=4, separators=(',', ': ')))
def package_build(dist=None, recipe='', upload=False):
''' build packages from recipe for dist '''
if dist:
dists = [dist]
else:
dists = [d for d, _ in PACKAGE_FORMAT.items()]
for dist in dists:
start_time = time.time()
execute('docker_run', dist, '/data/cook-recipe.sh {}'.format(recipe))
for root, dirs, files in os.walk(pj(PATH, 'recipes')):
for file in files:
if file == 'lastbuild' and os.path.getmtime(pj(root, file)) >= start_time:
package_name = ''
with open(pj(root, file), 'r') as fh:
package_name = fh.readline().strip()
print('package_name: {}'.format(package_name))
if package_name:
package_format = package_name.split('.')[-1]
dist = root.split('/')[-1]
if upload:
execute('repo_{0}_add'.format(package_format), pj(root, package_name), dist)
print 'task ran {0} seconds'.format(time.time() - start_time)
def setUp(self):
self.bin = pj(program_path, "DBNDownBeatTracker")
self.activations = Activations(
pj(ACTIVATIONS_PATH, "sample.downbeats_blstm.npz"))
self.result = np.loadtxt(
pj(DETECTIONS_PATH, "sample.dbn_downbeat_tracker.txt"))
self.downbeat_result = self.result[self.result[:, 1] == 1][:, 0]
def setUp(self):
self.bin = pj(program_path, "GMMPatternTracker")
self.activations = Activations(
pj(ACTIVATIONS_PATH, "sample.gmm_pattern_tracker.npz"))
self.result = np.loadtxt(
pj(DETECTIONS_PATH, "sample.gmm_pattern_tracker.txt"))
self.downbeat_result = self.result[self.result[:, 1] == 1][:, 0]
def default(self):
if CONDA_ENV:
qt_dir = CONDA_LIBRARY_PREFIX
self._default["QTDIR"] = qt_dir
self._default["QT5_BINPATH"] = pj(qt_dir, 'bin')
self._default["QT5_CPPPATH"] = pj(qt_dir, 'include')
self._default["QT5_LIBPATH"] = pj(qt_dir, 'lib')
self._default["QT5_FRAMEWORK"] = False
return
qt_dir = os.getenv("QTDIR")
qt_lib = '$QTDIR/lib'
qt_bin = '$QTDIR/bin'
qt_inc = '$QTDIR/include'
qt_fmk = False
if isinstance(platform, Linux):
# Use LSB spec
qt_dir = '/usr'
qt_bin = '/usr/bin'
qt_inc = '/usr/include/qt5'
qt_lib = '/usr/lib'
elif isinstance(platform, Darwin) and exists('/opt/local/libexec/qt5'):
qt_dir = '/opt/local/libexec/qt5'
qt_bin = '/opt/local/libexec/qt5/bin'
qt_inc = '/opt/local/libexec/qt5/include'
qt_lib = '/opt/local/libexec/qt5/lib'
elif not qt_dir:
try:
if isinstance(platform, Win32) or isinstance(platform, Darwin):
# Try to use openalea egg
from openalea.deploy import get_base_dir
qt_dir = get_base_dir("qt5-dev")
except:
if isinstance(platform, Win32):
try:
from openalea.deploy import get_base_dir
qt_dir = get_base_dir("qt5")
except:
# Try to locate bin/moc in PATH
qt_dir = find_executable_path_from_env("moc.exe", strip_bin=True)
elif isinstance(platform, Posix):
qt_dir = pj('/usr', 'lib', 'qt5')
if not exists(pj(qt_dir, 'bin')):
# Use LSB spec
qt_dir = '/usr'
qt_bin = '/usr/bin'
qt_inc = '/usr/include/qt5'
qt_lib = '/usr/lib'
if isinstance(platform, Darwin):
qt_fmk = True
self._default["QTDIR"] = qt_dir
self._default["QT5_BINPATH"] = qt_bin
self._default["QT5_CPPPATH"] = qt_inc
self._default["QT5_LIBPATH"] = qt_lib
self._default["QT5_FRAMEWORK"] = qt_fmk
def setUp(self):
self.bin = pj(program_path, "DBNBeatTracker")
self.activations = Activations(
pj(ACTIVATIONS_PATH, "sample.beats_blstm.npz"))
self.result = np.loadtxt(
pj(DETECTIONS_PATH, "sample.dbn_beat_tracker.txt"))
self.online_results = [0.47, 0.79, 1.48, 2.16, 2.5]
def default(self):
self._default['libs_suffix'] = '$compiler_libs_suffix'
self._default['flags'] = ''
self._default['defines'] = ''
isPosix = isinstance(platform, Posix)
# -- lets now look for decent flags --
self._default['flags'] = self.get_default_flags()
self._default['defines'] = self.get_default_defines()
# -- lets now look for decent include dirs --
try:
# Try to use openalea egg
from openalea.deploy import get_base_dir
try:
base_dir = get_base_dir("boost")
except:
base_dir = get_base_dir("boostpython")
self._default['include'] = pj(base_dir, 'include')
self._default['lib'] = pj(base_dir, 'lib')
self.__usingEgg = True
except:
try:
import openalea.config as conf
self._default['include'] = conf.include_dir
self._default['lib'] = conf.lib_dir
except ImportError, e:
self._default['include'] = '/usr/include' if isPosix else pj(os.getcwd(), "include")
self._default['lib'] = '/usr/lib' if isPosix else pj(os.getcwd(), "lib")
def install():
from openalea.deploy.shortcut import create_win_shortcut, set_win_reg, create_fd_shortcut
from openalea.deploy import get_base_dir
import sys
from os.path import join as pj
# Get the location of the installed egg
base_dir = get_base_dir('openalea.visualea')
share_dir = pj(base_dir, 'share')
winexe = sys.executable
winexe = winexe.replace('python.exe', 'pythonw.exe')
prefix = base_dir.lower().split("lib")[0]
create_win_shortcut(name = 'Visualea',
target = winexe,
arguments = '"'+pj(prefix, 'Scripts', 'visualea-script.pyw')+'"',
startin = "",
icon = pj(share_dir, 'openalea_icon.ico'),
description = "Visual programming",
menugroup = "OpenAlea")
create_win_shortcut(name = 'Python Shell',
target = winexe,
arguments = '"'+pj(prefix, 'Scripts', 'aleashell-script.pyw')+'"',
startin = "",
icon = "",
description = "Python Shell",
menugroup = "OpenAlea")
def main(top):
top = os.path.abspath(top)
if not exists(pj(top, "abipy")) or not exists(pj(top, "setup.py")):
raise ValueError("top %s is not the top-level abipy directory" % top)
os.chdir(top)
sys.path.insert(0, top)
import cProfile
import pstats
prof_file = pj(top, ".abipy_prof")
# Profile the import of the package
cProfile.run('abipy_import()', prof_file)
p = pstats.Stats(prof_file)
#If you were looking to see what functions were looping a lot, and taking a lot of time, you would do:
p.sort_stats('time').print_stats(10)
#This will sort all the statistics by file name, and then print out statistics for only the class init methods
#(since they are spelled with
#__init__ in them).
