def process_results(self, data, re_test=None): results = [] if data['status'] != 'success': ADDON.log(data) return [] gc.disable() for result in data['result']: title = self.normalize(result['title']) if re_test: if re_test.search(title) is None: continue sourcetitle = self.normalize(result['sourcetitle']) hoster = result['hosterurls'] extension = result['extension'] size = result['sizeinternal'] extension = result['extension'] host_name = result['hostername'] hosts = result['hosterurls'] if extension == 'rar': continue for host in hosts: if self.filter_host(host_name): url = "%s://%s" % (self.service, host['url']) quality = self.test_quality(title+sourcetitle+self.normalize(url)) result = ScraperResult(self.debrid_hosts, self.service, host_name, url, title) result.quality = quality result.size = int(size) result.extension = extension results += [result] gc.enable() return results
def test(self, view):
"""
Calls the given view and measures the time for it to return. The
garbage collector is diabled during execution.
"""
gc_old = gc.isenabled()
gc.disable()
try:
start = timeit.default_timer()
if view.method == 'GET':
response = self.client.get(view.url, view.data)
elif view.method == 'POST':
response = self.client.post(view.url, view.data)
else:
raise ValueError('Unknown view method: %s' % view.method)
end = timeit.default_timer()
# Return result in milliseconds
time_ms = (end - start) * 1000
# Try to get version information
version = subprocess.check_output(['git', 'describe'])
from .models import TestResult
return TestResult(view=view, time=time_ms, result=response,
result_code=response.status_code, version=version)
finally:
if gc_old:
gc.enable()
def test_trashcan(self):
class Ouch:
n = 0
def __del__(self):
Ouch.n = Ouch.n + 1
if Ouch.n % 17 == 0:
gc.collect()
# "trashcan" is a hack to prevent stack overflow when deallocating
# very deeply nested tuples etc. It works in part by abusing the
# type pointer and refcount fields, and that can yield horrible
# problems when gc tries to traverse the structures.
# If this test fails (as it does in 2.0, 2.1 and 2.2), it will
# most likely die via segfault.
# Note: In 2.3 the possibility for compiling without cyclic gc was
# removed, and that in turn allows the trashcan mechanism to work
# via much simpler means (e.g., it never abuses the type pointer or
# refcount fields anymore). Since it's much less likely to cause a
# problem now, the various constants in this expensive (we force a lot
# of full collections) test are cut back from the 2.2 version.
gc.enable()
N = 150
for count in range(2):
t = []
for i in range(N):
t = [t, Ouch()]
u = []
for i in range(N):
u = [u, Ouch()]
v = {}
for i in range(N):
v = {1: v, 2: Ouch()}
gc.disable()
def VTiter(self, *parsedArgs,**envars):
largs, dictargs = self.full_parse(parsedArgs)
if 'query' not in dictargs:
raise functions.OperatorError(__name__.rsplit('.')[-1],"No query argument ")
query=dictargs['query']
cur = envars['db'].cursor()
q = cur.execute(query, parse=False)
try:
yield list(cur.getdescriptionsafe())
except StopIteration:
try:
raise
finally:
try:
c.close()
except:
pass
gc.disable()
while True:
yield q.next()
gc.enable()
def test_get_stats(self):
stats = gc.get_stats()
self.assertEqual(len(stats), 3)
for st in stats:
self.assertIsInstance(st, dict)
self.assertEqual(set(st),
{"collected", "collections", "uncollectable"})
self.assertGreaterEqual(st["collected"], 0)
self.assertGreaterEqual(st["collections"], 0)
self.assertGreaterEqual(st["uncollectable"], 0)
# Check that collection counts are incremented correctly
if gc.isenabled():
self.addCleanup(gc.enable)
gc.disable()
old = gc.get_stats()
gc.collect(0)
new = gc.get_stats()
self.assertEqual(new[0]["collections"], old[0]["collections"] + 1)
self.assertEqual(new[1]["collections"], old[1]["collections"])
self.assertEqual(new[2]["collections"], old[2]["collections"])
gc.collect(2)
new = gc.get_stats()
self.assertEqual(new[0]["collections"], old[0]["collections"] + 1)
self.assertEqual(new[1]["collections"], old[1]["collections"])
self.assertEqual(new[2]["collections"], old[2]["collections"] + 1)
def main_measure(data, dataname,agentClassGenerator,params):
X_POINTS = params[scp.X_POINTS]
STEP = params[scp.STEP]
NUM_FOLDS = params[scp.NUM_FOLDS]
CLASSIFY_TIME = params[scp.CLASSIFY_TIME]
LEARN_TIME = params[scp.LEARN_TIME]
SEED = params[scp.SEED]
num_features_arr = [ i*STEP for i in range(1,X_POINTS +1) ]
print "\n============= Learning Curve ==================="
print "Evaluating", dataname
print "num_features,", "accuracy%,"
results=[]
for num_features in num_features_arr:
agentClass = agentClassGenerator(num_features)
try:
gc.disable()
confusion = AgentAnalyzer().run_one(data, agentClass, CLASSIFY_TIME, LEARN_TIME, num_folds=NUM_FOLDS, seed=SEED)
gc.enable()
gc.collect()
idf = s_common.idf(NUM_FOLDS, data, num_features)
results.append( (num_features,confusion, idf) )
print num_features, ',', confusion.getAccuracyStr()
except Exception, e:
print "Error:",e
print " Possible Timeout for", num_features
def getlines(a_filename):
# it gives chunks
fin = None
if a_filename == '-':
fin = sys.stdin
else:
fin = open(a_filename,'r')
header = dict()
first = True
while True:
lines = fin.readlines(10**8)
if not lines:
break
gc.disable()
lines = [line.rstrip('\r\n').split('\t') for line in lines if line.rstrip('\r\n')]
gc.enable()
for line in lines:
if line[0].startswith('@'):
if line[0].startswith('@SQ') and line[1].startswith('SN:') and line[2].startswith('LN:'):
k = line[1][3:]
v = int(line[2][3:])
header[k] = v
else:
pass
else:
if first:
first = False
yield header
header = None
yield line
if first and header:
yield header
fin.close()
def _exitfunc(cls):
# At shutdown invoke finalizers for which atexit is true.
# This is called once all other non-daemonic threads have been
# joined.
reenable_gc = False
try:
if cls._registry:
import gc
if gc.isenabled():
reenable_gc = True
gc.disable()
pending = None
while True:
if pending is None or finalize._dirty:
pending = cls._select_for_exit()
finalize._dirty = False
if not pending:
break
f = pending.pop()
try:
# gc is disabled, so (assuming no daemonic
# threads) the following is the only line in
# this function which might trigger creation
# of a new finalizer
f()
except Exception:
sys.excepthook(*sys.exc_info())
assert f not in cls._registry
finally:
# prevent any more finalizers from executing during shutdown
finalize._shutdown = True
if reenable_gc:
gc.enable()
def load(self):
try:
env = Environment.Environment(os.path.join(self.cachedir, "build.config.py"))
except (IOError, OSError):
pass
else:
if env["version"] < HEXVERSION:
raise Utils.WafError("Version mismatch! reconfigure the project")
for t in env["tools"]:
self.setup(**t)
try:
gc.disable()
f = data = None
Node.Nodu = self.node_class
try:
f = open(os.path.join(self.bdir, DBFILE), "rb")
except (IOError, EOFError):
pass
try:
if f:
data = cPickle.load(f)
except AttributeError:
if Logs.verbose > 1:
raise
if data:
for x in SAVED_ATTRS:
setattr(self, x, data[x])
else:
debug("build: Build cache loading failed")
finally:
if f:
f.close()
gc.enable()
def reads_from_fastq_file(file_name, size_read_buffer = 10**8):
fid = None
if file_name == '-':
fid = sys.stdin
elif file_name.lower().endswith('.gz'):
fid = gzip.open(file_name,'r')
else:
fid = open(file_name,'r')
piece = [None,None,None,None]
ij = 0
while True:
gc.disable()
lines = fid.readlines(size_read_buffer)
gc.enable()
if not lines:
break
for line in lines:
ij = ij + 1
piece[ij-1] = line
if ij == 4:
bucket = (piece[0].rstrip('\r\n')[1:],
piece[1].rstrip('\r\n'),
piece[3].rstrip('\r\n'))
yield bucket
piece = [None,None,None,None]
ij = 0
fid.close()
def testNoReferenceCyclesAfterCall(self):
class ChildNetwork(network.Network):
def __init__(self, name=None):
super(ChildNetwork, self).__init__(name=name)
def call(self, x):
return x * 2.
class ParentNetwork(network.Network):
def __init__(self, name=None):
super(ParentNetwork, self).__init__(name=name)
self.l1 = self.track_layer(ChildNetwork())
def call(self, x):
return self.l1(x)
one = constant_op.constant([[1.0]])
gc.disable()
gc.collect()
previous_gc_debug_flags = gc.get_debug()
gc.set_debug(gc.DEBUG_SAVEALL)
preexisting = len(gc.garbage)
net = ParentNetwork()
net(one)
del net
gc.collect()
# There should be no additional garbage requiring collection.
self.assertEqual(preexisting, len(gc.garbage))
gc.set_debug(previous_gc_debug_flags)
gc.enable()
def reads_from_fastq_file(f_name,size_read_buffer=10**8):
fid = None
if f_name == '-':
fid = sys.stdin
elif f_name.lower().endswith('.gz'):
fid = gzip.open(f_name,'r')
else:
fid = open(f_name,'r')
j = 0
p1 = None
p2 = None
while True:
gc.disable()
lines = fid.readlines(size_read_buffer)
gc.enable()
if not lines:
break
for a_line in lines:
j = j + 1
if j == 1:
p1 = a_line
elif j == 2:
p2 = a_line
elif j == 4:
yield (p1,p2,a_line)
p1 = None
p2 = None
j = 0
fid.close()
def add_line(self,line):
gc.disable()
self.data.append(line)
gc.enable()
self.size = self.size + len(line)
if self.size > self.size_buffer:
self.__write_buffer()
def main():
for key_file in glob.glob("*.pem"):
with open(key_file, "r") as f:
private_key = f.read()
# Check that the signatures match.
results = []
for name, func in get_signature.items():
if not available[name]:
continue
results.append((name, func(private_key, "foo bar")))
print "{} using {}:".format(key_file, ", ".join(r[0] for r in results))
signatures = dict((r[1], True) for r in results).keys()
if len(signatures) == 1:
print " EQUAL"
else:
print " NOT EQUAL"
# Simple benchmark.
iters = 500
s = get_random_string(500)
for name, func in get_signature.items():
if not available[name]:
continue
print "running {} iterations of {}".format(iters, name)
gc.disable()
tic = time.time()
for i in range(iters):
func(private_key, s)
toc = time.time()
gc.enable()
print " took {:.3f}s".format(toc - tic)
print
def _main():
"""
main loop
"""
user = None
GAME_SELECT_DELAY = .4
while True:
gc.disable()
hardware.reset()
if user is None:
user = persistence.get_anonymous()
# do game selection by good/bad light
hardware.write_message("Waiting for a game selection"," Choose 1 - %d" % len(games)).\
display_characters('H','I')
select = hardware.select_by_lights(len(games),9)
if select == 9:
for i in xrange(5):
hardware.display_characters('B','Y')\
.wait(.3)\
.display_characters(' ',' ')\
.wait(.2)
hardware.wait(1)\
.cleanup()
exit()
# game picked, construct it
(name,description,levels,author,date,ver) = games[select-1].GameInfo()
game = games[select-1]()
level = 1
if levels > 1:
hardware.display_characters('L','E')
level = hardware.select_by_lights(levels,9)
if level == 9:
continue
hardware.display_number(0)
game.initialize(hardware,user,level)
hardware.write_message("Playing game>",name)
hardware.write_debug(description,'by',author)
score = Score().load_at_start(name,ver,level,user)
persistence.save_score_start(score,user)
start = time.time()
score.score = game.play()
score.duration_sec = time.time() - start
persistence.save_score_end(score,user)
hardware.beep(2,.5)
hardware.blink_light_until_button(5)
gc.enable()
gc.collect()
def load(self, file):
# "file" could also be a socket
gc.disable()
try:
return pickle.load(file)
finally:
gc.enable()
def loadblk(self, blk, buf):
# we are in sighandler - establish cycle which also referenced obj_4del and trigger full GC
assert self.obj_4del is not None
w = weakref.ref(self.obj_4del)
assert w() is self.obj_4del
# establish cycle with leaf ref to obj_4del
a = C()
b = C()
a.b = b
b.a = a
a.obj_4del = self.obj_4del
self.obj_4del = None
assert w() is not None
# del a=b cycle - it should stay alice, while gc is disabled
gc_save = gc.isenabled()
gc.disable()
del a, b
assert w() is not None
# gc - a=b and obj_4del collected
gc.collect()
assert w() is None
if gc_save:
gc.enable()
self.marker_list.append(2)
def f(*k, **kw):
try:
gc.disable()
ret = fun(*k, **kw)
finally:
gc.enable()
return ret
def group_edges(cs):
plus = []
minus = []
pairs = []
gc.disable()
interval = 1000
for current, cl1 in enumerate(cs.clusters):
if (current % interval) == 0:
update_status(float(current) / len(cs.clusters), "Grouping all edges...")
bib1 = tuple(cl1.bibs)[0]
pointers = cl1.out_edges
for bib2 in xrange(len(cl1.out_edges)):
val = pointers[bib2]
if val[0] not in Bib_matrix.special_numbers:
if val[0] > edge_cut_prob:
pairs.append((bib1, bib2, val))
elif val[0] == Bib_matrix.special_symbols['+']:
plus.append((bib1, bib2))
elif val[0] == Bib_matrix.special_symbols['-']:
minus.append((bib1, bib2))
else:
assert val[0] == Bib_matrix.special_symbols[None], "Invalid Edge"
update_status_final("Finished with the edge grouping.")
bibauthor_print("Positive edges: %d, Negative edges: %d, Value edges: %d."