#p.sort_stats('file').print_stats('__init__')
return 0
def default(self):
if isinstance(platform, Win32):
self._default['flags'] = ''
self._default['defines'] = ''
try:
cgalroot = os.environ['CGALROOT']
self._default['include'] = pj(cgalroot,'auxiliary','gmp','include')
self._default['libpath'] = pj(cgalroot,'auxiliary','gmp','lib')
self._default['libs'] = 'libgmp-10'
except:
try:
import openalea.config as conf
self._default['include'] = conf.include_dir
self._default['libpath'] = conf.lib_dir
except ImportError, e:
try:
import pkg_resources as pkg
egg_env = pkg.Environment()
mingw_base = egg_env["mingw"][0].location
self._default['include'] = pj(mingw_base, "include")
self._default['libpath'] = pj(mingw_base, "lib")
except Exception, e:
self._default['include'] = 'C:' + os.sep
self._default['libpath'] = 'C:' + os.sep
self._default['libs'] = 'gmp'
def set_data_files(self, share_path):
for root, dirs, file_names in os.walk('data/'):
files = []
for file_name in file_names:
if file_name.endswith('.pyc'):
continue
elif file_name.endswith('.dat') and \
file_name.startswith('Geo'):
continue
elif file_name == "ooniprobe.conf.sample":
files.append(self.gen_config(share_path))
continue
files.append(pj(root, file_name))
self.distribution.data_files.append(
[
pj(share_path, 'ooni', root.replace('data/', '')),
files
]
)
settings = SafeConfigParser()
settings.add_section("directories")
settings.set("directories", "data_dir",
os.path.join(share_path, "ooni"))
with open("ooni/settings.ini", "w+") as fp:
settings.write(fp)
def find_login_passwd(allow_user_input=True):
home = ""
# Get password
if 'USERPROFILE' in os.environ:
home = os.environ['USERPROFILE']
elif 'HOME' in os.environ:
home = os.environ['HOME']
rc = pj(home, '.pypirc')
if not exists(rc):
matched = glob.glob(pj(home, "*pydistutils.cfg"))
if len(matched):
rc = matched[0]
username, password = None, None
if exists(rc):
print('Using PyPI login from %s' % (rc))
config = configparser.ConfigParser({
'username': '',
'password': '',
'repository': ''})
config.read(rc)
username = config.get('server-login', 'username')
password = config.get('server-login', 'password')
elif allow_user_input:
username = input("Enter your GForge login:"******"Enter you GForge password:")
return username, password
def regenerate_package(env, target=".", overwrite=False):
"""Rebuild all automatically generated files.
Args:
env (jinja2.Environment): current working environment
target (str): target directory to write into
overwrite (bool): default False, whether or not
to overwrite user modified files
Returns:
None
"""
# check consistency of env params
invalids = []
for option in installed_options(env):
for n in check_option_parameters(option, env):
invalids.append("%s.%s" % (option, n))
if len(invalids) > 0:
for param in invalids:
logger.warning("param %s is not valid", param)
return False
# check for potential conflicts
hm_ref = get_pkg_hash(target)
conflicted = []
for file_pth in hm_ref:
pth = pj(target, file_pth)
if exists(pth) and modified_file_hash(pth, hm_ref):
conflicted.append(pth)
else:
# file disappeared, regenerate_dir will reload it if managed by pkglts
pass
overwrite_file = {}
if len(conflicted) > 0:
if overwrite:
for name in conflicted:
logger.debug("conflicted, '%s'" % name)
overwrite_file[pth_as_key(name)] = True
else:
for name in conflicted:
print("A non editable section of %s has been modified" % name)
overwrite_file[pth_as_key(name)] = get_user_permission("overwrite", False)
# render files for all options
hm = {}
for name in installed_options(env):
opt_ref_dir = pj(get_data_dir(), 'base', name)
if not exists(opt_ref_dir):
logger.debug("option %s do not provide files" % name)
else:
logger.info("rendering option %s" % name)
loc_hm = regenerate_dir(opt_ref_dir, target, env, overwrite_file)
hm.update(loc_hm)
hm_ref.update(hm)
write_pkg_hash(hm_ref, target)
def clean(rep="."):
"""Thorough cleaning of all arborescence rooting at rep.
Todo: exception list instead of hardcoded one
Args:
rep (str): default ".", top directory to clean
Returns:
None
"""
for name in ("build", "dist", "doc/_dvlpt", "doc/build"):
pth = normpath(pj(rep, name))
if exists(pth):
rmtree(pth)
for root, dnames, fnames in walk(rep):
# do not walk directories starting with "."
for name in tuple(dnames):
if "clean.no" in listdir(pj(root, name)):
dnames.remove(name)
elif name.startswith("."):
dnames.remove(name)
elif name == "__pycache__":
rmtree(pj(root, name))
dnames.remove(name)
for name in fnames:
if not name.startswith("."):
if splitext(name)[1] in [".pyc", ".pyo"]:
remove(pj(root, name))
def protocol_repartition_depth(folder):
output_folder = pj(folder, "results", "protocol_repartition")
if not os.path.exists(output_folder):
os.makedirs(output_folder)
depth_df = pd.read_csv(pj(folder, "results", "depth.csv"))
depth_df.set_index("node", inplace=True)
pcap_df = pd.read_csv(pj(folder, "results", "pcap_relooked.csv"))
pcap_df = pcap_df.join(depth_df, on="mac_src")
res = pd.DataFrame()
res["rpl"] = pcap_df[pcap_df.icmpv6_type == "rpl"].groupby("depth").sum().length
res["udp"] = pcap_df[pcap_df.icmpv6_type == "udp"].groupby("depth").sum().length
RATE = 250000
res["rpl"] = 8.0 * res["rpl"] / RATE
res["udp"] = 8.0 * res["udp"] / RATE
ax = res.plot(kind="bar", stacked=True)
ax.set_ylabel('Time [s]')
ax.set_xlabel("Depth")
ax.set_xticklabels(res.index.map(int), rotation=0)
fig = plt.gcf()
plt.tight_layout()
# fig.set_size_inches(18.5, 10.5)
fig.savefig(pj(output_folder, 'protocol_repartition_depth.png'))
plt.close('all')
def remove(session, item):
"""Remove a given item from the database.
Args:
session: (DBSession)
item: (ContentItem)
Returns:
(True)
"""
pth = GalleryItem.gallery_pth(item.project)
# remove associated thumbnail
thumb_pth = pj(pth, "%s_thumb.png" % item.id)
if exists(thumb_pth):
remove(thumb_pth)
# remove associated resource if in gallery
if item.url.startswith("seeweb:"):
img_name = item.url.split("/")[-1]
img_pth = pj(pth, img_name)
if exists(img_pth):
remove(img_pth)
# delete item
session.delete(item)
return True
def protocol_repartition_aggregated(folder, BIN=25):
output_folder = pj(folder, "results", "protocol_repartition")
if not os.path.exists(output_folder):
os.makedirs(output_folder)
# depth_df = pd.read_csv(pj(folder, "results", "depth.csv"))
# depth_df.set_index("node", inplace=True)
pcap_df = pd.read_csv(pj(folder, "results", "pcap_relooked.csv"))
pcap_df["bin_start"] = BIN * (pcap_df.time // BIN)
res = pd.DataFrame()
RATE = 250000
res["udp"] = pcap_df[(pcap_df.icmpv6_type == "udp") & (pcap_df.time < 200)].groupby("bin_start").sum().length
res["rpl"] = pcap_df[(pcap_df.icmpv6_type == "rpl") & (pcap_df.time < 200)].groupby("bin_start").sum().length
res["rpl"] = 8.0 * res["rpl"] / RATE
res["udp"] = 8.0 * res["udp"] / RATE
ax = res[["rpl", "udp"]].plot(kind="bar", stacked=True)
ax.set_ylabel('Time [s]')
ax.set_xlabel("Time [s]")
ax.set_ylim(0, 4.5)
ax.set_xticklabels(res.index.map(int), rotation=0)
fig = plt.gcf()
plt.tight_layout()
# fig.set_size_inches(18.5, 10.5)
fig.savefig(pj(output_folder, 'protocol_repartition_aggregated.png'))
plt.close('all')
def overhead(folder):
"""
Plot the overhead (RPL, ACK,...).
This graph measures the amount of bytes sent by nodes that are not
application oriented (UDP, ping or CoAP) therefore we can see the amount
of bytes transmitted just to keep the network alive. This packets are
usually RPL and ACK packets.)
"""
fig = plt.figure()
ax1 = fig.add_subplot(111)
ax1.set_title('RPL traffic by time')
ax1.set_xlabel('Time (s)')
ax1.set_ylabel('RPL traffic (bytes)')
with open(pj(folder, "results", "io.csv")) as io_csv_f:
reader = DictReader(io_csv_f)
time, overhead_bytes = [], []
for row in reader:
time.append(float(row["bin_end"]))
overhead_bytes.append(float(row["total_bytes"])
- float(row["udp_bytes"])
- float(row["ping_bytes"])
- float(row["coap_bytes"]))
ax1.plot(time, overhead_bytes)
img_path = pj(self.img_dir, "overhead.png")
fig.savefig(img_path)
plt.close('all')
def generate(self, key, stage, piece):
"""Handles responsibility for replacing 'gen' with the correct
value"""
value = ''
if key == 'images':
value = self.generate_images(stage, piece)
if key == 'source':
raw_image = self.raw[0]
(dirname, base ) = os.path.split(raw_image)
pre = self.find_prefix(stage, piece)
value = pj(dirname, '%s%s%s' % ('mean', pre[1:], base))
if key == 'session':
for n_run in range(1,self.n_runs+1):
value += self.generate_session(n_run, piece)
if key == 'directory':
value = self.analysis_dir(piece['name'])
if key == 'spm_mat_file':
value = pj(self.analysis_dir(piece['name']), 'SPM.mat')
if key == 'contrast':
contrasts = self.paradigm['contrasts']
# for multiple model paradigms, contrasts will be a dictionary
if piece['name'] in contrasts:
contrasts = contrasts[piece['name']]
for n_con, contrast in enumerate(contrasts):
value += self.generate_contrast( n_con + 1, contrast, piece)
return value
def msgbakloop(_dir,q,sid):
_now = datetime.now()
_sdir = str(_now.month)
_np = pj(pj(_dir,q),_sdir)
for i in range(2,8): #
savedaymsg(sid,q,i,_np)
print ' ----------------\nDone!'
def run_model(i, task, params, include, group_only):
print '---', i, task, params, include, group_only, '---'
# create the file structure
root = ('congruence_and_modulation_affects_' +
'_'.join(p for p in params))
path = pj(
root, task, '_'.join(p for p in include),
'_'.join(p for p in group_only))
samples_filename = pj(path, 'samples_%i.pickle' % i)
model_filename = pj(path, 'model_%i.pickle' % i)
if not os.path.exists(path):
os.makedirs(path)
# fit and save the model
s = "%s~C(congruence, Treatment('control'))*C(modulation, Treatment('slow'))"
design = [s % p for p in params]
if not os.path.exists(model_filename):
data = pandas.read_pickle('data.pickle')
print data.head()
data = data[(data.task == task)]
model = hddm.HDDMRegressor(data,
design,
include=include,
group_only_nodes=group_only)
model.find_starting_values()
model.sample(10000, burn=500,
dbname=samples_filename, db='pickle')
model.save(model_filename)
return
def make(tool, mainFD):
makeFD=codecs.open(pj(args.reportOutDir, tool, "make.txt"), "w", encoding="utf-8")
try:
if not args.disableMake:
# make
errStr=""
if not args.disableMakeClean:
print("\n\nRUNNING make clean\n", file=makeFD); makeFD.flush()
if subprocess.call(["make", "clean"], stderr=subprocess.STDOUT, stdout=makeFD)!=0: errStr=errStr+"make clean failed; "
print("\n\nRUNNING make -k\n", file=makeFD); makeFD.flush()
if subprocess.call(["make", "-k", "-j", str(args.j)], stderr=subprocess.STDOUT, stdout=makeFD)!=0: errStr=errStr+"make failed; "
if not args.disableMakeInstall:
print("\n\nRUNNING make install\n", file=makeFD); makeFD.flush()
if subprocess.call(["make", "-k", "install"], stderr=subprocess.STDOUT, stdout=makeFD)!=0: errStr=errStr+"make install failed; "
if errStr!="": raise RuntimeError(errStr)
else:
print("make disabled", file=makeFD); makeFD.flush()
result="done"
except RuntimeError as ex:
result=str(ex)
if not args.disableMake:
print('<td class="%s"><span class="glyphicon glyphicon-%s"></span> '%("success" if result=="done" else "danger",
"ok-sign alert-success" if result=="done" else "exclamation-sign alert-danger"), file=mainFD)
print(' <a href="'+myurllib.pathname2url(pj(tool, "make.txt"))+'">'+result+'</a>', file=mainFD)
print('</td>', file=mainFD)
makeFD.close()
mainFD.flush()
if result!="done":
return 1
return 0
def tree(dname, padding, txt):
pkg_cfg = dict(base={"namespace": None, "owner": "owner", "pkgname": "pkgname", "url": None})
env = create_env(pkg_cfg)
files = [(isdir(pj(dname, fname)), fname) for fname in listdir(dname)]
files.sort()
count = 0
for is_dir, fname in files:
count += 1
txt += padding + "|\n"
fmt_name = nn(env, fname)
txt += padding + "+-" + fmt_name
path = pj(dname, fname)
if is_dir:
txt += "/\n"
if count == len(files):
txt = tree(path, padding + " " + " " * (len(fmt_name) / 2), txt)
else:
txt = tree(path, padding + "|" + " " * (len(fmt_name) / 2), txt)
else:
txt += "\n"
return txt
def message(folder):
"""
Message queue preparation
- Extract from powertracker all the message
- Extract from the serial logs all the message.
IMPORTANT: We only extract the message received from the root or send by
the root.
186572 ID:2 DATA send to 1 'Hello 1'
187124 ID:8 DATA recv 'Hello 1' from 2
197379 ID:8 REPLY send to 7 'Reply 1'
197702 ID:7 REPLY recv 'Reply 1' from 8
"""
stats = defaultdict(float)
sorted_messages = sorted(
chain(serial2message(folder, stats),
# powertracker2message(folder, stats)
),
key=operator.attrgetter("time"))
print_stats(stats)
# MESSAGE logging for debugging purposes #
log.info(pj(folder, "results", "messages.csv"))
with open(pj(folder, "results", "messages.csv"), "w") as f:
writer = DictWriter(f, fieldnames=fieldnames)
writer.writeheader()
for m in sorted_messages:
writer.writerow(m._asdict())
log.info("messages saved")
return sorted_messages
def run(self):
dst_path = "dist/ooni-resources.tar.gz"
try:
self.find_executables()
except ExecutableNotFound as enf:
print("ERR: Could not find '{0}'".format(enf.message))
return
tmp_dir = tempfile.mkdtemp()
pkg_dir = tempfile.mkdtemp()
os.mkdir(pj(pkg_dir, "resources"))
os.mkdir(pj(pkg_dir, "GeoIP"))
try:
geoip_asn_path = self.download(GEOIP_ASN_URL, tmp_dir, "GeoIPASNum.dat.gz")
except Exception as exc:
print(exc.message)
return
try:
geoip_path = self.download(GEOIP_URL, tmp_dir, "GeoIP.dat.gz")
except Exception as exc:
print(exc.message)
return
try:
test_lists_path = self.download(TEST_LISTS_URL, tmp_dir, "master.zip")
except Exception as exc:
print(exc.message)
return
run_command([self.gunzip, geoip_asn_path])
run_command([self.gunzip, geoip_path])
run_command([self.unzip, "-d", tmp_dir, test_lists_path])
shutil.move(pj(tmp_dir, "GeoIP.dat"),
pj(pkg_dir, "GeoIP", "GeoIP.dat"))
shutil.move(pj(tmp_dir, "GeoIPASNum.dat"),
pj(pkg_dir, "GeoIP", "GeoIPASNum.dat"))
shutil.move(pj(tmp_dir, "test-lists-master", "lists"),
pj(pkg_dir, "resources", "citizenlab-test-lists"))
# Don't include services and official lists
shutil.rmtree(
pj(pkg_dir,
"resources",
"citizenlab-test-lists",
"services"),
ignore_errors=True)
shutil.rmtree(
pj(pkg_dir,
"resources",
"citizenlab-test-lists",
"official"),
ignore_errors=True)
run_command([self.tar, "cvzf", dst_path, "-C", pkg_dir, "."])