% (len(plus), len(minus), len(pairs)))
gc.enable()
return plus, minus, pairs
def next_seed(self):
"""
Load next seed from disk
"""
seed = next(self._all_seeds)
folder = os.path.join(self._root, str(seed), self._subset)
self.data = []
silence = None
gc.disable()
for filename in os.listdir(folder):
command = os.path.splitext(os.path.basename(filename))[0]
with open(os.path.join(folder, filename), "r") as pkl_file:
audio = pickle.load(pkl_file)
# Check for 'silence'
if command == "silence":
silence = audio
else:
target = self.classes.index(os.path.basename(command))
self.data.extend(itertools.product(audio, [target]))
gc.enable()
target = self.classes.index("silence")
self.data += [(silence, target)] * int(len(self.data) * self._silence_percentage)
return seed
def timer(fxn, args):
gc.disable()
t1 = time.time()
R = fxn(*args)
t2 = time.time()
gc.enable()
return R, (t2 - t1)
def test_for_mem_leak():
from calibre.utils.mem import memory, gc_histogram, diff_hists
import gc
gc.disable()
scanner = DeviceScanner()
scanner.scan()
memory() # load the psutil library
for i in xrange(3):
gc.collect()
for reps in (1, 10, 100, 1000):
for i in xrange(3):
gc.collect()
h1 = gc_histogram()
startmem = memory()
for i in xrange(reps):
scanner.scan()
for i in xrange(3):
gc.collect()
usedmem = memory(startmem)
prints('Memory used in %d repetitions of scan(): %.5f KB'%(reps,
1024*usedmem))
prints('Differences in python object counts:')
diff_hists(h1, gc_histogram())
prints()
def handoff_all():
main_logger.info("Preparing to re-exec with handoffs")
handoff_data = {}
for name, manager in clients.items():
data = manager.handoff()
if data:
handoff_data[name] = data
main_logger.info("Final handoff data: {!r}".format(handoff_data))
for manager in clients.values():
manager.get()
main_logger.info("All managers stopped")
# TODO this will fail if the bytes contain invalid utf-8
handoff_data_str = json.dumps(handoff_data)
env = os.environ.copy()
env['handoff_data'] = handoff_data_str
main_logger.info("Calling execve({!r}, {!r}, {!r})".format(sys.executable, sys.argv, env))
# critical section - absolutely no blocking calls beyond this point
gc.disable() # we don't want any destructors running
open_fds = set(map(int, os.listdir('/proc/self/fd')))
for fd in open_fds - {0, 1, 2} - set(data['fd'] for data in handoff_data.values()):
try:
os.close(fd)
except OSError:
pass # this is probably EBADF, but even if it isn't we can't do anything about it
os.execve(sys.executable, [sys.executable, '-m', 'ekimbot'] + sys.argv[1:], env)
def arrayFP(input): outfp = [] for i in input: gc.disable() outfp.append(calcFingerprints(i[0])) gc.enable() return np.array(outfp, dtype=np.uint8)
def wrapped(self, *args, **kwargs):
import gc
gc.disable()
gc.collect()
deltas = []
d = None
try:
while True:
d = gettotalrefcount()
method(self, *args, **kwargs)
if hasattr(self, 'cleanup'):
self.cleanup()
if 'urlparse' in sys.modules:
sys.modules['urlparse'].clear_cache()
d = gettotalrefcount() - d
deltas.append(d)
if 2 <= len(deltas) <= 3 and deltas[-2:] == [0, 0]:
break
if 3 <= len(deltas) and deltas[-3:] == [0, 0, 0]:
break
if len(deltas) >= 6:
raise AssertionError('refcount increased by %r' % (deltas, ))
finally:
gc.collect()
gc.enable()
def test_cyclic_dependency_withKeepRef(self):
"""Create 2 objects with a cyclic dependency, so that they can
only be removed by the garbage collector, and then invoke the
garbage collector in a different thread.
"""
import gc
class CyclicChildObject(ObjectView):
def __init__(self, model):
super(CyclicChildObject, self).__init__(None)
self.setModel(model)
class CyclicObject(ObjectModel):
def __init__(self):
super(CyclicObject, self).__init__()
self._view = CyclicChildObject(self)
# turn off automatic garbage collection, to be able to trigger it
# at the 'right' time
gc.disable()
alive = lambda :sum(isinstance(o, CyclicObject) for o in gc.get_objects() )
#
# first proof that the wizard is only destructed by the garbage
# collector
#
cycle = CyclicObject()
self.assertTrue(alive())
del cycle
self.assertTrue(alive())
gc.collect()
self.assertFalse(alive())
def import_data(filepath):
conn = sqlite3.connect(filepath)
c = conn.cursor()
count = 0
gc.disable()
for d in c.execute('''
select
hyg.*,
trek.name as trek,
gal.x,
gal.y,
gal.z
from
tblhyg hyg
inner join tblgalactic gal on gal.starid = hyg.starid
left outer join tblstartrek trek on trek.starid = hyg.starid'''):
starid, hip, hd, hr, gliese, bayerflam, proper, ra, dec, dist, mag,\
absmag, spectrum, color, trek, x, y, z = d
hd = 'HD%d' % hd if hd else None
hip = 'HIP%d' % hip if hip else None
hr = 'HR%d' % hr if hr else None
name = proper or hd or hip or hr
pos = vec3(x, y, z) * _km_in_ly
obj = SpaceObject.create(names=(name,))
obj.coords = pos
count += 1
if count % 1000 == 0:
gc.collect()
print count
gc.enable()
conn.close()
def main():
gc.disable()
app = QtGui.QApplication(sys.argv)
dialog = MyDialog()
dialog.show()
retval = app.exec_()
sys.exit(retval)
def wrapper(*args, **kwargs): try: gc.disable() return fun(*args, **kwargs) finally: gc.enable() gc.collect()
def newfunc(*args,**kargs):
try:
gc.disable()
result = func( *args, **kargs)
finally:
gc.enable()
return result
#!/usr/bin/env python ###################################################### # # howdoi - instant coding answers via the command line # written by Benjamin Gleitzman ([email protected]) # inspired by Rich Jones ([email protected]) # ###################################################### import gc gc.disable() # disable right at the start, we don't need it import argparse import glob import os import random import re import requests import requests_cache import sys from . import __version__ from pygments import highlight from pygments.lexers import guess_lexer, get_lexer_by_name from pygments.formatters.terminal import TerminalFormatter from pygments.util import ClassNotFound from pyquery import PyQuery as pq from requests.exceptions import ConnectionError from requests.exceptions import SSLError
import clr
clr.AddReference('MinadNet')
from System import AppDomain
AppDomain.CurrentDomain.SetData("APP_CONFIG_FILE", "./app.config")
import time
import gc
from System import Console
from System.Net.Sockets import *
from System.Net import *
from MinadNet import *
from MinadNet.Pools import BufferPool
gc.disable()
MainBufferPool = BufferPool(1024)
class EchoServer(Listener):
def OnConnection(self, sock):
#print("New connection from %s"%sock.RemoteEndPoint.ToString())
newClient = EchoClient(sock, False)
newClient.setReadBuf()
newClient.setName("CLIENT-SERVER")
newClient.Receive(newClient.ReadBuf)
class EchoClient(Client):
def setName(self, name):
self.name = name
def setReadBuf(self):
def main(cmd_args):
import optparse
global options, PSYCO
usage = "\n%prog [options] command [input-file-patterns]\n" + cmd_doc
oparser = optparse.OptionParser(usage)
oparser.add_option(
"-l", "--logfilename",
default="",
help="contains error messages")
oparser.add_option(
"-v", "--verbosity",
type="int", default=0,
help="level of information and diagnostics provided")
oparser.add_option(
"-m", "--mmap",
type="int", default=-1,
help="1: use mmap; 0: don't use mmap; -1: accept heuristic")
oparser.add_option(
"-e", "--encoding",
default="",
help="encoding override")
oparser.add_option(
"-f", "--formatting",
type="int", default=0,
help="0 (default): no fmt info\n"
"1: fmt info (all cells)\n"
)
oparser.add_option(
"-g", "--gc",
type="int", default=0,
help="0: auto gc enabled; 1: auto gc disabled, manual collect after each file; 2: no gc")
oparser.add_option(
"-s", "--onesheet",
default="",
help="restrict output to this sheet (name or index)")
oparser.add_option(
"-u", "--unnumbered",
action="store_true", default=0,
help="omit line numbers or offsets in biff_dump")
oparser.add_option(
"-d", "--on-demand",
action="store_true", default=0,
help="load sheets on demand instead of all at once")
oparser.add_option(
"-t", "--suppress-timing",
action="store_true", default=0,
help="don't print timings (diffs are less messy)")
oparser.add_option(
"-r", "--ragged-rows",
action="store_true", default=0,
help="open_workbook(..., ragged_rows=True)")
options, args = oparser.parse_args(cmd_args)
if len(args) == 1 and args[0] in ("version", ):
pass
elif len(args) < 2:
oparser.error("Expected at least 2 args, found %d" % len(args))
cmd = args[0]
xlrd_version = getattr(xlrd, "__VERSION__", "unknown; before 0.5")
if cmd == 'biff_dump':
xlrd.dump(args[1], unnumbered=options.unnumbered)
sys.exit(0)
if cmd == 'biff_count':
xlrd.count_records(args[1])
sys.exit(0)
if cmd == 'version':
print("xlrd: %s, from %s" % (xlrd_version, xlrd.__file__))
print("Python:", sys.version)
sys.exit(0)
if options.logfilename:
logfile = LogHandler(open(options.logfilename, 'w'))
else:
logfile = sys.stdout
mmap_opt = options.mmap
mmap_arg = xlrd.USE_MMAP
if mmap_opt in (1, 0):
mmap_arg = mmap_opt
elif mmap_opt != -1:
print('Unexpected value (%r) for mmap option -- assuming default' % mmap_opt)
fmt_opt = options.formatting | (cmd in ('xfc', ))
gc_mode = options.gc
if gc_mode:
gc.disable()
for pattern in args[1:]:
for fname in glob.glob(pattern):
print("\n=== File: %s ===" % fname)
if logfile != sys.stdout:
logfile.setfileheading("\n=== File: %s ===\n" % fname)
if gc_mode == 1:
n_unreachable = gc.collect()
if n_unreachable:
print("GC before open:", n_unreachable, "unreachable objects")
if PSYCO:
import psyco
psyco.full()
PSYCO = 0
try:
t0 = time.time()
bk = xlrd.open_workbook(fname,
verbosity=options.verbosity, logfile=logfile,
use_mmap=mmap_arg,
encoding_override=options.encoding,
formatting_info=fmt_opt,
on_demand=options.on_demand,
ragged_rows=options.ragged_rows,
)
t1 = time.time()
if not options.suppress_timing:
print("Open took %.2f seconds" % (t1-t0,))
except xlrd.XLRDError as e:
print("*** Open failed: %s: %s" % (type(e).__name__, e))
continue
except KeyboardInterrupt:
print("*** KeyboardInterrupt ***")
traceback.print_exc(file=sys.stdout)
sys.exit(1)
except BaseException as e:
print("*** Open failed: %s: %s" % (type(e).__name__, e))
traceback.print_exc(file=sys.stdout)
continue
t0 = time.time()
if cmd == 'hdr':
bk_header(bk)
elif cmd == 'ov': # OverView
show(bk, 0)
elif cmd == 'show': # all rows
show(bk)
elif cmd == '2rows': # first row and last row
show(bk, 2)
elif cmd == '3rows': # first row, 2nd row and last row
show(bk, 3)
elif cmd == 'bench':
show(bk, printit=0)
elif cmd == 'fonts':
bk_header(bk)
show_fonts(bk)
elif cmd == 'names': # named reference list
show_names(bk)
elif cmd == 'name_dump': # named reference list
show_names(bk, dump=1)
elif cmd == 'labels':
show_labels(bk)
elif cmd == 'xfc':
count_xfs(bk)
else:
print("*** Unknown command <%s>" % cmd)
sys.exit(1)
del bk
if gc_mode == 1:
n_unreachable = gc.collect()
if n_unreachable:
print("GC post cmd:", fname, "->", n_unreachable, "unreachable objects")
if not options.suppress_timing:
t1 = time.time()
print("\ncommand took %.2f seconds\n" % (t1-t0,))
return None
def setUp(self):
self.using_gc = gc.isenabled()
gc.disable()
def _execute_child(self, args, executable, preexec_fn, cwd, env,
universal_newlines, shell, p2cread, p2cwrite, c2pread,
c2pwrite, errread, errwrite):
"""Execute program (POSIX version)"""
if isinstance(args, basestring):
args = [args]
else:
args = list(args)
if shell:
args = ["/bin/sh", "-c"] + args
if executable is None:
executable = args[0]
gc_was_enabled = gc.isenabled()
# Disable gc to avoid bug where gc -> file_dealloc ->
# write to stderr -> hang. http://bugs.python.org/issue1336
gc.disable()
try:
self.pid = os.fork()
except:
if gc_was_enabled:
gc.enable()
raise
self._child_created = True
if self.pid == 0:
# Child
try:
# Close parent's pipe ends
if p2cwrite is not None:
os.close(p2cwrite)
if c2pread is not None:
os.close(c2pread)
if errread is not None:
os.close(errread)
# Dup fds for child
if p2cread is not None:
os.dup2(p2cread, 0)
if c2pwrite is not None:
os.dup2(c2pwrite, 1)
if errwrite is not None:
os.dup2(errwrite, 2)
# Close pipe fds. Make sure we don't close the same
# fd more than once, or standard fds.
if p2cread is not None and p2cread not in (0, ):
os.close(p2cread)
if c2pwrite is not None and c2pwrite not in (p2cread, 1):
os.close(c2pwrite)
if errwrite is not None and errwrite not in \
(p2cread, c2pwrite, 2):
os.close(errwrite)
if cwd is not None:
os.chdir(cwd)
if preexec_fn:
preexec_fn()
if env is None:
os.execvp(executable, args)
else:
os.execvpe(executable, args, env)
except:
# Child execution failure
os._exit(255)
# Parent
if gc_was_enabled:
gc.enable()
if p2cread is not None and p2cwrite is not None:
os.close(p2cread)
if c2pwrite is not None and c2pread is not None:
os.close(c2pwrite)
if errwrite is not None and errread is not None:
os.close(errwrite)
def main(self):
rounds = self.values['-n']
reportfile = self.values['-f']
show_bench = self.values['-s']
compare_to = self.values['-c']
hidenoise = self.values['-d']
warp = int(self.values['-w'])
withgc = self.values['--with-gc']
limitnames = self.values['-t']
if limitnames:
if _debug:
print('* limiting test names to one with substring "%s"' % \
limitnames)
limitnames = re.compile(limitnames, re.I)
else:
limitnames = None
verbose = self.verbose
withsyscheck = self.values['--with-syscheck']
calibration_runs = self.values['-C']
timer = self.values['--timer']
print('-' * LINE)
print('PYBENCH %s' % __version__)
print('-' * LINE)
print('* using %s %s' % (
getattr(platform, 'python_implementation', lambda:'Python')(),
' '.join(sys.version.split())))
# Switch off garbage collection
if not withgc:
try:
import gc
except ImportError:
print('* Python version doesn\'t support garbage collection')
else:
try:
gc.disable()
except NotImplementedError:
print('* Python version doesn\'t support gc.disable')
else:
print('* disabled garbage collection')
# "Disable" sys check interval
if not withsyscheck:
# Too bad the check interval uses an int instead of a long...
value = 2147483647
try:
sys.setcheckinterval(value)
except (AttributeError, NotImplementedError):
print('* Python version doesn\'t support sys.setcheckinterval')
else:
print('* system check interval set to maximum: %s' % value)
if timer == TIMER_SYSTIMES_PROCESSTIME:
import systimes
print('* using timer: systimes.processtime (%s)' % \
systimes.SYSTIMES_IMPLEMENTATION)
else:
# Check that the clock function does exist
try:
get_timer(timer)
except TypeError:
print("* Error: Unknown timer: %s" % timer)
return
print('* using timer: %s' % timer)
if hasattr(time, 'get_clock_info'):
info = time.get_clock_info(timer[5:])
print('* timer: resolution=%s, implementation=%s'
% (info.resolution, info.implementation))
print()
if compare_to:
try:
f = open(compare_to,'rb')
bench = pickle.load(f)
bench.name = compare_to
f.close()
compare_to = bench
except IOError as reason:
print('* Error opening/reading file %s: %s' % (
repr(compare_to),
reason))
compare_to = None
if show_bench:
try:
f = open(show_bench,'rb')
bench = pickle.load(f)
bench.name = show_bench
f.close()
bench.print_header()
if compare_to:
bench.print_comparison(compare_to,
hidenoise=hidenoise,
limitnames=limitnames)
else:
bench.print_benchmark(hidenoise=hidenoise,
limitnames=limitnames)
except IOError as reason:
print('* Error opening/reading file %s: %s' % (
repr(show_bench),
reason))
print()
return
if reportfile:
print('Creating benchmark: %s (rounds=%i, warp=%i)' % \
(reportfile, rounds, warp))
print()
# Create benchmark object
bench = Benchmark(reportfile,
verbose=verbose,
timer=timer,
warp=warp,
calibration_runs=calibration_runs)
bench.rounds = rounds
bench.load_tests(Setup, limitnames=limitnames)
try:
bench.calibrate()
bench.run()
except KeyboardInterrupt:
print()
print('*** KeyboardInterrupt -- Aborting')
print()
return
bench.print_header()
if compare_to:
bench.print_comparison(compare_to,
hidenoise=hidenoise,
limitnames=limitnames)
else:
bench.print_benchmark(hidenoise=hidenoise,
limitnames=limitnames)
# Ring bell
sys.stderr.write('\007')
if reportfile:
try:
f = open(reportfile,'wb')
bench.name = reportfile
pickle.dump(bench,f)
f.close()
except IOError as reason:
print('* Error opening/writing reportfile')
except IOError as reason:
print('* Error opening/writing reportfile %s: %s' % (
reportfile,
reason))
print()
def config_rt_process(core, priority): gc.disable() set_realtime_priority(priority) set_core_affinity(core)
def main(argv):
global BLEU_THRESHOLD_REACHED, START_TIME
if len(argv) > 1:
raise app.UsageError('Too many command-line arguments.')
init_mllogger()
mllogger.event('cache_clear')
mllogger.start('init_start')
mllogger.event('submission_org', 'Google')
mllogger.event('submission_platform',
'TPUv3-{}'.format(jax.device_count()))
mllogger.event('submission_division', 'closed')
mllogger.event('submission_status', 'research')
mllogger.event('submission_benchmark', 'transformer')
mllogger.event('train_samples', input_pipeline.N_TRAIN)
mllogger.event('eval_samples', input_pipeline.N_EVAL)
tf.enable_v2_behavior()
# Use hardware RNG for bernoulli randoms in dropout mask creation.
if FLAGS.hardware_rng:
models.set_hardware_bernoulli()
num_partitions = FLAGS.num_partitions
batch_size = FLAGS.batch_size
if batch_size is None:
batch_size = min(16 * jax.device_count() // num_partitions, 2048)
mllogger.event('global_batch_size', batch_size)
num_eval_steps = FLAGS.num_eval_steps
max_target_length = FLAGS.max_target_length
max_eval_target_length = FLAGS.max_eval_target_length
max_length = max(max_target_length, max_eval_target_length)
mllogger.event('max_sequence_length',
max_length,
metadata={'method': 'discard'})
if FLAGS.random_seed is not None:
seed = FLAGS.random_seed
else:
seed = np.uint32(time.time() if jax.host_id() == 0 else 0)
seed = per_host_sum_pmap(seed)
mllogger.event('seed', int(seed))
steps_per_epoch = int(math.ceil(input_pipeline.N_TRAIN / batch_size))
logging.info('steps per epoch: %d', steps_per_epoch)
num_replicas = jax.local_device_count() // num_partitions
device_train_input_shape = (batch_size //
(num_replicas * jax.host_count()),
max_target_length)
# This is per-host; in principle 64/replica or more should fit
eval_batch_size = min(
32 * num_replicas,
int(
math.ceil(input_pipeline.N_EVAL /
(num_replicas * jax.host_count()))) * num_replicas)
logging.info('eval batch size: %d', eval_batch_size)
pred_batches = int(
math.ceil(input_pipeline.N_EVAL /
(jax.host_count() * eval_batch_size)))
logging.info('pred batches: %d', pred_batches)
broadcast = functools.partial(_broadcast,
num_replicas=num_replicas,
num_partitions=num_partitions)
if jax.host_id() == 0:
train_summary_writer = tensorboard.SummaryWriter(
os.path.join(FLAGS.model_dir, 'train'))
eval_summary_writer = tensorboard.SummaryWriter(
os.path.join(FLAGS.model_dir, 'eval'))
else:
train_summary_writer = None
eval_summary_writer = None
# Write summaries in background thread to avoid blocking on device sync
summary_thread = thread.ThreadPoolExecutor(1, 'summary')
if FLAGS.infeed:
# Infeed is currently synchronous, so do it in a background thread too
infeed_pool = thread.ThreadPoolExecutor(jax.local_device_count(),
'infeed')
# MLPerf 2020 WMT en-de dataset uses a custom T2T dataset:
# Shared 32K subword tokenization
# 256-length packed training examples from WMT17
# 97-length unpacked evaluation examples from WMT14
train_keys = [
'inputs', 'targets', 'inputs_position', 'targets_position',
'inputs_segmentation', 'targets_segmentation'
]
encoder = mlperf_encoder.SubwordTextEncoder(filename=FLAGS.vocab_path)
input_encoder = encoder
target_encoder = encoder
vocab_size = input_encoder.vocab_size
output_vocab_size = target_encoder.vocab_size
input_shape = (batch_size, max_target_length)
target_shape = (batch_size, max_target_length)
transformer_kwargs = flax.core.FrozenDict({
'vocab_size': vocab_size,
'output_vocab_size': output_vocab_size,
'emb_dim': 1024,
'num_heads': 16,
'num_layers': 6,
'qkv_dim': 1024,
'mlp_dim': 4096,
'max_len': max_length,
'share_embeddings': FLAGS.share_embeddings,
'logits_via_embedding': FLAGS.logits_via_embedding,
'num_partitions': num_partitions,
})
rng = random.PRNGKey(seed)
rng, init_rng = random.split(rng)
logging.info('initializing model')
model, cache_def = create_model(init_rng, tuple(input_shape),
tuple(target_shape), transformer_kwargs)
mllogger.event('opt_name', 'adam')
if batch_size < 1024:
learning_rate = 4.0 # 0.0625
warmup_steps = 1000
beta1 = 0.9
beta2 = 0.98
if batch_size < 2048:
learning_rate = 2.0
warmup_steps = 500 # ??
beta1 = 0.9 # ??
beta2 = 0.98 # ??
else:
learning_rate = 3.3092157691415953
warmup_steps = 664
beta1 = 0.9086575725261137
beta2 = 0.9198719118104947
epsilon = 1e-9
if FLAGS.learning_rate is not None:
learning_rate = FLAGS.learning_rate
mllogger.event('opt_adam_beta_1', beta1)
mllogger.event('opt_adam_beta_2', beta2)
mllogger.event('opt_adam_epsilon', epsilon)
logging.info('initializing optimizer')
optimizer_def = optim.Adam(learning_rate,
beta1=beta1,
beta2=beta2,
eps=epsilon,
weight_decay=FLAGS.weight_decay)
optimizer = optimizer_def.create(model)
del model # don't keep a copy of the initial model
# Build parameter partition annotations for preserving partitions from train
# to eval.
partition_rules = [
(('encoder', 'posembed_input'), partitions.empty_dict),
(('decoder', 'posembed_targets'), partitions.empty_dict),
(('embedding', ), partitions.spec(num_partitions, 1)),
((r'LayerNorm_\d+', '(bias|scale)'), None),
((r'encoder(decoder)?_norm', '(bias|scale)'), None),
((r'MultiHeadDotProductAttention_\d+', '(query|key|value)', 'kernel'),
partitions.spec(1, num_partitions, 1)),
((r'MultiHeadDotProductAttention_\d+', 'out', 'kernel'),
partitions.spec(num_partitions, 1, 1)),
((r'MlpBlock_\d+', r'Dense_\d+', 'bias'), None),
((r'MlpBlock_\d+', 'Dense_0', 'kernel'),
partitions.spec(1, num_partitions)),
((r'MlpBlock_\d+', 'Dense_1', 'kernel'),
partitions.spec(num_partitions, 1)),
(('state', 'step'), None),
]
optimizer_partitions = optimizer.restore_state(
partitions.set_partitions(partition_rules, optimizer.state_dict()))
optimizer = broadcast(optimizer)
empty_metrics = broadcast({'loss': 0.0, 'accuracy': 0, 'denominator': 0})
learning_rate_fn = create_learning_rate_scheduler(
base_learning_rate=learning_rate,
warmup_steps=warmup_steps,
hidden_size=transformer_kwargs['qkv_dim'])
p_train_step = jax.pmap(functools.partial(
train_step, learning_rate_fn=learning_rate_fn),
axis_name='batch',
in_axes=(None, 0, 0, 0))
if num_partitions > 1:
sharded_predict_step = sharded_jit(
predict_step,
in_parts=(None, optimizer_partitions.target, None),
out_parts=None)
else:
sharded_predict_step = predict_step
if FLAGS.extra_eval_metrics:
p_eval_step = jax.pmap(eval_step, axis_name='batch', in_axes=(None, 0))
p_pred_step = jax.pmap(sharded_predict_step,
axis_name='batch',
in_axes=(0, None, None))
p_allreduce_metrics = jax.pmap(functools.partial(lax.psum,
axis_name='batch'),
axis_name='batch')
def device_train_loop_cond(args):
_, _, _, _, step, epoch = args
return step // steps_per_epoch == epoch
def device_train_loop_body(args):
optimizer, dropout_rngs, metrics, token, step, epoch = args
input_data, token = lax.infeed(token,
shape=tuple([
jax.ShapedArray(
device_train_input_shape,
jnp.int32) for _ in train_keys
]))
batch = {k: v for k, v in zip(train_keys, input_data)}
optimizer, metrics, dropout_rngs = train_step(optimizer,
batch,
metrics,
learning_rate_fn,
dropout_rng=dropout_rngs)
step += 1
return optimizer, dropout_rngs, metrics, token, step, epoch
def device_train_loop(optimizer, dropout_rngs, metrics, step, epoch):
token = lax.create_token(step)
optimizer, dropout_rngs, metrics, _, step, _ = lax.while_loop(
device_train_loop_cond, device_train_loop_body,
(optimizer, dropout_rngs, metrics, token, step, epoch))
return optimizer, dropout_rngs, metrics, step
if num_partitions > 1:
device_train_loop = sharded_jit(device_train_loop,
in_parts=(optimizer_partitions, None,
None, None, None),
out_parts=(optimizer_partitions, None,
None, None))
p_train_epoch = jax.pmap(device_train_loop,
axis_name='batch',
in_axes=(None, 0, 0, None, None))
p_allreduce_metrics_train = functools.partial(lax.psum, axis_name='batch')
if num_partitions > 1:
p_allreduce_metrics_train = sharded_jit(p_allreduce_metrics_train,
in_parts=None,
out_parts=None,
num_partitions=num_partitions)
p_allreduce_metrics_train = jax.pmap(p_allreduce_metrics_train,
axis_name='batch')
# Precompile all needed computations with fake data so as not to include
# compilation time in MLPerf metrics.
if FLAGS.precompile:
logging.info('precompiling step/epoch functions')
if FLAGS.infeed:
# the device training loop condition will immediately be false, but
# the optimizer tree will be resharded here
optimizer, *_ = p_train_epoch(unbroadcast(optimizer),
random.split(rng, num_replicas),
empty_metrics,
jnp.array(0, dtype=jnp.int32), 1)
else:
metrics = empty_metrics
train_input_shape = (num_replicas, batch_size // num_replicas,
input_pipeline.MAX_TRAIN_LEN)
fake_batch = {
k: jnp.ones(train_input_shape, jnp.int32)
for k in train_keys
}
p_train_step(unbroadcast(optimizer),
fake_batch,
metrics,
dropout_rng=random.split(rng, num_replicas))
eval_input_shape = (num_replicas, eval_batch_size // num_replicas,
input_pipeline.MAX_EVAL_LEN)
fake_eval_batch = {
'inputs': jnp.ones(eval_input_shape, jnp.int32),
'targets': jnp.ones(eval_input_shape, jnp.int32),
}
if FLAGS.extra_eval_metrics:
p_eval_step(unbroadcast(optimizer.target), fake_eval_batch)
fake_cache = cache_def.initialize_cache(
(eval_input_shape[1], FLAGS.max_predict_length))
p_pred_step(fake_eval_batch['inputs'], unbroadcast(optimizer.target),
fake_cache)
time.sleep(20)
sync_devices()
fake_bleu_1 = np.zeros((4, ), dtype=np.int32)
fake_bleu_2 = np.zeros((), dtype=np.int32)
per_host_sum_pmap((fake_bleu_1, fake_bleu_1, fake_bleu_2, fake_bleu_2))
sync_devices()
p_allreduce_metrics_train(empty_metrics)
sync_devices()
logging.info('finished precompiling step/epoch functions')
# We init the first set of dropout PRNG keys, but update it afterwards inside
# the main pmap'd training update for performance.
dropout_rngs = random.split(rng, num_replicas)
# Record time-0 metrics for proper tensorboard plot x-axis scaling.
if jax.host_id() == 0:
if FLAGS.compute_train_metrics:
train_summary_writer.scalar('loss', 9.999, 0)
train_summary_writer.scalar('accuracy', 0.0, 0)
train_summary_writer.flush()
eval_summary_writer.scalar('bleu', 0.0, 0)
eval_summary_writer.flush()
train_ds = input_pipeline.get_wmt_dataset(batch_size=batch_size //
jax.host_count(),
train=True)
eval_ds = input_pipeline.get_wmt_dataset(batch_size=eval_batch_size,
train=False)
train_iter = iter(train_ds)
eval_iter = iter(eval_ds)
local_devices = jax.local_devices()
host_step, device_step = 0, broadcast(0)
gc.disable()
mllogger.end('init_stop')
if jax.host_id() == 0:
mllogger.start('run_start')
START_TIME = time.time()
for epoch in range(FLAGS.num_epochs):
if jax.host_id() == 0 and not BLEU_THRESHOLD_REACHED:
mllogger.start('block_start',
metadata={
'first_epoch_num': epoch + 1,
'epoch_count': 1
})
metrics = empty_metrics
if FLAGS.infeed:
optimizer, dropout_rngs, metrics, device_step = p_train_epoch(
unbroadcast(optimizer), dropout_rngs, metrics,
unbroadcast(device_step), epoch)
while int(host_step // steps_per_epoch) == epoch:
# pylint: disable=protected-access
batch = jax.tree_map(lambda x: x._numpy(), next(train_iter))