# Cleanup
shutil.rmtree(pkg_dir, ignore_errors=True)
shutil.rmtree(tmp_dir, ignore_errors=True)
print("Written ooniresources to {0}".format(dst_path))
def _build_kernel_specs(self):
"""Returns the dict of name -> kernel_spec for all environments"""
# This is called much too often and the conda calls are really expensive :-(
if hasattr(self, "_build_kernel_specs_cache"):
return getattr(self, "_build_kernel_specs_cache")
venv_dirs = self.find_envs()
kspecs = {}
if platform.system() == "Windows":
python_exe_name = "python.exe"
else:
python_exe_name = "python"
for venv_name, venv_dir in venv_dirs.items():
# conda on windows has python.exe directly in the env
exe_name = pj(venv_dir, python_exe_name)
if not os.path.exists(exe_name):
exe_name = pj(venv_dir, "bin", python_exe_name)
kspec_dict = {"argv": [exe_name,
"-m",
"IPython.kernel",
"-f",
"{connection_file}"],
"language": "python",
"display_name": self.display_name_template.format(venv_name),
"env": {}}
# This should probably use self.kernel_spec_class instead of the direct class
kspecs.update({venv_name: KernelSpec(resource_dir=RESOURCES, **kspec_dict)})
self._build_kernel_specs_cache = kspecs
return kspecs
def check(tool, mainFD):
checkFD=codecs.open(pj(args.reportOutDir, tool, "check.txt"), "w", encoding="utf-8")
if not args.disableMakeCheck:
# make check
print("RUNNING make check\n", file=checkFD); checkFD.flush()
if simplesandbox.call(["make", "-j", str(args.j), "check"], envvar=["PKG_CONFIG_PATH", "LD_LIBRARY_PATH"], shareddir=["."],
stderr=subprocess.STDOUT, stdout=checkFD)==0:
result="done"
else:
result="failed"
else:
print("make check disabled", file=checkFD); checkFD.flush()
result="done"
foundTestSuiteLog=False
testSuiteLogFD=codecs.open(pj(args.reportOutDir, tool, "test-suite.log.txt"), "w", encoding="utf-8")
for rootDir,_,files in os.walk('.'): # append all test-suite.log files
if "test-suite.log" in files:
testSuiteLogFD.write('\n\n')
testSuiteLogFD.write(open(pj(rootDir, "test-suite.log")).read())
foundTestSuiteLog=True
testSuiteLogFD.close()
if not args.disableMakeCheck:
print('<td class="%s"><span class="glyphicon glyphicon-%s"></span> '%("success" if result=="done" else "danger",
"ok-sign alert-success" if result=="done" else "exclamation-sign alert-danger"), file=mainFD)
print(' <a href="'+myurllib.pathname2url(pj(tool, "check.txt"))+'">'+result+'</a>', file=mainFD)
if foundTestSuiteLog:
print(' <a href="'+myurllib.pathname2url(pj(tool, "test-suite.log.txt"))+'">test-suite.log</a>', file=mainFD)
print('</td>', file=mainFD)
checkFD.close()
mainFD.flush()
if result!="done":
return 1
return 0
# encoding: utf-8
# pylint: skip-file
"""
This file contains tests for the madmom.features.tempo module.
"""
from __future__ import absolute_import, division, print_function
import unittest
from os.path import join as pj
from . import ACTIVATIONS_PATH
from madmom.features.tempo import *
act_file = np.load(pj(ACTIVATIONS_PATH, "sample.beats_blstm.npz"))
act = act_file['activations'].astype(np.float)
fps = float(act_file['fps'])
COMB_TEMPI = np.array([[176.470, 0.475], [117.647, 0.177], [240.0, 0.154],
[68.966, 0.099], [82.192, 0.096]])
HIST = interval_histogram_comb(act, 0.79, min_tau=24, max_tau=150)
class TestIntervalHistogramAcfFunction(unittest.TestCase):
def test_values(self):
hist = interval_histogram_acf(act, min_tau=24, max_tau=150)
self.assertTrue(
np.allclose(hist[0][:6], [
0.10034907, 0.10061631, 0.11078519, 0.13461014, 0.17694432,
def setUp(self):
self.bin = pj(program_path, "BarTracker")
self.act = np.load(pj(ACTIVATIONS_PATH, 'sample.bar_tracker.npz'))
self.beats = [[0.091, 1], [0.8, 2], [1.481, 3], [2.148, 1]]
self.downbeats = [0.091, 2.148]
try:
from cStringIO import StringIO
except ImportError:
from io import StringIO
import numpy as np
from madmom.features import Activations
from madmom.features.chords import load_chords
from madmom.evaluation.key import load_key
from . import AUDIO_PATH, ACTIVATIONS_PATH, ANNOTATIONS_PATH, DETECTIONS_PATH
tmp_act = tempfile.NamedTemporaryFile(delete=False).name
tmp_result = tempfile.NamedTemporaryFile(delete=False).name
sample_file = pj(AUDIO_PATH, 'sample.wav')
sample2_file = pj(AUDIO_PATH, 'sample2.wav')
sample_file_22050 = pj(AUDIO_PATH, 'sample_22050.wav')
sample_beats = pj(ANNOTATIONS_PATH, 'sample.beats')
stereo_sample_file = pj(AUDIO_PATH, 'stereo_sample.wav')
program_path = os.path.dirname(os.path.realpath(__file__)) + '/../bin/'
# prevent writing compiled Python files to disk
sys.dont_write_bytecode = True
def run_program(program):
# import module, capture stdout
test = imp.load_source('test', program[0])
sys.argv = program
backup = sys.stdout
inputs_v = moving_average(np.random.randn(seq_size),
10).reshape(seq_size, batch_size, 1)
sess = tf.Session()
saver = tf.train.Saver()
model_fname = env.run("glm_model.ckpt")
if os.path.exists(model_fname):
print "Restoring from {}".format(model_fname)
saver.restore(sess, model_fname)
epochs = 0
else:
sess.run(tf.global_variables_initializer())
reward_v, reward_mean_v = None, None
tmp_dir = pj(os.environ["HOME"], "tmp")
[os.remove(pj(tmp_dir, f)) for f in os.listdir(tmp_dir) if f[-4:] == ".png"]
epochs = 1000
for e in xrange(epochs):
state_v = GLMStateTuple(
np.zeros((batch_size, net_size)),
np.zeros((batch_size, filters_num + net_size)),
np.zeros((batch_size, net_size)),
np.zeros((batch_size, net_size)),
np.zeros((batch_size, L, net_size)),
np.zeros((batch_size, filters_num + net_size, net_size)),
np.zeros((batch_size, L, filters_num)),
np.zeros((batch_size, L, 1)),
)
from sklearn.ensemble import GradientBoostingRegressor
from sklearn.linear_model import LassoCV
from sklearn.linear_model import RidgeCV
from sklearn.cross_decomposition import PLSRegression
from sklearn.model_selection import GridSearchCV
from sklearn.pipeline import Pipeline
from sklearn.model_selection import cross_val_score, cross_val_predict
# %%
data_dir = Path('path/to/Data').resolve()