# Shard data to devices and do a training step.
batch = jax.tree_map(
lambda x: x.reshape((num_replicas, -1) + x.shape[1:]), batch)
if FLAGS.infeed:
for i, device in enumerate(local_devices):
replica_id = i // num_partitions
input_tuple = tuple(
[batch[k][replica_id] for k in train_keys])
assert input_tuple[0].shape == device_train_input_shape, (
'infeed shape error %s != %s' %
(input_tuple[0].shape, device_train_input_shape))
assert input_tuple[0].dtype == jnp.int32, (
'infeed dtype error %s != %s' %
(input_tuple[0].dtype, jnp.int32))
infeed_pool.submit(
functools.partial(device.transfer_to_infeed,
input_tuple))
else:
optimizer, metrics, dropout_rngs = p_train_step(
unbroadcast(optimizer),
batch,
metrics,
dropout_rng=dropout_rngs)
host_step += 1
if FLAGS.compute_train_metrics:
metrics = p_allreduce_metrics_train(metrics)
# Schedule training metric handling.
summary_thread.submit(
functools.partial(write_train_summary, metrics,
train_summary_writer, host_step))
# Optional, extra evaluation metrics.
if FLAGS.extra_eval_metrics:
eval_metrics = []
eval_iter = iter(eval_ds)
for _, eval_batch in zip(range(num_eval_steps), eval_iter):
eval_batch = common_utils.shard(eval_batch)
metrics = p_eval_step(unbroadcast(optimizer.target),
eval_batch)
eval_metrics.append(metrics)
eval_metrics = p_allreduce_metrics(eval_metrics)
# Schedule metric summarization/logging.
summary_thread.submit(
functools.partial(write_eval_summary, eval_metrics,
eval_summary_writer, host_step))
# Translation and BLEU Score.
all_predicted, all_targets, all_bs = [], [], []
for i in range(pred_batches):
# pylint: disable=protected-access
pred_batch = jax.tree_map(lambda x: x._numpy(), next(eval_iter))
# Handle final odd-sized batch by padding instead of dropping it.
cur_pred_batch_size = pred_batch['inputs'].shape[0]
if cur_pred_batch_size != eval_batch_size:
pred_batch = jax.tree_map(
lambda x: pad_examples(x, eval_batch_size), pred_batch)
pred_batch = jax.tree_map(
lambda x: x.reshape((num_replicas, -1) + x.shape[1:]),
pred_batch)
per_device_batchsize = pred_batch['inputs'].shape[1]
cache = cache_def.initialize_cache(
(per_device_batchsize, FLAGS.max_predict_length))
all_predicted.append(
p_pred_step(pred_batch['inputs'],
unbroadcast(optimizer.target), cache))
all_targets.append(pred_batch['targets'])
all_bs.append(cur_pred_batch_size)
# Schedule BLEU calculation and summarization/logging.
# We use the ICI as part of BLEU score computation, so we call this from the
# main thread so the BLEU pmap runs before the next train epoch pmap
write_predict_summary(all_predicted, all_targets, all_bs,
target_encoder, eval_summary_writer, epoch,
host_step, summary_thread)
# Wait until computations are done before exiting
sync_devices()
if jax.host_id() == 0:
summary_thread.shutdown()
if not BLEU_THRESHOLD_REACHED:
mllogger.end('run_stop', metadata={'status': 'aborted'})
def map2dict(a_file,
column,
limit_counts_reads=7 * (10**7),
size_buffer=10**8):
# get read name and mismatches
fi = open(a_file, 'r')
pack = None
last_read = None
base = []
while True:
gc.disable()
du = fi.readlines(size_buffer)
gc.enable()
if not du:
break
gc.disable()
du = [giveRC(d.rstrip('\r\n').split('\t'), column) for d in du]
gc.enable()
di = []
for d in du:
if last_read != d[0]:
gc.disable()
di.append((d[0], 0 if not d[1] else d[1].count(':')))
gc.enable()
last_read = d[0]
if di:
gc.disable()
base.extend(di)
gc.enable()
di = []
if len(base) > limit_counts_reads:
gc.disable()
base = dict(base)
gc.enable()
yield base
base = []
if base:
gc.disable()
base = dict(base)
gc.enable()
yield base
base = []
fi.close()
def __enter__(self):
self.gcold = gc.isenabled()
if self.disable_gc:
gc.disable()
self.start_time = timeit.default_timer()
return self
def disableGarbageCollector():
gc.disable()
print('Disabled Garbage Collector.')
def _fork(self, path, uid, gid, executable, args, environment, **kwargs):
"""
Fork and then exec sub-process.
@param path: the path where to run the new process.
@type path: C{str}
@param uid: if defined, the uid used to run the new process.
@type uid: C{int}
@param gid: if defined, the gid used to run the new process.
@type gid: C{int}
@param executable: the executable to run in a new process.
@type executable: C{str}
@param args: arguments used to create the new process.
@type args: C{list}.
@param environment: environment used for the new process.
@type environment: C{dict}.
@param kwargs: keyword arguments to L{_setupChild} method.
"""
settingUID = (uid is not None) or (gid is not None)
if settingUID:
curegid = os.getegid()
currgid = os.getgid()
cureuid = os.geteuid()
curruid = os.getuid()
if uid is None:
uid = cureuid
if gid is None:
gid = curegid
# prepare to change UID in subprocess
os.setuid(0)
os.setgid(0)
collectorEnabled = gc.isenabled()
gc.disable()
try:
self.pid = os.fork()
except:
# Still in the parent process
if settingUID:
os.setregid(currgid, curegid)
os.setreuid(curruid, cureuid)
if collectorEnabled:
gc.enable()
raise
else:
if self.pid == 0: # pid is 0 in the child process
# do not put *ANY* code outside the try block. The child process
# must either exec or _exit. If it gets outside this block (due
# to an exception that is not handled here, but which might be
# handled higher up), there will be two copies of the parent
# running in parallel, doing all kinds of damage.
# After each change to this code, review it to make sure there
# are no exit paths.
try:
# Stop debugging. If I am, I don't care anymore.
sys.settrace(None)
self._setupChild(**kwargs)
self._execChild(path, settingUID, uid, gid, executable,
args, environment)
except:
# If there are errors, bail and try to write something
# descriptive to stderr.
# XXX: The parent's stderr isn't necessarily fd 2 anymore, or
# even still available
# XXXX: however even libc assumes write(2, err) is a useful
# thing to attempt
try:
stderr = os.fdopen(2, 'w')
stderr.write(
"Upon execvpe %s %s in environment %s\n:" %
(executable, str(args), "id %s" % id(environment)))
traceback.print_exc(file=stderr)
stderr.flush()
for fd in range(3):
os.close(fd)
except:
pass # make *sure* the child terminates
# Did you read the comment about not adding code here?
os._exit(1)
# we are now in parent process
if settingUID:
os.setregid(currgid, curegid)
os.setreuid(curruid, cureuid)
if collectorEnabled:
gc.enable()
self.status = -1 # this records the exit status of the child
def opt_disablegc(self):
"""Disable the garbage collector"""
gc.disable()
def disableGC(self) -> None:
gc.collect()
gc.disable()
def main(sm=None, pm=None):
gc.disable()
set_realtime_priority(5)
if sm is None:
sm = messaging.SubMaster(['liveLocationKalman', 'carState'],
poll=['liveLocationKalman'])
if pm is None:
pm = messaging.PubMaster(['liveParameters'])
params_reader = Params()
# wait for stats about the car to come in from controls
cloudlog.info("paramsd is waiting for CarParams")
CP = car.CarParams.from_bytes(params_reader.get("CarParams", block=True))
cloudlog.info("paramsd got CarParams")
min_sr, max_sr = 0.5 * CP.steerRatio, 2.0 * CP.steerRatio
params = params_reader.get("LiveParameters") if not params_reader.get_bool(
'dp_reset_live_param_on_start') else None
# Check if car model matches
if params is not None:
params = json.loads(params)
if params.get('carFingerprint', None) != CP.carFingerprint:
cloudlog.info("Parameter learner found parameters for wrong car.")
params = None
# Check if starting values are sane
if params is not None:
try:
angle_offset_sane = abs(params.get('angleOffsetAverageDeg')) < 10.0
steer_ratio_sane = min_sr <= params['steerRatio'] <= max_sr
params_sane = angle_offset_sane and steer_ratio_sane
if not params_sane:
cloudlog.info(f"Invalid starting values found {params}")
params = None
except Exception as e:
cloudlog.info(f"Error reading params {params}: {str(e)}")
params = None
# TODO: cache the params with the capnp struct
if params is None:
params = {
'carFingerprint': CP.carFingerprint,
'steerRatio': CP.steerRatio,
'stiffnessFactor': 1.0,
'angleOffsetAverageDeg': 0.0,
}
cloudlog.info("Parameter learner resetting to default values")
# When driving in wet conditions the stiffness can go down, and then be too low on the next drive
# Without a way to detect this we have to reset the stiffness every drive
params['stiffnessFactor'] = 1.0
learner = ParamsLearner(CP, params['steerRatio'],
params['stiffnessFactor'],
math.radians(params['angleOffsetAverageDeg']))
angle_offset_average = params['angleOffsetAverageDeg']
angle_offset = angle_offset_average
while True:
sm.update()
for which in sorted(sm.updated.keys(),
key=lambda x: sm.logMonoTime[x]):
if sm.updated[which]:
t = sm.logMonoTime[which] * 1e-9
learner.handle_log(t, which, sm[which])
if sm.updated['liveLocationKalman']:
x = learner.kf.x
P = np.sqrt(learner.kf.P.diagonal())
if not all(map(math.isfinite, x)):
cloudlog.error(
"NaN in liveParameters estimate. Resetting to default values"
)
learner = ParamsLearner(CP, CP.steerRatio, 1.0, 0.0)
x = learner.kf.x
angle_offset_average = clip(
math.degrees(x[States.ANGLE_OFFSET]),
angle_offset_average - MAX_ANGLE_OFFSET_DELTA,
angle_offset_average + MAX_ANGLE_OFFSET_DELTA)
angle_offset = clip(
math.degrees(x[States.ANGLE_OFFSET] +
x[States.ANGLE_OFFSET_FAST]),
angle_offset - MAX_ANGLE_OFFSET_DELTA,
angle_offset + MAX_ANGLE_OFFSET_DELTA)
msg = messaging.new_message('liveParameters')
msg.logMonoTime = sm.logMonoTime['carState']
msg.liveParameters.posenetValid = True
msg.liveParameters.sensorValid = True
msg.liveParameters.steerRatio = float(x[States.STEER_RATIO])
msg.liveParameters.stiffnessFactor = float(x[States.STIFFNESS])
msg.liveParameters.angleOffsetAverageDeg = angle_offset_average
msg.liveParameters.angleOffsetDeg = angle_offset
msg.liveParameters.valid = all((
abs(msg.liveParameters.angleOffsetAverageDeg) < 10.0,
abs(msg.liveParameters.angleOffsetDeg) < 10.0,
0.2 <= msg.liveParameters.stiffnessFactor <= 5.0,
min_sr <= msg.liveParameters.steerRatio <= max_sr,
))
msg.liveParameters.steerRatioStd = float(P[States.STEER_RATIO])
msg.liveParameters.stiffnessFactorStd = float(P[States.STIFFNESS])
msg.liveParameters.angleOffsetAverageStd = float(
P[States.ANGLE_OFFSET])
msg.liveParameters.angleOffsetFastStd = float(
P[States.ANGLE_OFFSET_FAST])
if sm.frame % 1200 == 0: # once a minute
params = {
'carFingerprint':
CP.carFingerprint,
'steerRatio':
msg.liveParameters.steerRatio,
'stiffnessFactor':
msg.liveParameters.stiffnessFactor,
'angleOffsetAverageDeg':
msg.liveParameters.angleOffsetAverageDeg,
}
put_nonblocking("LiveParameters", json.dumps(params))
pm.send('liveParameters', msg)
def _teardown():
gc.disable() # do not interfere with timing tests
yield None
gc.enable()
gc.collect()
def __init__(self, fname=None, fdata=None, decompress=False,
disable_gc=True, slow_parsing=True):
# Runs a lot faster with GC off.
disable_gc = disable_gc and gc.isenabled()
try:
if disable_gc:
gc.disable()
if fname is not None:
assert fdata is None
# Allow reading preexisting streams like pyPdf
if hasattr(fname, 'read'):
fdata = fname.read()
else:
try:
f = open(fname, 'rb')
fdata = f.read()
f.close()
except IOError:
raise PdfParseError('Could not read PDF file %s' %
fname)
assert fdata is not None
if not fdata.startswith('%PDF-'):
startloc = fdata.find('%PDF-')
if startloc >= 0:
log.warning('PDF header not at beginning of file')
else:
lines = fdata.lstrip().splitlines()
if not lines:
raise PdfParseError('Empty PDF file!')
raise PdfParseError('Invalid PDF header: %s' %
repr(lines[0]))
endloc = fdata.rfind('%EOF')
if endloc < 0:
#log.error('EOF mark not found: %s' %
# repr(fdata[-20:]))
endloc = len(fdata) - 6
endloc += 6
junk = fdata[endloc:]
# Done: It is not necessary to truncate the string.
# Some PDFs just use wrong EOF at the end to confuse parsers.
#fdata = fdata[:endloc]
if junk.rstrip('\00').strip():
log.warning('Extra data at end of file')
private = self.private
private.indirect_objects = {}
private.deferred_objects = set()
private.special = {'<<': self.readdict,
'[': self.readarray,
'endobj': self.empty_obj,
}
for tok in r'\ ( ) < > { } ] >> %'.split():
self.special[tok] = self.badtoken
if slow_parsing == True:
startloc = 0
source = PdfTokens(fdata, startloc, True)
private.source = source
# Calling next() just for complete the structure of source by adding source.current.
source.next()
source.all_offsets = []
source.obj_offsets = {}
self.slow_parse_xref(source)
# Done: add slow parsing for multiple trailers.
trailer_loc = fdata.find('trailer')
newdict = None
while trailer_loc >= 0:
source.floc = trailer_loc
assert source.next() == "trailer" # trailer
tok = source.next() # <<
if tok != '<<':
source.exception('Expected "<<" starting catalog')
# Ignored the corrupted trailer.
try:
tmpdict = self.readdict(source)
except:
pass
else:
if not newdict:
newdict = tmpdict
else:
newdict.update(tmpdict)
finally:
trailer_loc = fdata.find('trailer', trailer_loc+1)
if newdict is not None:
newdict.Prev = None
else:
source.exception("No trailer.")