results_dir = Path('path/to/results').resolve()
# %% ## Loading and cleaning data
data = pd.read_excel(
pj(data_dir, 'macro_dataset'),
header=0,
sep=',', #### load data as pandas dataframe, set sample IDs as index
index_col='Subject_ID')
data.dropna(axis='columns', how='any', inplace=True)
data = data[data['report_bias'] ==
0] ### dropping data from biased questionaires
# %% ### Select features and target columns of the dataset (continuous and categorical)
feats = [column for column in data.columns if 'delta' in column]
feats_cat = [
column for column in data.columns
def main(home_dir):
from os.path import join as pj
# Check if we are in the top of the ABINIT source tree
my_name = os.path.basename(__file__) + ".main"
if (not os.path.exists(pj(home_dir, "configure.ac"))
or not os.path.exists(pj(home_dir, "src/98_main/abinit.F90"))):
print("%s: You must be in the top of an ABINIT source tree." % my_name)
print("%s: Aborting now." % my_name)
sys.exit(1)
# Init
cnf_bin = MyConfigParser()
cnf_fname = pj(home_dir, "config/specs/binaries.conf")
cnf_bin.read(cnf_fname)
bin_list = cnf_bin.sections()
bin_list.sort()
dep_order = dict()
lib_order = dict()
re_num = re.compile("[0-9][0-9]_")
# Check order of dependencies and libraries
for prg in bin_list:
if cnf_bin.has_option(prg, "dependencies"):
bin_deps = cnf_bin.get(prg, "dependencies").split()
else:
bin_deps = list()
dep_old = 100
dep_new = 100
for dep in bin_deps:
if dep in dep_levels:
dep_new = dep_levels[dep]
else:
sys.stderr.write("%s: Error: unregistered dependency '%s'\n" %
(my_name, dep))
sys.exit(10)
if dep_new > dep_old:
if not prg in dep_order:
dep_order[prg] = list()
dep_order[prg].append(dep)
dep_old = dep_new
if cnf_bin.has_option(prg, "libraries"):
bin_libs = cnf_bin.get(prg, "libraries").split()
else:
bin_libs = list()
lib_old = 100
lib_new = 100
for lib in bin_libs:
if re_num.match(lib):
lib_new = int(re.sub("_.*", "", lib))
if lib_new > lib_old:
if not prg in lib_order:
lib_order[prg] = list()
lib_order[prg].append(lib)
lib_old = lib_new
# Report any disorder
nerr = len(dep_order) + len(lib_order)
if nerr > 0:
sys.stderr.write("%s: reporting disordered libraries\n\n" %
(os.path.basename(sys.argv[0])))
sys.stderr.write("X: D=Dependency / L=Library\n\n")
sys.stderr.write("%s %-24s %-24s\n" %
("X", "Program", "Dependency/Library"))
sys.stderr.write("%s %s %s\n" % ("-", "-" * 24, "-" * 24))
dep_keys = list(dep_order.keys())
dep_keys.sort()
for prg in dep_keys:
for dep in dep_order[prg]:
sys.stderr.write("%s %-24s %-24s\n" % ("D", prg, dep))
lib_keys = list(lib_order.keys())
lib_keys.sort()
for prg in lib_keys:
for lib in lib_order[prg]:
sys.stderr.write("%s %-24s %-24s\n" % ("L", prg, lib))
sys.stderr.write("\n")
return nerr
import os
import json
from os.path import join as pj
from fabric.api import task
from jinja2 import Environment, FileSystemLoader
# Default values
ROOT_DIR = os.path.dirname(__file__)
EXPERIMENT_FOLDER = pj(ROOT_DIR, "experiments")
TEMPLATE_FOLDER = pj(ROOT_DIR, "templates")
TEMPLATE_ENV = Environment(loader=FileSystemLoader(TEMPLATE_FOLDER))
@task
def my_special_function(name):
""" This function will create 42 C files. """
path = pj(EXPERIMENT_FOLDER, name)
if not os.path.exists(path):
os.makedirs(path)
c_template = TEMPLATE_ENV.get_template("dummy_template.c")
# We make the id start at 1 and finish at 42
for value in range(1, 43):
with open(pj(path, "dummy_%d.c" % value), "w") as f:
f.write(c_template.render(my_value=value))
# If you change the platform target and want to push to iotlab
# don't forget to update the nodes names
makefile_template = TEMPLATE_ENV.get_template("dummy_makefile")
with open(pj(path, "Makefile"), "w") as f:
def _load_gml(name, filename):
return ImportedGraphBuilder() \
.set_name(name) \
.import_gml(pj(Datasets.DATASETS_ROOT_PATH, name, filename)) \
.build()
def _load_polbooks_or_football(name, nodes_path, edges_path):
return ImportedGraphBuilder() \
.set_name(name) \
.import_nodes(pj(Datasets.DATASETS_ROOT_PATH, name, nodes_path), name_col_idx='idx', class_col_idx=2) \
.import_edges(pj(Datasets.DATASETS_ROOT_PATH, name, edges_path)) \
.build()
class Datasets:
DATASETS_ROOT_PATH = pj(os.path.dirname(os.path.abspath(__file__)), 'datasets')
_lazy_datasets = {
# 'as': lambda: Datasets._load_altsoph('as', 'as.clusters', 'as.edges'), # TOO BIG
# 'citeseer': lambda: Datasets._load_webkb_like('citeseer', 'citeseer.nodes', 'citeseer.edges'), # BROKEN
# 'cora_full': lambda: Datasets._load_altsoph('cora_full', '_old.clusters', '_old.edges'), # TOO BIG
'dolphins': lambda: Datasets._load_altsoph('dolphins', 'dolphins.clusters', 'dolphins.edges'),
'eu-core': lambda: Datasets._load_altsoph('eu-core', 'eu-core.clusters', 'eu-core.edges'),
'football': lambda: Datasets._load_gml('football', 'football.gml'),
'karate': lambda: Datasets._load_altsoph('karate', 'karate.clusters', 'karate.edges'),
'news_2cl_1': lambda: Datasets._load_newsgroup('news_2cl_1', 'news_2cl_1_classeo.csv', 'news_2cl_1_Docr.csv'),
'news_2cl_2': lambda: Datasets._load_newsgroup('news_2cl_2', 'news_2cl_2_classeo.csv', 'news_2cl_2_Docr.csv'),
'news_2cl_3': lambda: Datasets._load_newsgroup('news_2cl_3', 'news_2cl_3_classeo.csv', 'news_2cl_3_Docr.csv'),
'news_3cl_1': lambda: Datasets._load_newsgroup('news_3cl_1', 'news_3cl_1_classeo.csv', 'news_3cl_1_Docr.csv'),
'news_3cl_2': lambda: Datasets._load_newsgroup('news_3cl_2', 'news_3cl_2_classeo.csv', 'news_3cl_2_Docr.csv'),
'news_3cl_3': lambda: Datasets._load_newsgroup('news_3cl_3', 'news_3cl_3_classeo.csv', 'news_3cl_3_Docr.csv'),
'news_5cl_1': lambda: Datasets._load_newsgroup('news_5cl_1', 'news_5cl_1_classeo.csv', 'news_5cl_1_Docr.csv'),
'news_5cl_2': lambda: Datasets._load_newsgroup('news_5cl_2', 'news_5cl_2_classeo.csv', 'news_5cl_2_Docr.csv'),
'news_5cl_3': lambda: Datasets._load_newsgroup('news_5cl_3', 'news_5cl_3_classeo.csv', 'news_5cl_3_Docr.csv'),
'polblogs': lambda: Datasets._load_gml('polblogs', 'polblogs.gml'),
'polbooks': lambda: Datasets._load_gml('polbooks', 'polbooks.gml'),
# 'webkb_cornel': lambda: Datasets._load_webkb('webkb_cornell', 'cornell/webkb-cornell.nodes',
# 'cornell/webkb-cornell.edges'), # POSSIBLY BROKEN
# 'webkb_texas': lambda: Datasets._load_webkb('webkb_texas', 'texas/webkb-texas.nodes',
# 'texas/webkb-texas.edges'), # POSSIBLY BROKEN
# 'webkb_washington': lambda: Datasets._load_webkb('webkb_washington', 'washington/webkb-washington.nodes',
# 'washington/webkb-washington.edges'), # POSSIBLY BROKEN
# 'webkb_wisconsin': lambda: Datasets._load_webkb('webkb_wisconsin', 'wisconsin/webkb-wisconsin.nodes',
# 'wisconsin/webkb-wisconsin.edges') # POSSIBLY BROKEN
}
_loaded_datasets = {}
@property
def newsgroup(self):
newsgroup_names = ['news_2cl_1', 'news_2cl_2', 'news_2cl_3',
'news_3cl_1', 'news_3cl_2', 'news_3cl_3',
'news_5cl_1', 'news_5cl_2', 'news_5cl_3']
return [self[x] for x in newsgroup_names]
@property
def webkb(self):
webkb_names = ['webkb_cornel', 'webkb_texas', 'webkb_washington', 'webkb_wisconsin']
return [self[x] for x in webkb_names]
@property
def all(self):
return [self[x] for x in self._lazy_datasets.keys()]
@staticmethod
def _load_altsoph(name, nodes_path, edges_path):
return ImportedGraphBuilder() \
.set_name(name) \
.import_nodes(pj(Datasets.DATASETS_ROOT_PATH, name, nodes_path)) \
.import_edges(pj(Datasets.DATASETS_ROOT_PATH, name, edges_path)) \
.build()
@staticmethod
def _load_polbooks_or_football(name, nodes_path, edges_path):
return ImportedGraphBuilder() \
.set_name(name) \
.import_nodes(pj(Datasets.DATASETS_ROOT_PATH, name, nodes_path), name_col_idx='idx', class_col_idx=2) \
.import_edges(pj(Datasets.DATASETS_ROOT_PATH, name, edges_path)) \
.build()
@staticmethod
def _load_polblogs_or_zachary(name, graph_path):
return ImportedGraphBuilder() \
.set_name(name) \
.import_nodes_and_edges(pj(Datasets.DATASETS_ROOT_PATH, name, graph_path)) \
.build()
@staticmethod
def _load_newsgroup(name, nodes_path, edges_path):
return ImportedGraphBuilder() \
.set_name(name) \
.import_nodes(pj(Datasets.DATASETS_ROOT_PATH, 'newsgroup', nodes_path),
name_col_idx='idx', class_col_idx=0) \
.import_adjacency_matrix(pj(Datasets.DATASETS_ROOT_PATH, 'newsgroup', edges_path)) \
.build()
@staticmethod
def _load_webkb(name, nodes_path, edges_path):
return ImportedGraphBuilder() \
.set_name(name) \
.import_nodes(pj(Datasets.DATASETS_ROOT_PATH, 'webkb', nodes_path),
startline=2, name_col_idx=0, class_col_idx=-1) \
.import_edges(pj(Datasets.DATASETS_ROOT_PATH, 'webkb', edges_path),
startline=3, node1_col_idx=1, node2_col_idx=3) \
.build()
@staticmethod
def _load_webkb_like(name, nodes_path, edges_path):
return ImportedGraphBuilder() \
.set_name(name) \
.import_nodes(pj(Datasets.DATASETS_ROOT_PATH, name, nodes_path),
startline=2, name_col_idx=0, class_col_idx=-1) \
.import_edges(pj(Datasets.DATASETS_ROOT_PATH, name, edges_path),
startline=3, node1_col_idx=1, node2_col_idx=3) \
.build()
@staticmethod
def _load_gml(name, filename):
return ImportedGraphBuilder() \
.set_name(name) \
.import_gml(pj(Datasets.DATASETS_ROOT_PATH, name, filename)) \
.build()
def __getitem__(self, item):
if item not in self._loaded_datasets:
self._loaded_datasets[item] = self._lazy_datasets[item]()
return self._loaded_datasets[item]
def __getattr__(self, name):
return self[name]
author_email = ''
url = 'http://openalea.gforge.inria.fr'
license = 'Cecill-C'
packages = [
namespace + "." + pkg for pkg in find_packages('src')
if 'vplants' not in pkg
]
print pkg_name
setup(
name=name,
version=version,
description=description,
author=author,
author_email=author_email,
url=url,
license=license,
namespace_packages=['vplants'],
create_namespaces=True,
zip_safe=False,
packages=packages,
package_dir={
pkg_name: pj('src', 'tree_reduction'),
'': 'src'
},
# Dependencies
install_requires=['openalea.deploy'],
dependency_links=['http://openalea.gforge.inria.fr/pi'],
)
def _load_polblogs_or_zachary(name, graph_path):
return ImportedGraphBuilder() \
.set_name(name) \
.import_nodes_and_edges(pj(Datasets.DATASETS_ROOT_PATH, name, graph_path)) \
.build()
def output_filename(self, filename):
folder = re.search('shared.*(?=/)', filename).group(0)
shortname = re.search('[^/]+$', filename).group(0)
return pj(self.conf['dirs']['output'], folder), shortname
def _load_altsoph(name, nodes_path, edges_path):
return ImportedGraphBuilder() \
.set_name(name) \
.import_nodes(pj(Datasets.DATASETS_ROOT_PATH, name, nodes_path)) \
.import_edges(pj(Datasets.DATASETS_ROOT_PATH, name, edges_path)) \
.build()
return int(tol[-1])
def xlat_pmax(pmax):
if '/' in pmax:
a = [float(x) for x in pmax.split("/")]
return a[0] / a[1]
else:
return float(pmax)
pool_path = os.getenv("HORIZON_POOL")
if pool_path is None:
raise IOError("need HORIOZN_POOL")
base_path = pj(pool_path, "parts", "passive", "resistor", "samsung", "rc")
gen = util.UUIDGenerator("uu.txt")
pmaxs = set()
for row in j_raw["rows"]:
mpn = row["parnum"]
pkg = row["sizcd_eia"]
rval = row["rval"][:-1]
if rval[-1] in muls:
value = float(rval[:-1]) * muls[rval[-1]]
else:
value = float(rval)
tol = xlat_tol(row['tol'])
pmax = row['ratpow']
"""
from __future__ import absolute_import, division, print_function
import os
import sys
import tempfile
import unittest
from os.path import join as pj
from madmom.io.audio import *
from . import AUDIO_PATH, DATA_PATH
from .test_audio_comb_filters import sig_1d, sig_2d
sample_file = pj(AUDIO_PATH, 'sample.wav')
sample_file_22k = pj(AUDIO_PATH, 'sample_22050.wav')
stereo_sample_file = pj(AUDIO_PATH, 'stereo_sample.wav')
tmp_file = tempfile.NamedTemporaryFile(delete=False).name
class TestLoadWaveFileFunction(unittest.TestCase):
def test_types(self):
signal, sample_rate = load_wave_file(sample_file)
self.assertIsInstance(signal, np.ndarray)
self.assertTrue(signal.dtype == np.int16)
self.assertTrue(type(sample_rate) == int)
def test_file_handle(self):
# open file handle
This file contains test functions for the madmom.utils module.