else:
startloc, source = self.findxref(fdata)
private.source = source
xref_table_list = []
source.all_offsets = []
while 1:
source.obj_offsets = {}
# Loop through all the cross-reference tables
self.parsexref(source)
tok = source.next()
if tok != '<<':
source.exception('Expected "<<" starting catalog')
newdict = self.readdict(source)
token = source.next()
if token != 'startxref' and not xref_table_list:
source.warning('Expected "startxref" at end of xref table')
# Loop if any previously-written tables.
prev = newdict.Prev
if prev is None:
break
if not xref_table_list:
newdict.Prev = None
original_indirect = self.indirect_objects.copy()
original_newdict = newdict
source.floc = int(prev)
xref_table_list.append(source.obj_offsets)
self.indirect_objects.clear()
if xref_table_list:
for update in reversed(xref_table_list):
source.obj_offsets.update(update)
self.indirect_objects.clear()
self.indirect_objects.update(original_indirect)
newdict = original_newdict
self.update(newdict)
# self.read_all_indirect(source)
private.pages = self.readpages(self.Root)
if decompress:
self.uncompress()
# For compatibility with pyPdf
private.numPages = len(self.pages)
finally:
if disable_gc:
gc.enable()
# load the trace
fname_trace = fname + '.trace'
if os.path.isfile(fname_trace):
f = open(fname_trace, 'rb')
private.active_trace = pickle.load(f)
f.close()
# WARNING - this application disables python garbage collection, to slightly improve performance # comment this out if you see randon crashes import RPi.GPIO as GPIO # to allow control of GPIO import time # for delay timing from time import sleep # for delay timing import datetime # for real time from datetime import datetime # for real time numbers import glob # needed to support automatic setup of DS18B20 # try commenting these 2 lines out if you see random crashes import gc # for managing garbage collection gc.disable() # disable garabage collect, minimises some random dislay blips # on slower Pi's #define GPIO pin usage - BCM numbering assumed throughout GPIO.setmode(GPIO.BCM) #turn off warning GPIO.setwarnings(False) #segments segA = 5 segB = 6 segC = 13 segD = 19
def controlsd_thread(gctx=None):
gc.disable()
# start the loop
set_realtime_priority(3)
params = Params()
# Pub Sockets
sendcan = messaging.pub_sock(service_list['sendcan'].port)
controlsstate = messaging.pub_sock(service_list['controlsState'].port)
carstate = messaging.pub_sock(service_list['carState'].port)
carcontrol = messaging.pub_sock(service_list['carControl'].port)
carevents = messaging.pub_sock(service_list['carEvents'].port)
carparams = messaging.pub_sock(service_list['carParams'].port)
is_metric = params.get("IsMetric") == "1"
passive = params.get("Passive") != "0"
sm = messaging.SubMaster([
'thermal', 'health', 'liveCalibration', 'driverMonitoring', 'plan',
'pathPlan'
])
logcan = messaging.sub_sock(service_list['can'].port)
# wait for health and CAN packets
hw_type = messaging.recv_one(sm.sock['health']).health.hwType
is_panda_black = hw_type == log.HealthData.HwType.blackPanda
wait_for_can(logcan)
CI, CP = get_car(logcan, sendcan, is_panda_black)
logcan.close()
# TODO: Use the logcan socket from above, but that will currenly break the tests
can_timeout = None if os.environ.get('NO_CAN_TIMEOUT', False) else 100
can_sock = messaging.sub_sock(service_list['can'].port,
timeout=can_timeout)
car_recognized = CP.carName != 'mock'
# If stock camera is disconnected, we loaded car controls and it's not chffrplus
controller_available = CP.enableCamera and CI.CC is not None and not passive
read_only = not car_recognized or not controller_available
if read_only:
CP.safetyModel = car.CarParams.SafetyModel.elm327 # diagnostic only
# Write CarParams for radard and boardd safety mode
params.put("CarParams", CP.to_bytes())
params.put("LongitudinalControl",
"1" if CP.openpilotLongitudinalControl else "0")
CC = car.CarControl.new_message()
AM = AlertManager()
startup_alert = get_startup_alert(car_recognized, controller_available)
AM.add(sm.frame, startup_alert, False)
LoC = LongControl(CP, CI.compute_gb)
VM = VehicleModel(CP)
if CP.lateralTuning.which() == 'pid':
LaC = LatControlPID(CP)
elif CP.lateralTuning.which() == 'indi':
LaC = LatControlINDI(CP)
elif CP.lateralTuning.which() == 'lqr':
LaC = LatControlLQR(CP)
driver_status = DriverStatus()
state = State.disabled
soft_disable_timer = 0
v_cruise_kph = 255
v_cruise_kph_last = 0
overtemp = False
free_space = False
cal_status = Calibration.INVALID
cal_perc = 0
mismatch_counter = 0
low_battery = False
events_prev = []
sm['pathPlan'].sensorValid = True
sm['pathPlan'].posenetValid = True
# detect sound card presence
sounds_available = not os.path.isfile('/EON') or (
os.path.isdir('/proc/asound/card0')
and open('/proc/asound/card0/state').read().strip() == 'ONLINE')
# controlsd is driven by can recv, expected at 100Hz
rk = Ratekeeper(100, print_delay_threshold=None)
prof = Profiler(False) # off by default
while True:
start_time = sec_since_boot()
prof.checkpoint("Ratekeeper", ignore=True)
# Sample data and compute car events
CS, events, cal_status, cal_perc, overtemp, free_space, low_battery, mismatch_counter =\
data_sample(CI, CC, sm, can_sock, cal_status, cal_perc, overtemp, free_space, low_battery,
driver_status, state, mismatch_counter, params)
prof.checkpoint("Sample")
# Create alerts
if not sm.all_alive_and_valid():
events.append(
create_event('commIssue', [ET.NO_ENTRY, ET.SOFT_DISABLE]))
if not sm['pathPlan'].mpcSolutionValid:
events.append(
create_event('plannerError',
[ET.NO_ENTRY, ET.IMMEDIATE_DISABLE]))
if not sm['pathPlan'].sensorValid:
events.append(
create_event('sensorDataInvalid', [ET.NO_ENTRY, ET.PERMANENT]))
if not sm['pathPlan'].paramsValid:
events.append(create_event('vehicleModelInvalid', [ET.WARNING]))
if not sm['pathPlan'].posenetValid:
events.append(
create_event('posenetInvalid', [ET.NO_ENTRY, ET.SOFT_DISABLE]))
if not sm['plan'].radarValid:
events.append(
create_event('radarFault', [ET.NO_ENTRY, ET.SOFT_DISABLE]))
if sm['plan'].radarCanError:
events.append(
create_event('radarCanError', [ET.NO_ENTRY, ET.SOFT_DISABLE]))
if not CS.canValid:
events.append(
create_event('canError', [ET.NO_ENTRY, ET.IMMEDIATE_DISABLE]))
if not sounds_available:
events.append(
create_event('soundsUnavailable', [ET.NO_ENTRY, ET.PERMANENT]))
# Only allow engagement with brake pressed when stopped behind another stopped car
if CS.brakePressed and sm[
'plan'].vTargetFuture >= STARTING_TARGET_SPEED and not CP.radarOffCan and CS.vEgo < 0.3:
events.append(
create_event('noTarget', [ET.NO_ENTRY, ET.IMMEDIATE_DISABLE]))
if not read_only:
# update control state
state, soft_disable_timer, v_cruise_kph, v_cruise_kph_last = \
state_transition(sm.frame, CS, CP, state, events, soft_disable_timer, v_cruise_kph, AM)
prof.checkpoint("State transition")
# Compute actuators (runs PID loops and lateral MPC)
actuators, v_cruise_kph, driver_status, v_acc, a_acc, lac_log = \
state_control(sm.frame, sm.rcv_frame, sm['plan'], sm['pathPlan'], CS, CP, state, events, v_cruise_kph, v_cruise_kph_last, AM, rk,
driver_status, LaC, LoC, VM, read_only, is_metric, cal_perc)
prof.checkpoint("State Control")
# Publish data
CC, events_prev = data_send(sm, CS, CI, CP, VM, state, events,
actuators, v_cruise_kph, rk, carstate,
carcontrol, carevents, carparams,
controlsstate, sendcan, AM, driver_status,
LaC, LoC, read_only, start_time, v_acc,
a_acc, lac_log, events_prev)
prof.checkpoint("Sent")
rk.monitor_time()
prof.display()
def main(args):
global running_loss, start_t
# logging info that needs to persist across iterations
viz = utils.visualize.VisdomVisualize(env_name=args.env_name)
viz.viz.text(str(args))
running_loss = None
running_accs = {
qtype: 0.5
for qtype, _ in enumerate(utils.QTYPE_ID_TO_META)
}
start_t = None
# data
dataset = FigQADataset(args.figqa_dir, args.figqa_pre, split='train1')
dataloader = DataLoader(dataset,
batch_size=args.batch_size,
num_workers=args.workers,
pin_memory=True,
shuffle=bool(args.shuffle_train))
val_dataset = FigQADataset(args.figqa_dir,
args.figqa_pre,
split=args.val_split)
val_dataloader = DataLoader(val_dataset,
batch_size=args.batch_size,
num_workers=args.workers,
pin_memory=True,
shuffle=True)
# model
if args.start_from:
model = utils.load_model(fname=args.start_from, ngpus=args.cuda)
else:
model_args = figqa.options.model_args(args)
model = utils.load_model(model_args, ngpus=args.cuda)
# optimization
optimizer = torch.optim.Adam(model.parameters(),
lr=args.lr,
weight_decay=args.weight_decay)
def exp_lr(epoch):
iters = epoch * len(dataloader)
return args.lr_decay**iters
scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer, exp_lr)
criterion = nn.NLLLoss()
# training
for epoch in range(args.epochs):
checkpoint_stuff(**locals())
scheduler.step()
start_t = timer()
# TODO: understand when/why automatic garbage collection slows down
# the train loop
gc.disable()
data_time = time.time()
start = torch.cuda.Event(enable_timing=True)
end = torch.cuda.Event(enable_timing=True)
end.record()
for local_iter_idx, batch in batch_iter(dataloader, args):
start.record()
torch.cuda.synchronize()
iter_idx = local_iter_idx + epoch * len(dataloader)
# forward + update
optimizer.zero_grad()
pred = model(batch)
loss = criterion(pred, batch['answer'])
loss.backward()
optimizer.step()
#Chargrid: Gradient Clipping
#torch.nn.utils.clip_grad_value_([model.module.decision_weight],0.1)
# visualize, log, checkpoint
#print("visualize")
log_stuff(**locals())
#for idx,timing in enumerate(["FW","Loss","Backw","Opt","log"]):
#print(timing,times[idx+1]-times[idx])
#data_time = time.time()
end.record()
torch.cuda.synchronize()
#print("Training: ",start.elapsed_time(end))
gc.enable()
def controlsd_thread(gctx=None, rate=100, default_bias=0.):
gc.disable()
# start the loop
set_realtime_priority(3)
context = zmq.Context()
params = Params()
# Pub Sockets
live100 = messaging.pub_sock(context, service_list['live100'].port)
carstate = messaging.pub_sock(context, service_list['carState'].port)
carcontrol = messaging.pub_sock(context, service_list['carControl'].port)
livempc = messaging.pub_sock(context, service_list['liveMpc'].port)
is_metric = params.get("IsMetric") == "1"
passive = params.get("Passive") != "0"
# No sendcan if passive
if not passive:
sendcan = messaging.pub_sock(context, service_list['sendcan'].port)
else:
sendcan = None
# Sub sockets
poller = zmq.Poller()
thermal = messaging.sub_sock(context, service_list['thermal'].port, conflate=True, poller=poller)
health = messaging.sub_sock(context, service_list['health'].port, conflate=True, poller=poller)
cal = messaging.sub_sock(context, service_list['liveCalibration'].port, conflate=True, poller=poller)
driver_monitor = messaging.sub_sock(context, service_list['driverMonitoring'].port, conflate=True, poller=poller)
gps_location = messaging.sub_sock(context, service_list['gpsLocationExternal'].port, conflate=True, poller=poller)
logcan = messaging.sub_sock(context, service_list['can'].port)
CC = car.CarControl.new_message()
CI, CP = get_car(logcan, sendcan, 1.0 if passive else None)
if CI is None:
raise Exception("unsupported car")
# if stock camera is connected, then force passive behavior
if not CP.enableCamera:
passive = True
sendcan = None
if passive:
CP.safetyModel = car.CarParams.SafetyModels.noOutput
# Get FCW toggle from settings
fcw_enabled = params.get("IsFcwEnabled") == "1"
geofence = None
PL = Planner(CP, fcw_enabled)
LoC = LongControl(CP, CI.compute_gb)
VM = VehicleModel(CP)
LaC = LatControl(CP)
AM = AlertManager()
driver_status = DriverStatus()
if not passive:
AM.add("startup", False)
# Write CarParams for radard and boardd safety mode
params.put("CarParams", CP.to_bytes())
state = State.disabled
soft_disable_timer = 0
v_cruise_kph = 255
v_cruise_kph_last = 0
overtemp = False
free_space = False
cal_status = Calibration.INVALID
cal_perc = 0
mismatch_counter = 0
low_battery = False
rk = Ratekeeper(rate, print_delay_threshold=2./1000)
# Read angle offset from previous drive, fallback to default
angle_offset = default_bias
calibration_params = params.get("CalibrationParams")
if calibration_params:
try:
calibration_params = json.loads(calibration_params)
angle_offset = calibration_params["angle_offset2"]
except (ValueError, KeyError):
pass
prof = Profiler(False) # off by default
while True:
prof.checkpoint("Ratekeeper", ignore=True)
# Sample data and compute car events
CS, events, cal_status, cal_perc, overtemp, free_space, low_battery, mismatch_counter = data_sample(CI, CC, thermal, cal, health,
driver_monitor, gps_location, poller, cal_status, cal_perc, overtemp, free_space, low_battery, driver_status, geofence, state, mismatch_counter, params)
prof.checkpoint("Sample")
# Define longitudinal plan (MPC)
plan, plan_ts = calc_plan(CS, CP, events, PL, LaC, LoC, v_cruise_kph, driver_status, geofence)
prof.checkpoint("Plan")
if not passive:
# update control state
state, soft_disable_timer, v_cruise_kph, v_cruise_kph_last = \
state_transition(CS, CP, state, events, soft_disable_timer, v_cruise_kph, AM)
prof.checkpoint("State transition")
# Compute actuators (runs PID loops and lateral MPC)
actuators, v_cruise_kph, driver_status, angle_offset = state_control(plan, CS, CP, state, events, v_cruise_kph,
v_cruise_kph_last, AM, rk, driver_status, PL, LaC, LoC, VM, angle_offset, passive, is_metric, cal_perc)
prof.checkpoint("State Control")
# Publish data
CC = data_send(PL.perception_state, plan, plan_ts, CS, CI, CP, VM, state, events, actuators, v_cruise_kph, rk, carstate, carcontrol,
live100, livempc, AM, driver_status, LaC, LoC, angle_offset, passive)
prof.checkpoint("Sent")
rk.keep_time() # Run at 100Hz
prof.display()
def loads(self, msg, encoding=None, raw=False):
'''
Run the correct loads serialization format
:param encoding: Useful for Python 3 support. If the msgpack data
was encoded using "use_bin_type=True", this will
differentiate between the 'bytes' type and the
'str' type by decoding contents with 'str' type
to what the encoding was set as. Recommended
encoding is 'utf-8' when using Python 3.