"""
from __future__ import absolute_import, division, print_function
import unittest
from os.path import join as pj
from madmom.utils import *
from . import (ACTIVATIONS_PATH, ANNOTATIONS_PATH, AUDIO_PATH, DATA_PATH,
DETECTIONS_PATH)
from .test_evaluation_notes import ANNOTATIONS as NOTES
FILE_LIST = [
pj(DATA_PATH, 'README'),
pj(DATA_PATH, 'commented_txt'),
pj(DATA_PATH, 'events.txt')
]
AUDIO_FILES = [
pj(AUDIO_PATH, 'sample.wav'),
pj(AUDIO_PATH, 'sample2.wav'),
pj(AUDIO_PATH, 'sample_22050.wav'),
pj(AUDIO_PATH, 'stereo_chirp.wav'),
pj(AUDIO_PATH, 'stereo_chirp_rg.flac'),
pj(AUDIO_PATH, 'stereo_sample.flac'),
pj(AUDIO_PATH, 'stereo_sample.m4a'),
pj(AUDIO_PATH, 'stereo_sample_rg.flac'),
pj(AUDIO_PATH, 'stereo_sample.wav')
]
def mobileNetV2_tst():
with tf.Graph().as_default():
is_training = tf.placeholder(tf.bool, [])
url = 'https://upload.wikimedia.org/wikipedia/commons/7/70/EnglishCockerSpaniel_simon.jpg'
image_string = urllib.urlopen(url).read()
image = tf.image.decode_jpeg(image_string, channels=3)
processed_image = inception_preprocessing.preprocess_image(
image, image_size, image_size, is_training=False)
processed_images = tf.expand_dims(processed_image, 0)
# print(processed_images.shape)
print(processed_images.dtype)
# Create the model, use the default arg scope to configure the batch norm parameters.
with slim.arg_scope(mobilenet_v2.training_scope(is_training=False)):
global_pool, endpoints = mobilenet_v2.mobilenet(processed_images,
num_classes=None)
# print(type(endpoints))
# print(len(endpoints.keys()))
# print(endpoints['layer_18/output'].shape) # (1, 7, 7, 320)
# print(endpoints['layer_18'].shape) # (1, 7, 7, 320)
# print(endpoints['layer_19'].shape) # (1, 7, 7, 320)
# print(endpoints['global_pool'].shape) # (1, 1, 1, 1280)
# pprint(endpoints.keys())
# variables_to_restore = slim.get_variables_to_restore(exclude=['MobilenetV2/Logits/Conv2d_1c_1x1'])
variables_to_restore = slim.get_variables_to_restore()
print(len(variables_to_restore)) # 260
restorer = tf.train.Saver(variables_to_restore)
# print(variables_to_restore)
dropout_keep_prob = 0.5
n_classes = 2
weight_decay = 0.05
with tf.variable_scope('addition', 'fc'):
# flatten = tf.flatten(endpoints['global_pool'])
flatten = slim.flatten(global_pool)
with slim.arg_scope(
[slim.fully_connected],
weights_regularizer=slim.l2_regularizer(weight_decay),
weights_initializer=tc.layers.xavier_initializer(
tf.float32),
# weights_initializer=tf.truncated_normal_initializer(stddev=0.1),
activation_fn=None) as sc:
net = slim.fully_connected(flatten,
128,
activation_fn=None,
scope='fc1')
net = slim.dropout(net,
dropout_keep_prob,
is_training=is_training,
scope='dropout')
logits = slim.fully_connected(net,
n_classes,
activation_fn=None,
scope='fc2')
# with tf.name_scope('loss') :
probabilities = tf.nn.softmax(logits)
ckptDir = r'D:\Lab408\tfslim\mobileNetV2'
ckptPath = pj(ckptDir, 'mobilenet_v2_1.0_224.ckpt')
# variables_to_restore = slim.get_variables('MobilenetV2/Logits/Conv2d_1c_1x1')
# variables_to_save = slim.get_variables_to_restore(exclude=['MobilenetV2/Logits/Conv2d_1c_1x1'])
variables_to_save = slim.get_variables_to_restore()
print(len(variables_to_save)) # 264 两层 fc 4个参数
saver = tf.train.Saver(variables_to_save)
# print(variables_to_save)
# pprint(variables_to_restore)
# variables_to_restore = slim.get_model_variables()
# print(len(variables_to_restore))
op_init1 = tf.variables_initializer(tf.global_variables())
op_init2 = tf.variables_initializer(tf.local_variables())
op_group = tf.group(op_init1, op_init2)
saveto = r'D:\Lab408\tfslim\mobileNetV2-finetue\a2'
# ckptPath = saveto
# variables_to_restore = variables_to_save
saveto = r'D:\Lab408\tfslim\mobileNetV2-finetue\a3'
init_fn0 = slim.assign_from_checkpoint_fn(ckptPath,
variables_to_restore)
sess_conf = tf.ConfigProto()
sess_conf.gpu_options.allow_growth = True
with tf.Session(config=sess_conf) as sess:
sess.run(op_group) # fc参数需要初始化
init_fn0(sess)
np_image, probabilities = sess.run([image, probabilities],
feed_dict={is_training: False})
probabilities = probabilities[0, 0:]
sorted_inds = [
i[0]
for i in sorted(enumerate(-probabilities), key=lambda x: x[1])
]
# restorer.save(sess, saveto)
saver.save(sess, saveto)
names = imagenet.create_readable_names_for_imagenet_labels()
for i in range(n_classes):
index = sorted_inds[i]
outstr = 'Probability {:.2%} => {}'.format(probabilities[index],
names[index])
print(outstr)
# print('Probability %0.2f%% => [%s]' % (probabilities[index] * 100, names[index]))
# Probability 62.93% => tench, Tinca tinca
# Probability 37.07% => background
return
def test_write_and_read_events(self):
write_events(EVENTS, pj(DATA_PATH, 'events.txt'))
annotations = load_events(pj(DATA_PATH, 'events.txt'))
self.assertTrue(np.allclose(annotations, EVENTS))
def test_errors(self):
with self.assertRaises(IOError):
search_path(pj(DATA_PATH, 'README'))
def test_write_events_to_file(self):
result = write_events(EVENTS, pj(DATA_PATH, 'events.txt'))
self.assertEqual(EVENTS, result)
def get_profile_folder_by_name(self, profname=None):
if profname is None:
profname = self.conf['prof']['name']
return pj(self.conf['prof']['basedir'], profname)
def test_load_file_with_comments_and_empty_lines(self):
events = load_events(pj(DATA_PATH, 'commented_txt'))
self.assertTrue(np.allclose(events, [1.1, 2.1]))
def test_write_events_to_file_handle(self):
file_handle = open(pj(DATA_PATH, 'events.txt'), 'wb')
result = write_events(EVENTS, file_handle)
self.assertEqual(EVENTS, result)
file_handle.close()
def test_read_events_from_file_handle(self):
file_handle = open(pj(DATA_PATH, 'events.txt'))
events = load_events(file_handle)
self.assertIsInstance(events, np.ndarray)
file_handle.close()
def test_load_only_timestamps(self):
events = load_events(pj(ANNOTATIONS_PATH, 'stereo_sample.notes'))
self.assertTrue(
np.allclose(
events,
[0.147, 1.567, 2.526, 2.549, 2.563, 2.577, 3.369, 3.449]))
def main():
cwd = cf.cwd
log_dir = pj(cwd, 'train_eval_log')
ckpt_dir = path.join(log_dir, 'ckpts')
videoRoot = cf.videoRoot
# videoRoot = pj(cwd, 'all_data')
labeljson = cf.labeljson
evalSet = cf.evalSet
# evalSet = [47, 48, 49, 51, 52, 59, 61, 62, 63, 65]
# 47: {'fire': 1601, 'fireless': 57},
# 48: {'fire': 3748, 'fireless': 98},
# 49: {'fire': 3714, 'fireless': 40},
# 51: {'fire': 4120, 'fireless': 21},
# 52: {'fire': 4451, 'fireless': 45},
# 59: {'fire': 6911, 'fireless': 70},
# 61: {'fire': 1298, 'fireless': 0},
# 62: {'fire': 3275, 'fireless': 0},
# 63: {'fire': 5055, 'fireless': 0},
# 65: {'fire': 6913, 'fireless': 64},
if not path.exists(log_dir):
os.mkdir(log_dir)
if not path.exists(ckpt_dir):
os.mkdir(ckpt_dir)
eval_interval = 20
save_summary_steps = 5
save_ckpts_steps = 100
train_batchsz = 50
eval_batchsz = 100
# eval_steps = 40
epoch = 900
img_num = 870 * 2
max_steps = (img_num * epoch) // train_batchsz
# ------------------------------ prepare input ------------------------------
_h = 240
_w = 320
dset = cnn_dataset.MyDataset(videoRoot,
labeljson,
evalSet,
resize=(_h, _w))
dset.setTrainParams(train_batchsz, prefetch=10)
dset.setEvalParams(eval_batchsz, prefetch=3)
iter_dict = {'train': dset.makeTrainIter(), 'eval': dset.makeEvalIter()}
holder_handle = tf.placeholder(tf.string, [])
iter = tf.data.Iterator.from_string_handle(holder_handle,
dset.output_types)
# next_elem = iter.get_next()
inputx, labels, filename = iter.get_next()
inputx = tf.reshape(inputx, [-1, _h, _w, 3])
# eval_x.set_shape([eval_batchsz, 200, 250, 3])
# train_x, train_y, train_fname = dset.train(train_batchsz, prefetch_batch)
# train_x.set_shape([train_batchsz, 200, 250, 3])
# eval_x, eval_y, eval_fname = dset.eval(eval_batchsz, prefetch_batch)
# eval_x.set_shape([eval_batchsz, 200, 250, 3])
# \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\ build graph \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\
model = cnn.CNN('NCHW')
# model = smodel.Simplest('NHWC')
logits = model(inputx, castFromUint8=False)
with tf.name_scope('cross_entropy'):
loss = tf.losses.sparse_softmax_cross_entropy(labels, logits)
with tf.name_scope('accuracy'):
_pred = tf.argmax(logits, axis=1, output_type=tf.int32)
acc_vec = tf.equal(labels, _pred)
acc = tf.reduce_mean(tf.cast(acc_vec, tf.float32))
update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS)
with tf.name_scope('optimizer'):
optimizer = tf.train.AdamOptimizer(1e-4)
with tf.control_dependencies(update_ops):
train_op = optimizer.minimize(loss,
tf.train.get_or_create_global_step())
# |||||||||||||||||||||||||||||| hooks ||||||||||||||||||||||||||||||
# >>> logging
tf.logging.set_verbosity(tf.logging.INFO)
# global_step = tf.train.get_or_create_global_step()
# tf.identity(global_step, 'g_step')
# tf.identity(loss, 'cross_entropy')
# tf.identity(acc, 'accuracy')
# tensor_lr = optimizer._lr_t
tensors = {
'step': tf.train.get_or_create_global_step().name,
'loss': loss.name,
'accuracy': acc.name
}
logging_hook = tf.train.LoggingTensorHook(tensors=tensors, every_n_iter=10)
# >>> summary
summary_conf = {'dir': log_dir, 'saved_steps': save_summary_steps}
summary_protobuf = {
'loss': tf.summary.scalar('cross_entropy', loss),
'accuracy': tf.summary.scalar('accuracy', acc)
}
# print(list(summary_protobuf.values()))
# summary_loss = tf.summary.scalar('cross_entropy', loss)
# summary_acc = tf.summary.scalar('accuracy', acc)
# summary_lr = tf.summary.scalar('lr', optimizer._lr_t)
# global_step =
# merged_op = tf.summary.merge_all()
# summary_hook = tf.train.SummarySaverHook(
# save_steps=1,
# output_dir= ckpt_dir,
# summary_op= merged_op
# )
# >>> main run hook
eval_hook = runhooks.RunHook(
iter_dict=iter_dict,
eval_steps=eval_interval,
train_op=train_op,
training=model.is_training,
holder_handle=holder_handle,
summary_conf=summary_conf,
summary_protobuf=summary_protobuf,
)
# >>> checkpoint saver
ckpt_saver_hook = runhooks.CkptSaverHook(ckpt_dir,
save_steps=save_ckpts_steps)
# ckpt_saver_hook = tf.train.CheckpointSaverHook(
# checkpoint_dir= ckpt_dir,
# save_steps= save_ckpts_steps,
# )
all_hooks = [
# logging_hook,
# summary_hook,
eval_hook,
ckpt_saver_hook,
# tf.train.StopAtStepHook(max_steps),
# tf.train.NanTensorHook(loss)
]
# ////////////////////////////// session config //////////////////////////////
sess_conf = tf.ConfigProto()
sess_conf.gpu_options.allow_growth = True
# sess_conf.gpu_options.per_process_gpu_memory_fraction = 0.9
sess_creator = tf.train.ChiefSessionCreator(
# scaffold=scaffold,
# master='',
config=sess_conf,
checkpoint_dir=ckpt_dir)
# print('end')
# return
# ------------------------------ start ------------------------------
with tf.train.MonitoredSession(session_creator=sess_creator,
hooks=all_hooks,
stop_grace_period_secs=3600) as mon_sess:
while not mon_sess.should_stop():
step = mon_sess.run(tf.train.get_global_step()
) # arg from retval of _EvalHook before_run()
# training, step = mon_sess.run([model.is_training, tf.train.get_global_step()]) # arg from retval of _EvalHook before_run()
# if not training:
# print('step {}: eval xxxxxxxxx'.format(step))
# print(lr)
return
def test_read_onset_annotations(self):
events = load_events(pj(ANNOTATIONS_PATH, 'sample.onsets'))
self.assertTrue(np.allclose(events, ONSET_ANNOTATIONS))
import time
import sys
import os
import os.path as path
from os.path import join as pj
import tensorflow as tf
import var_config as cf
import cnn_dataset
import cnn
repoRoot = r'E:\github_repo\tensorflow_intro_practice'
sys.path.append(pj(repoRoot, 'proj', 'needle_mushroom'))
import runhooks
def main():
cwd = cf.cwd
log_dir = pj(cwd, 'train_eval_log')
ckpt_dir = path.join(log_dir, 'ckpts')
videoRoot = cf.videoRoot
# videoRoot = pj(cwd, 'all_data')
labeljson = cf.labeljson
evalSet = cf.evalSet
# evalSet = [47, 48, 49, 51, 52, 59, 61, 62, 63, 65]
# 47: {'fire': 1601, 'fireless': 57},
# 48: {'fire': 3748, 'fireless': 98},
# 49: {'fire': 3714, 'fireless': 40},
# 51: {'fire': 4120, 'fireless': 21},
# 52: {'fire': 4451, 'fireless': 45},
# 59: {'fire': 6911, 'fireless': 70},
def test_read_events_from_file(self):
events = load_events(pj(DATA_PATH, 'events.txt'))
self.assertIsInstance(events, np.ndarray)