If the msgpack data was not encoded using
"use_bin_type=True", it will try to decode
all 'bytes' and 'str' data (the distinction has
been lost in this case) to what the encoding is
set as. In this case, it will fail if any of
the contents cannot be converted.
'''
try:
def ext_type_decoder(code, data):
if code == 78:
data = salt.utils.stringutils.to_unicode(data)
return datetime.datetime.strptime(data, '%Y%m%dT%H:%M:%S.%f')
return data
gc.disable() # performance optimization for msgpack
loads_kwargs = {'use_list': True,
'ext_hook': ext_type_decoder}
if msgpack.version >= (0, 4, 0):
# msgpack only supports 'encoding' starting in 0.4.0.
# Due to this, if we don't need it, don't pass it at all so
# that under Python 2 we can still work with older versions
# of msgpack.
if msgpack.version >= (0, 5, 2):
if encoding is None:
loads_kwargs['raw'] = True
else:
loads_kwargs['raw'] = False
else:
loads_kwargs['encoding'] = encoding
try:
ret = msgpack.loads(msg, **loads_kwargs)
except UnicodeDecodeError:
# msg contains binary data
loads_kwargs.pop('raw', None)
loads_kwargs.pop('encoding', None)
ret = msgpack.loads(msg, **loads_kwargs)
else:
ret = msgpack.loads(msg, **loads_kwargs)
if six.PY3 and encoding is None and not raw:
ret = salt.transport.frame.decode_embedded_strs(ret)
except Exception as exc:
log.critical(
'Could not deserialize msgpack message. This often happens '
'when trying to read a file not in binary mode. '
'To see message payload, enable debug logging and retry. '
'Exception: %s', exc
)
log.debug('Msgpack deserialization failure on message: %s', msg)
gc.collect()
raise six.raise_from(
SaltDeserializationError(
'Could not deserialize msgpack message.'
' See log for more info.'
),
exc,
)
finally:
gc.enable()
return ret
def majorUpgrade(self, doSave=False):
"""
"""
try:
xtime = time.time()
wasEnabled = gc.isenabled()
if wasEnabled:
gc.disable()
newRoot = self.newRoot
# old to new data object map
self.oldToNew = {self.oldRoot: newRoot}
# Modify old object hierarchy in place
self.massageData()
# load all new data before modifying IO map
from memops.general.Util import loadAllData
loadAllData(newRoot)
tt = xtime
xtime = time.time()
print('+++ Done load new data - time %s' % (xtime - tt))
# check for existing files, move old file as needed, and save
oldPath = getDataDir(self.oldRoot, self.oldVersionStr)
newPath = getDataDir(newRoot, self.newVersionStr)
# save destination tests. NB done regardless to avoid accidental overwrites
if oldPath == newPath:
# using same name - change old version to '_bak'
dest = oldPath + '_bak'
if os.path.exists(dest):
raise ApiError("Cannot move old data, %s already exists" %
dest)
# move old data - must be done here to avoid accidental reading instead
# files on oldPath location
if os.path.exists(oldPath):
os.rename(oldPath, dest)
# check that new locations are free:
if os.path.exists(newPath):
raise ApiError("Cannot save new data, %s already exists" %
newPath)
# Make new data and do minor upgrade
self.transferData()
finally:
if wasEnabled:
gc.enable()
tt = xtime
xtime = time.time()
print('+++ Done transfer data - time %s ' % (xtime - tt))
# check validity
# NB complete=True checking will be slow - may be relaxed later
newRoot.checkAllValid(complete=True)
tt = xtime
xtime = time.time()
print('+++ Done complete checkValid time %s' % (xtime - tt))
if doSave:
# save new data
newRoot.saveModified()
tt = xtime
xtime = time.time()
print('+++ Done save result - time %s' % (xtime - tt))
#
return newRoot
def closest_genes(inputfile=None,
outputfile=None,
from_region_type=None,
no_header=False,
nb_neighbors=1,
to_region_type=None,
same_strandedness=False,
diff_strandedness=False,
text_format=False,
identifier="gene_id",
collapse=False):
"""
Find the n closest genes for each gene.
"""
if same_strandedness and diff_strandedness:
message(
"--same-strandedness and --diff-strandedness are "
"mutually exclusive.",
type="ERROR")
# ----------------------------------------------------------------------
# load GTF
# ----------------------------------------------------------------------
gtf = GTF(inputfile)
gn_gtf = gtf.select_by_key("feature", "gene")
gn_ids = gn_gtf.get_gn_ids(nr=True)
if len(gn_gtf) == 0:
message("No gene feature found. Please use convert_ensembl.",
type="ERROR")
if nb_neighbors >= (len(gn_gtf) - 1):
message("Two much neighbors", type="ERROR")
all_ids = gn_gtf.extract_data(identifier, as_list=True, no_na=False)
if "." in all_ids:
message("Some identifiers are undefined ('.').", type="ERROR")
if len(all_ids) == 0:
message("The identifier was not found.", type="ERROR")
# ----------------------------------------------------------------------
# load GTF and requested regions (for source/'from' transcript)
# ----------------------------------------------------------------------
if from_region_type == 'tss':
from_regions = gn_gtf.get_5p_end(
feat_type="gene",
name=[identifier],
).cut([0, 1, 2, 3, 4, 5]).sort()
elif from_region_type == 'tts':
from_regions = gn_gtf.get_3p_end(
feat_type="gene",
name=[identifier],
).cut([0, 1, 2, 3, 4, 5]).sort()
elif from_region_type == 'gene':
from_regions = gn_gtf.to_bed(name=[identifier
], ).cut([0, 1, 2, 3, 4, 5]).sort()
else:
message("Unknown type.", type="ERROR")
# ----------------------------------------------------------------------
# load GTF and requested regions (for dest/'to' transcript)
# ----------------------------------------------------------------------
if to_region_type == 'tss':
to_regions = gn_gtf.get_5p_end(
feat_type="gene",
name=[identifier],
).cut([0, 1, 2, 3, 4, 5]).sort()
elif to_region_type == 'tts':
to_regions = gn_gtf.get_3p_end(
feat_type="gene",
name=[identifier],
).cut([0, 1, 2, 3, 4, 5]).sort()
elif to_region_type == 'gene':
to_regions = gn_gtf.to_bed(name=[identifier], ).cut([0, 1, 2, 3, 4,
5]).sort()
else:
message("Unknown type.", type="ERROR")
# ----------------------------------------------------------------------
# Search closest genes
# ----------------------------------------------------------------------
gene_closest = defaultdict(list)
gene_closest_dist = defaultdict(list)
closest_bo = from_regions.closest(b=to_regions,
k=nb_neighbors,
N=True,
s=same_strandedness,
S=diff_strandedness,
d=True)
for i in closest_bo:
gene_closest[i[3]] += [i[9]]
gene_closest_dist[i[3]] += [i[12]]
if not text_format:
if len(gene_closest):
gtf = gtf.add_attr_from_dict(feat="gene",
key=identifier,
a_dict=gene_closest,
new_key="closest_gn")
gtf = gtf.add_attr_from_dict(feat="gene",
key=identifier,
a_dict=gene_closest_dist,
new_key="closest_dist")
gtf.write(outputfile, gc_off=True)
else:
if not no_header:
outputfile.write("genes\tclosest_genes\tdistances\n")
for gene in gn_ids:
if not collapse:
outputfile.write("\t".join([
gene, ",".join(gene_closest[gene]), ",".join(
gene_closest_dist[gene])
]) + "\n")
else:
for closest, dist in zip(gene_closest[gene],
gene_closest_dist[gene]):
outputfile.write("\t".join([gene, closest, dist]) + "\n")
gc.disable()
close_properly(outputfile, inputfile)
def sra2illumina(input_file,
output_file,
tag_read = None,
tag='',
phred_conversion = False,
operation = 'change',
tmp_dir = None,
size_read_buffer = 10**8):
"""
It converts the FASTQ file (PHRED-33 qualities and SRA read names) downloaded
from Short Read Archive (SRA) to Illumina FASTQ file (PHRED-64 Illumina v1.5
and Illumina read names).
"""
temp_file = None
if phred_conversion:
temp_file = give_me_temp_filename(tmp_dir)
else:
temp_file = output_file
read_name = file(input_file,'r').readline().rstrip('\r\n')
sra = False
e = read_name.partition(" ")[0]
if read_name.startswith('@') and ( not(e.endswith('/1') or e.endswith('/2'))):
sra = True
if operation == 'change' or sra:
fid = open(input_file,'r')
fod = open(temp_file,'w')
i = 0
r = 0
while True:
gc.disable()
lines = fid.readlines(size_read_buffer)
gc.enable()
if not lines:
break
n = len(lines)
for j in xrange(n):
r = r + 1
i = i + 1
if i == 1:
if tag_read:
lines[j] = '@%s%s%s\n' % (tag_read ,int2str(r,12) , tag)
else: # if there is no tag_read then the original SRA id is left
lines[j] = '%s%s\n' % (lines[j][:-1].partition(" ")[0], tag)
#lines[j] = lines[j].rstrip('\r\n').upper().split(' ')[1]+tag+'\n'
elif i == 3:
lines[j] = "+\n"
elif i == 4:
i = 0
fod.writelines(lines)
fid.close()
fod.close()
if phred_conversion == '64':
phred.fq2fq(temp_file,'sanger',output_file,'illumina-1.5',tmp_dir = tmp_dir)
os.remove(temp_file)
elif phred_conversion == '33':
phred.fq2fq(temp_file,'auto-detect',output_file,'sanger',tmp_dir = tmp_dir)
os.remove(temp_file)
else:
print "No changes are done!"
if os.path.isfile(output_file):
os.remove(output_file)
if operation == 'soft':
if os.path.islink(input_file):
linkto = os.readlink(input_file)
os.symlink(linkto,ooutput_file)
else:
os.symlink(input_file,output_file)
elif operation == 'hard':
linkto = input_file
if os.path.islink(input_file):
linkto = os.readlink(input_file)
try:
os.link(linkto,output_file)
except OSError as er:
print >>sys.stderr,"WARNING: Cannot do hard links ('%s' and '%s')!" % (linkto,output_file)
shutil.copyfile(linkto,output_file)
# if er.errno == errno.EXDEV:
# # they are on different partitions
# # [Errno 18] Invalid cross-device link
# shutil.copyfile(linkto,output_file)
# else:
# print >>sys.stderr,"ERROR: Cannot do hard links ('%s' and '%s')!" % (linkto,output_file)
# print >>sys.stderr,er
# sys.exit(1)
elif operation == 'copy':
shutil.copyfile(input_file, output_file)
else:
print >>sys.stderr, "ERROR: unknown operation of linking!", operation
sys.exit(1)
def preprocessing(args):
gc.enable()
# 目的変数
target_name = 'TARGET'
#===========================
# 無用なデータを除外（結合前）
#===========================
# application_{train|test}
df_application_train = pd.read_csv(
os.path.join(args.dataset_dir, "application_train.csv"))
df_application_test = pd.read_csv(
os.path.join(args.dataset_dir, "application_test.csv"))
#df_application_train.drop(['FLAG_DOCUMENT_2', 'FLAG_DOCUMENT_4', 'FLAG_DOCUMENT_5', 'FLAG_DOCUMENT_7', 'FLAG_DOCUMENT_9', 'FLAG_DOCUMENT_10', 'FLAG_DOCUMENT_11', 'FLAG_DOCUMENT_12', 'FLAG_DOCUMENT_13', 'FLAG_DOCUMENT_14', 'FLAG_DOCUMENT_15', 'FLAG_DOCUMENT_16', 'FLAG_DOCUMENT_17', 'FLAG_DOCUMENT_18', 'FLAG_DOCUMENT_19', 'FLAG_DOCUMENT_20', 'FLAG_DOCUMENT_21'], axis=1, inplace=True)
#df_application_test.drop(['FLAG_DOCUMENT_2', 'FLAG_DOCUMENT_4', 'FLAG_DOCUMENT_5', 'FLAG_DOCUMENT_7', 'FLAG_DOCUMENT_9', 'FLAG_DOCUMENT_10', 'FLAG_DOCUMENT_11', 'FLAG_DOCUMENT_12', 'FLAG_DOCUMENT_13', 'FLAG_DOCUMENT_14', 'FLAG_DOCUMENT_15', 'FLAG_DOCUMENT_16', 'FLAG_DOCUMENT_17', 'FLAG_DOCUMENT_18', 'FLAG_DOCUMENT_19', 'FLAG_DOCUMENT_20', 'FLAG_DOCUMENT_21'], axis=1, inplace=True)
#===========================
# サブ構造の結合
#===========================
# 元データ
df_train = df_application_train
df_test = df_application_test
#---------------------------
# bureau
#---------------------------
df_bureau = pd.read_csv(os.path.join(args.dataset_dir, "bureau.csv"))
for col in df_bureau.columns:
# ラベル情報のエンコード
if (df_bureau[col].dtypes == "object"):
label_encoder = LabelEncoder()
label_encoder.fit(list(df_bureau[col]))
df_bureau[col] = label_encoder.transform(list(df_bureau[col]))
# 同じ SK_ID_CURR の行を 過去の申込み回数（SK_ID_CURR あたりの SK_ID_BUREAU の個数）, 各々の特徴量の mean, max, min, で集約する。
df_bureau_agg = df_bureau.groupby('SK_ID_CURR', as_index=False).agg(
['count', 'mean', 'max', 'min']).reset_index()
df_bureau_agg.columns = rename_columns_levels(df_bureau_agg, "bureau",
'SK_ID_CURR')
#print( df_bureau_agg.shape )
#print( df_bureau_agg.head() )
# 元のデータに統合
df_train = pd.merge(df_train, df_bureau_agg, on='SK_ID_CURR', how='left')
df_test = pd.merge(df_test, df_bureau_agg, on='SK_ID_CURR', how='left')
# 不要になったメモリを解放
del df_bureau_agg
gc.collect()
#---------------------------
# bureau_balance
#---------------------------
df_bureau_balance = pd.read_csv(
os.path.join(args.dataset_dir, "bureau_balance.csv"))
for col in df_bureau_balance.columns:
# ラベル情報のエンコード
if (df_bureau_balance[col].dtypes == "object"):
label_encoder = LabelEncoder()
label_encoder.fit(list(df_bureau_balance[col]))
df_bureau_balance[col] = label_encoder.transform(
list(df_bureau_balance[col]))
# 同じ SK_ID_BUREAU を集約
df_bureau_balance_agg = df_bureau_balance.groupby(
'SK_ID_BUREAU', as_index=False).agg(['count', 'mean', 'max',
'min']).reset_index()
df_bureau_balance_agg.columns = rename_columns_levels(
df_bureau_balance_agg, "bureau_balance", 'SK_ID_BUREAU')
# 親データ （df_bureau） の 'SK_ID_CURR' に、対応する 'SK_ID_BUREAU' を紐付け
df_bureau_balance_agg = df_bureau[['SK_ID_BUREAU', 'SK_ID_CURR'
]].merge(df_bureau_balance_agg,
on='SK_ID_BUREAU',
how='left')
# １つの `SK_ID_CURR` に対して、複数の `SK_ID_BUREAU` が存在することになるので、`SK_ID_CURR` を集約
df_bureau_balance_agg = df_bureau_balance_agg.drop(
columns=['SK_ID_BUREAU']).groupby('SK_ID_CURR', as_index=False).agg(
['count', 'mean', 'max', 'min']).reset_index()
df_bureau_balance_agg.columns = rename_columns_levels(
df_bureau_balance_agg, "bureau_balance", 'SK_ID_CURR')
#print( df_bureau_balance_agg.shape )
#print( df_bureau_balance_agg.head() )
# 元のデータに統合
df_train = pd.merge(df_train,
df_bureau_balance_agg,
on='SK_ID_CURR',
how='left')
df_test = pd.merge(df_test,
df_bureau_balance_agg,
on='SK_ID_CURR',
how='left')
#print( df_train.shape )
#print( df_train.head() )
# 不要になったメモリを解放
del df_bureau, df_bureau_balance, df_bureau_balance_agg
gc.collect()
#---------------------------
# previous_application
#--------------------------
df_previous_application = pd.read_csv(
os.path.join(args.dataset_dir, "previous_application.csv"))
for col in df_previous_application.columns:
# ラベル情報のエンコード
if (df_previous_application[col].dtypes == "object"):
label_encoder = LabelEncoder()
label_encoder.fit(list(df_previous_application[col]))
df_previous_application[col] = label_encoder.transform(
list(df_previous_application[col]))
df_previous_application_agg = df_previous_application.groupby(
'SK_ID_CURR', as_index=False).agg(['count', 'mean', 'max',
'min']).reset_index()
df_previous_application_agg.columns = rename_columns_levels(
df_previous_application_agg, "revious_application", 'SK_ID_CURR')
# 元データに統合
df_train = pd.merge(df_train,
df_previous_application_agg,
on='SK_ID_CURR',
how='left')
df_test = pd.merge(df_test,
df_previous_application_agg,
on='SK_ID_CURR',
how='left')
# 不要になったメモリを解放
del df_previous_application_agg
gc.collect()
#---------------------------
# pos_cash_balance
#---------------------------
df_pos_cash_balance = pd.read_csv(
os.path.join(args.dataset_dir, "POS_CASH_balance.csv"))
for col in df_pos_cash_balance.columns:
# ラベル情報のエンコード
if (df_pos_cash_balance[col].dtypes == "object"):
label_encoder = LabelEncoder()
label_encoder.fit(list(df_pos_cash_balance[col]))
df_pos_cash_balance[col] = label_encoder.transform(
list(df_pos_cash_balance[col]))
df_pos_cash_balance_agg = df_pos_cash_balance.groupby(
'SK_ID_PREV', as_index=False).agg(['count', 'mean', 'max',
'min']).reset_index()
#print( df_pos_cash_balance_agg.head() )
df_pos_cash_balance_agg.columns = rename_columns_levels(
df_pos_cash_balance_agg, "pos_cash_balance", 'SK_ID_PREV')
# 親データの 'SK_ID_CURR' に、対応する 'SK_ID_PREV' を紐付け
df_pos_cash_balance_agg = df_previous_application[[
'SK_ID_PREV', 'SK_ID_CURR'
]].merge(df_pos_cash_balance_agg, on='SK_ID_PREV', how='left')
#print( df_pos_cash_balance_agg.head() )
# １つの `SK_ID_CURR` に対して、複数の `SK_ID_PREV` が存在することになるので、`SK_ID_CURR` を集約
df_pos_cash_balance_agg = df_pos_cash_balance_agg.drop(
columns=['SK_ID_PREV']).groupby('SK_ID_CURR', as_index=False).agg(
['count', 'mean', 'max', 'min']).reset_index()
df_pos_cash_balance_agg.columns = rename_columns_levels(
df_pos_cash_balance_agg, "bureau_balance", 'SK_ID_CURR')
#print( df_pos_cash_balance_agg.head() )
# 元データに統合
df_train = pd.merge(df_train,
df_pos_cash_balance_agg,
on='SK_ID_CURR',
how='left')
df_test = pd.merge(df_test,
df_pos_cash_balance_agg,
on='SK_ID_CURR',
how='left')
# 不要になったメモリを解放
del df_pos_cash_balance, df_pos_cash_balance_agg
gc.collect()
#---------------------------
# installments_payments
#---------------------------
df_installments_payments = pd.read_csv(
os.path.join(args.dataset_dir, "installments_payments.csv"))
for col in df_installments_payments.columns:
# ラベル情報のエンコード
if (df_installments_payments[col].dtypes == "object"):
label_encoder = LabelEncoder()
label_encoder.fit(list(df_installments_payments[col]))
df_installments_payments[col] = label_encoder.transform(
list(df_installments_payments[col]))
df_installments_payments_agg = df_installments_payments.groupby(
'SK_ID_PREV', as_index=False).agg(['count', 'mean', 'max',
'min']).reset_index()
df_installments_payments_agg.columns = rename_columns_levels(
df_installments_payments_agg, "installments_payments", 'SK_ID_PREV')
# 親データの 'SK_ID_CURR' に、対応する 'SK_ID_PREV' を紐付け
df_installments_payments_agg = df_previous_application[[
'SK_ID_PREV', 'SK_ID_CURR'
]].merge(df_installments_payments_agg, on='SK_ID_PREV', how='left')
# １つの `SK_ID_CURR` に対して、複数の `SK_ID_PREV` が存在することになるので、`SK_ID_CURR` を集約
df_installments_payments_agg = df_installments_payments_agg.drop(
columns=['SK_ID_PREV']).groupby('SK_ID_CURR', as_index=False).agg(
['count', 'mean', 'max', 'min']).reset_index()
df_installments_payments_agg.columns = rename_columns_levels(
df_installments_payments_agg, "installments_payments", 'SK_ID_CURR')
# 元データに統合
df_train = pd.merge(df_train,
df_installments_payments_agg,
on='SK_ID_CURR',
how='left')
df_test = pd.merge(df_test,
df_installments_payments_agg,
on='SK_ID_CURR',
how='left')
# 不要になったメモリを解放
del df_installments_payments, df_installments_payments_agg
gc.collect()
#---------------------------
# credit_card_balance
#---------------------------
df_credit_card_balance = pd.read_csv(
os.path.join(args.dataset_dir, "credit_card_balance.csv"))
for col in df_credit_card_balance.columns:
# ラベル情報のエンコード
if (df_credit_card_balance[col].dtypes == "object"):
label_encoder = LabelEncoder()
label_encoder.fit(list(df_credit_card_balance[col]))
df_credit_card_balance[col] = label_encoder.transform(
list(df_credit_card_balance[col]))
df_credit_card_balance_agg = df_credit_card_balance.groupby(
'SK_ID_PREV', as_index=False).agg(['count', 'mean', 'max',
'min']).reset_index()
df_credit_card_balance_agg.columns = rename_columns_levels(
df_credit_card_balance_agg, "credit_card_balance", 'SK_ID_PREV')
# 親データの 'SK_ID_CURR' に、対応する 'SK_ID_PREV' を紐付け
df_credit_card_balance_agg = df_previous_application[[
'SK_ID_PREV', 'SK_ID_CURR'
]].merge(df_credit_card_balance_agg, on='SK_ID_PREV', how='left')
# １つの `SK_ID_CURR` に対して、複数の `SK_ID_PREV` が存在することになるので、`SK_ID_CURR` を集約
df_credit_card_balance_agg = df_credit_card_balance_agg.drop(
columns=['SK_ID_PREV']).groupby('SK_ID_CURR', as_index=False).agg(
['count', 'mean', 'max', 'min']).reset_index()
df_credit_card_balance_agg.columns = rename_columns_levels(
df_credit_card_balance_agg, "installments_payments", 'SK_ID_CURR')
# 元データに統合
df_train = pd.merge(df_train,
df_credit_card_balance_agg,
on='SK_ID_CURR',
how='left')
df_test = pd.merge(df_test,
df_credit_card_balance_agg,
on='SK_ID_CURR',
how='left')
# 不要になったメモリを解放
del df_credit_card_balance, df_credit_card_balance_agg
gc.collect()
#===========================
# 特徴量の追加（結合後）
#===========================
df_train[
'DAYS_EMPLOYED_ANOM'] = df_train["DAYS_EMPLOYED"] == 365243 # 異常値のフラグ
df_train['DAYS_EMPLOYED'].replace({365243: np.nan}, inplace=True)
df_test[
'DAYS_EMPLOYED_ANOM'] = df_test["DAYS_EMPLOYED"] == 365243 # 異常値のフラグ
df_test['DAYS_EMPLOYED'].replace({365243: np.nan}, inplace=True)
# 時系列データ
df_train['DAYS_BIRTH'] = -1 * df_train['DAYS_BIRTH']
df_test['DAYS_BIRTH'] = -1 * df_test['DAYS_BIRTH']
df_train['YEARS_BIRTH'] = df_train['DAYS_BIRTH'] / 365
df_test['YEARS_BIRTH'] = df_test['DAYS_BIRTH'] / 365
#df_train['YEARS_BINNED'] = pd.cut(df_train['YEARS_BIRTH'], bins = np.linspace(20, 70, num = 11))
#df_test['YEARS_BINNED'] = pd.cut(df_test['YEARS_BIRTH'], bins = np.linspace(20, 70, num = 11))
#===========================
# 無用なデータを除外（結合後）
#===========================
if 'SK_ID_CURR' in df_train.columns:
df_train.drop(['SK_ID_CURR'], axis=1, inplace=True)
df_test.drop(['SK_ID_CURR'], axis=1, inplace=True)
if 'SK_ID_BUREAU' in df_train.columns:
df_train.drop(['SK_ID_BUREAU'], axis=1, inplace=True)
df_test.drop(['SK_ID_BUREAU'], axis=1, inplace=True)
if 'SK_ID_PREV' in df_train.columns:
df_train.drop(['SK_ID_PREV'], axis=1, inplace=True)
df_test.drop(['SK_ID_PREV'], axis=1, inplace=True)
#===========================
# 全特徴量を一括で処理
#===========================
# 全データセット
df_data = pd.concat([df_train, df_test], sort=False)
for col in df_train.columns:
# 目的変数
if (col in [target_name]):
continue
#-----------------------------
# ラベル情報のエンコード
#-----------------------------
if (df_train[col].dtypes == "object"):
label_encoder = LabelEncoder()
label_encoder.fit(list(df_data[col]))
df_train[col] = label_encoder.transform(list(df_train[col]))
label_encoder = LabelEncoder()
label_encoder.fit(list(df_data[col]))
df_test[col] = label_encoder.transform(list(df_test[col]))
#-----------------------------
# 欠損値の埋め合わせ
#-----------------------------
# NAN 値の埋め合わせ（平均値）
if (col in ["OWN_CAR_AGE"]):
# データセット全体 df_data での平均値とする
df_train[col].fillna(np.mean(df_data[col]), inplace=True)
df_test[col].fillna(np.mean(df_data[col]), inplace=True)
# NAN 値の埋め合わせ（ゼロ値）/ int 型
elif (df_train[col].dtypes in [
"int8", "int16", "int32", "int64", "uint8", "uint16", "uint32",
"uint64"
]):
df_train[col].fillna(0, inplace=True)
df_test[col].fillna(0, inplace=True)
# NAN 値の埋め合わせ（ゼロ値）/ float 型
elif (df_train[col].dtypes
in ["float16", "float32", "float64", "float128"]):
df_train[col].fillna(0.0, inplace=True)
df_test[col].fillna(0.0, inplace=True)
# NAN 値の補完（None値）/ object 型
else:
df_train[col] = df_train[col].fillna('NA')
df_test[col] = df_test[col].fillna('NA')
#-----------------------------
# 正規化処理
#-----------------------------
"""
if( df_train[col].dtypes in ["float16", "float32", "float64", "float128"] ):
scaler = StandardScaler()
scaler.fit( df_train[col].values.reshape(-1,1) )
df_train[col] = scaler.transform( df_train[col].values.reshape(-1,1) )
df_test[col] = scaler.transform( df_test[col].values.reshape(-1,1) )
"""
#-----------------------------
# 値が単一の特徴量をクレンジング
#-----------------------------
"""
if( df_train[col].nunique() == 1 ):
print( "remove {} : {}".format(col,df_train[col].nunique()) )
df_train.drop([col], axis=1, inplace=True)
df_test.drop([col], axis=1, inplace=True)
"""
gc.disable()
return df_train, df_test
def dictionaryAttack(attack_dict):
suffix_list = [""]
custom_wordlist = [""]
hash_regexes = []
results = []
resumes = []
user_hash = []
processException = False
foundHash = False
for (_, hashes) in attack_dict.items():
for hash_ in hashes:
if not hash_:
continue
hash_ = hash_.split()[0] if hash_ and hash_.strip() else hash_
regex = hashRecognition(hash_)
if regex and regex not in hash_regexes:
hash_regexes.append(regex)
infoMsg = "using hash method '%s'" % __functions__[
regex].func_name
logger.info(infoMsg)
for hash_regex in hash_regexes:
keys = set()
attack_info = []
for (user, hashes) in attack_dict.items():
for hash_ in hashes:
if not hash_:
continue
foundHash = True
hash_ = hash_.split()[0] if hash_ and hash_.strip() else hash_
if re.match(hash_regex, hash_):
item = None
if hash_regex not in (HASH.CRYPT_GENERIC, HASH.WORDPRESS):
hash_ = hash_.lower()
if hash_regex in (HASH.MYSQL, HASH.MYSQL_OLD,
HASH.MD5_GENERIC, HASH.SHA1_GENERIC):
item = [(user, hash_), {}]
elif hash_regex in (HASH.ORACLE_OLD, HASH.POSTGRES):
item = [(user, hash_), {'username': user}]
elif hash_regex in (HASH.ORACLE, ):
item = [(user, hash_), {'salt': hash_[-20:]}]
elif hash_regex in (HASH.MSSQL, HASH.MSSQL_OLD,
HASH.MSSQL_NEW):
item = [(user, hash_), {'salt': hash_[6:14]}]
elif hash_regex in (HASH.CRYPT_GENERIC, ):
item = [(user, hash_), {'salt': hash_[0:2]}]
elif hash_regex in (HASH.WORDPRESS, ):
if ITOA64.index(hash_[3]) < 32:
item = [(user, hash_), {
'salt': hash_[4:12],
'count': 1 << ITOA64.index(hash_[3]),
'prefix': hash_[:12]
}]
else:
warnMsg = "invalid hash '%s'" % hash_
logger.warn(warnMsg)
if item and hash_ not in keys:
resumed = hashDBRetrieve(hash_)
if not resumed:
attack_info.append(item)
user_hash.append(item[0])
else:
infoMsg = "resuming password '%s' for hash '%s'" % (
resumed, hash_)
if user and not user.startswith(DUMMY_USER_PREFIX):
infoMsg += " for user '%s'" % user
logger.info(infoMsg)
resumes.append((user, hash_, resumed))
keys.add(hash_)
if not attack_info:
continue
if not kb.wordlists:
while not kb.wordlists:
# the slowest of all methods hence smaller default dict
if hash_regex in (HASH.ORACLE_OLD, HASH.WORDPRESS):
dictPaths = [paths.SMALL_DICT]
else:
dictPaths = [paths.WORDLIST]
message = "what dictionary do you want to use?\n"
message += "[1] default dictionary file '%s' (press Enter)\n" % dictPaths[
0]
message += "[2] custom dictionary file\n"
message += "[3] file with list of dictionary files"
choice = readInput(message, default='1')
try:
if choice == '2':
message = "what's the custom dictionary's location?\n"
_ = readInput(message)
if _:
dictPaths = [readInput(message)]
logger.info("using custom dictionary")
elif choice == '3':
message = "what's the list file location?\n"
listPath = readInput(message)
checkFile(listPath)
dictPaths = getFileItems(listPath)
logger.info("using custom list of dictionaries")
else:
logger.info("using default dictionary")
dictPaths = filter(None, dictPaths)
for dictPath in dictPaths:
checkFile(dictPath)
if os.path.splitext(dictPath)[1].lower() == ".zip":
_ = zipfile.ZipFile(dictPath, 'r')
if len(_.namelist()) == 0:
errMsg = "no file(s) inside '%s'" % dictPath
raise SqlmapDataException(errMsg)
else:
_.open(_.namelist()[0])
kb.wordlists = dictPaths
except Exception, ex:
warnMsg = "there was a problem while loading dictionaries"
warnMsg += " ('%s')" % getSafeExString(ex)
logger.critical(warnMsg)
message = "do you want to use common password suffixes? (slow!) [y/N] "
if readInput(message, default='N', boolean=True):
suffix_list += COMMON_PASSWORD_SUFFIXES
infoMsg = "starting dictionary-based cracking (%s)" % __functions__[
hash_regex].func_name
logger.info(infoMsg)
for item in attack_info:
((user, _), _) = item
if user and not user.startswith(DUMMY_USER_PREFIX):
custom_wordlist.append(normalizeUnicode(user))
if hash_regex in (HASH.MYSQL, HASH.MYSQL_OLD, HASH.MD5_GENERIC,
HASH.SHA1_GENERIC):
for suffix in suffix_list:
if not attack_info or processException:
break
if suffix:
clearConsoleLine()
infoMsg = "using suffix '%s'" % suffix
logger.info(infoMsg)
retVal = None
processes = []
try:
if _multiprocessing:
if _multiprocessing.cpu_count() > 1:
infoMsg = "starting %d processes " % _multiprocessing.cpu_count(
)
singleTimeLogMessage(infoMsg)
gc.disable()
retVal = _multiprocessing.Queue()
count = _multiprocessing.Value(
'i', _multiprocessing.cpu_count())
for i in xrange(_multiprocessing.cpu_count()):
process = _multiprocessing.Process(
target=_bruteProcessVariantA,
args=(attack_info, hash_regex, suffix, retVal,
i, count, kb.wordlists, custom_wordlist,
conf.api))
processes.append(process)
for process in processes:
process.daemon = True
process.start()
while count.value > 0:
time.sleep(0.5)
else:
warnMsg = "multiprocessing hash cracking is currently "
warnMsg += "not supported on this platform"
singleTimeWarnMessage(warnMsg)
retVal = Queue()
_bruteProcessVariantA(attack_info, hash_regex, suffix,
retVal, 0, 1, kb.wordlists,
custom_wordlist, conf.api)
except KeyboardInterrupt:
print
processException = True
warnMsg = "user aborted during dictionary-based attack phase (Ctrl+C was pressed)"
logger.warn(warnMsg)
for process in processes:
try:
process.terminate()
process.join()
except (OSError, AttributeError):
pass
finally:
if _multiprocessing:
gc.enable()
if retVal:
conf.hashDB.beginTransaction()
while not retVal.empty():
user, hash_, word = item = retVal.get(block=False)
attack_info = filter(
lambda _: _[0][0] != user or _[0][1] != hash_,
attack_info)
hashDBWrite(hash_, word)
results.append(item)
conf.hashDB.endTransaction()
clearConsoleLine()
else:
for ((user, hash_), kwargs) in attack_info:
if processException:
break
if any(_[0] == user and _[1] == hash_ for _ in results):
continue
count = 0
found = False
for suffix in suffix_list:
if found or processException:
break
if suffix:
clearConsoleLine()
infoMsg = "using suffix '%s'" % suffix
logger.info(infoMsg)
retVal = None
processes = []
try:
if _multiprocessing:
if _multiprocessing.cpu_count() > 1:
infoMsg = "starting %d processes " % _multiprocessing.cpu_count(
)
singleTimeLogMessage(infoMsg)
gc.disable()
retVal = _multiprocessing.Queue()
found_ = _multiprocessing.Value('i', False)
count = _multiprocessing.Value(
'i', _multiprocessing.cpu_count())
for i in xrange(_multiprocessing.cpu_count()):
process = _multiprocessing.Process(
target=_bruteProcessVariantB,
args=(user, hash_, kwargs, hash_regex,
suffix, retVal, found_, i, count,
kb.wordlists, custom_wordlist,
conf.api))
processes.append(process)
for process in processes:
process.daemon = True
process.start()
while count.value > 0:
time.sleep(0.5)
found = found_.value != 0
else:
warnMsg = "multiprocessing hash cracking is currently "
warnMsg += "not supported on this platform"
singleTimeWarnMessage(warnMsg)
class Value():
pass
retVal = Queue()
found_ = Value()
found_.value = False
_bruteProcessVariantB(user, hash_, kwargs,
hash_regex, suffix, retVal,
found_, 0, 1, kb.wordlists,
custom_wordlist, conf.api)
found = found_.value
except KeyboardInterrupt:
print
processException = True
warnMsg = "user aborted during dictionary-based attack phase (Ctrl+C was pressed)"
logger.warn(warnMsg)
for process in processes:
try:
process.terminate()
process.join()
except (OSError, AttributeError):
pass
finally:
if _multiprocessing:
gc.enable()
if retVal:
conf.hashDB.beginTransaction()
while not retVal.empty():
user, hash_, word = item = retVal.get(
block=False)
hashDBWrite(hash_, word)
results.append(item)
conf.hashDB.endTransaction()
clearConsoleLine()
Parse function for setuptools_scm that ignores tags for non-C++
subprojects, e.g. apache-arrow-js-XXX tags.
"""
from setuptools_scm.git import parse
kwargs['describe_command'] = \
"git describe --dirty --tags --long --match 'apache-arrow-[0-9].*'"
return parse(root, **kwargs)
__version__ = setuptools_scm.get_version('../', parse=parse_git)
except ImportError:
__version__ = None
# ARROW-8684: Disable GC while initializing Cython extension module,
# to workaround Cython bug in https://github.com/cython/cython/issues/3603
_gc_enabled = _gc.isenabled()
_gc.disable()
import pyarrow.lib as _lib
if _gc_enabled:
_gc.enable()
from pyarrow.lib import cpu_count, set_cpu_count
from pyarrow.lib import (
null, bool_, int8, int16, int32, int64, uint8, uint16, uint32, uint64,
time32, time64, timestamp, date32, date64, duration, float16, float32,
float64, binary, string, utf8, large_binary, large_string, large_utf8,
decimal128, list_, large_list, map_, struct, union, dictionary, field,
type_for_alias, DataType, DictionaryType, StructType, ListType,
LargeListType, MapType, FixedSizeListType, UnionType, TimestampType,
Time32Type, Time64Type, DurationType, FixedSizeBinaryType, Decimal128Type,
BaseExtensionType, ExtensionType, PyExtensionType, UnknownExtensionType,
register_extension_type, unregister_extension_type, DictionaryMemo,
def main():
print "\tCreating 1 unreachable object => should be destroyed automatically"
make_unreachable_object()
print "\tCollect...=> no unreachable or uncollectable should be visible"
collected = gc.collect(0)
assert (collected == 0)
assert (gc.garbage == [])
print "\tCreating 1 unreachable object with ref cycles"
make_unreachable_object_with_ref_cycles()
print "\tCollect...=> 1 unreachable and 0 uncollectable should be visible"
collected = gc.collect(0)
assert (collected == 1)
assert (gc.garbage == [])
print "\tCreating 1 uncollectable object"
make_uncollectable_object()
print "\tCollect...=> Unreachables and uncollectables should be visible"
collected = gc.collect(0)
assert (collected != 0)
assert (gc.garbage != [])
if __name__ == "__main__":
gc.disable() # disable automatic garbage collection
gc.set_debug(gc.DEBUG_STATS | gc.DEBUG_COLLECTABLE | gc.DEBUG_UNCOLLECTABLE
| gc.DEBUG_INSTANCES
| gc.DEBUG_OBJECTS) # activate gc debug flags
main()
gc.set_debug(0) # deactivate gc debug flags
def controlsd_thread(sm=None, pm=None, can_sock=None):
gc.disable()
# start the loop
set_realtime_priority(3)
params = Params()
is_metric = params.get("IsMetric", encoding='utf8') == "1"
is_ldw_enabled = params.get("IsLdwEnabled", encoding='utf8') == "1"
passive = params.get("Passive", encoding='utf8') == "1"
openpilot_enabled_toggle = params.get("OpenpilotEnabledToggle", encoding='utf8') == "1"
community_feature_toggle = params.get("CommunityFeaturesToggle", encoding='utf8') == "1"
passive = passive or not openpilot_enabled_toggle
# Pub/Sub Sockets
if pm is None:
pm = messaging.PubMaster(['sendcan', 'controlsState', 'carState', 'carControl', 'carEvents', 'carParams'])
if sm is None:
sm = messaging.SubMaster(['thermal', 'health', 'liveCalibration', 'driverMonitoring', 'plan', 'pathPlan', \
'model', 'gpsLocation'], ignore_alive=['gpsLocation'])
if can_sock is None:
can_timeout = None if os.environ.get('NO_CAN_TIMEOUT', False) else 100
can_sock = messaging.sub_sock('can', timeout=can_timeout)
# wait for health and CAN packets
hw_type = messaging.recv_one(sm.sock['health']).health.hwType
has_relay = hw_type in [HwType.blackPanda, HwType.uno]
print("Waiting for CAN messages...")
messaging.get_one_can(can_sock)
CI, CP = get_car(can_sock, pm.sock['sendcan'], has_relay)
car_recognized = CP.carName != 'mock'
# If stock camera is disconnected, we loaded car controls and it's not chffrplus
controller_available = CP.enableCamera and CI.CC is not None and not passive
community_feature_disallowed = CP.communityFeature and not community_feature_toggle
read_only = not car_recognized or not controller_available or CP.dashcamOnly or community_feature_disallowed
if read_only:
CP.safetyModel = car.CarParams.SafetyModel.noOutput
# Write CarParams for radard and boardd safety mode
params.put("CarParams", CP.to_bytes())
params.put("LongitudinalControl", "1" if CP.openpilotLongitudinalControl else "0")
CC = car.CarControl.new_message()
AM = AlertManager()
startup_alert = get_startup_alert(car_recognized, controller_available)
AM.add(sm.frame, startup_alert, False)
LoC = LongControl(CP, CI.compute_gb)
VM = VehicleModel(CP)
if CP.lateralTuning.which() == 'pid':
LaC = LatControlPID(CP)
elif CP.lateralTuning.which() == 'indi':
LaC = LatControlINDI(CP)
elif CP.lateralTuning.which() == 'lqr':
LaC = LatControlLQR(CP)
driver_status = DriverStatus()
is_rhd = params.get("IsRHD")
if is_rhd is not None:
driver_status.is_rhd = bool(int(is_rhd))
state = State.disabled
soft_disable_timer = 0
v_cruise_kph = 255
v_cruise_kph_last = 0
mismatch_counter = 0
last_blinker_frame = 0
events_prev = []
sm['liveCalibration'].calStatus = Calibration.INVALID
sm['pathPlan'].sensorValid = True
sm['pathPlan'].posenetValid = True
sm['thermal'].freeSpace = 1.
# detect sound card presence
sounds_available = not os.path.isfile('/EON') or (os.path.isdir('/proc/asound/card0') and open('/proc/asound/card0/state').read().strip() == 'ONLINE')
# controlsd is driven by can recv, expected at 100Hz
rk = Ratekeeper(100, print_delay_threshold=None)
internet_needed = params.get("Offroad_ConnectivityNeeded", encoding='utf8') is not None
prof = Profiler(False) # off by default
while True:
start_time = sec_since_boot()
prof.checkpoint("Ratekeeper", ignore=True)
# Sample data and compute car events
CS, events, cal_perc, mismatch_counter = data_sample(CI, CC, sm, can_sock, driver_status, state, mismatch_counter, params)
prof.checkpoint("Sample")
# Create alerts
if not sm.all_alive_and_valid():
events.append(create_event('commIssue', [ET.NO_ENTRY, ET.SOFT_DISABLE]))
if not sm['pathPlan'].mpcSolutionValid:
events.append(create_event('plannerError', [ET.NO_ENTRY, ET.IMMEDIATE_DISABLE]))
if not sm['pathPlan'].sensorValid:
events.append(create_event('sensorDataInvalid', [ET.NO_ENTRY, ET.PERMANENT]))
if not sm['pathPlan'].paramsValid:
events.append(create_event('vehicleModelInvalid', [ET.WARNING]))
if not sm['pathPlan'].posenetValid:
events.append(create_event('posenetInvalid', [ET.NO_ENTRY, ET.WARNING]))
if not sm['plan'].radarValid:
events.append(create_event('radarFault', [ET.NO_ENTRY, ET.SOFT_DISABLE]))
if sm['plan'].radarCanError:
events.append(create_event('radarCanError', [ET.NO_ENTRY, ET.SOFT_DISABLE]))
if not CS.canValid:
events.append(create_event('canError', [ET.NO_ENTRY, ET.IMMEDIATE_DISABLE]))
if not sounds_available:
events.append(create_event('soundsUnavailable', [ET.NO_ENTRY, ET.PERMANENT]))
if internet_needed:
events.append(create_event('internetConnectivityNeeded', [ET.NO_ENTRY, ET.PERMANENT]))
if community_feature_disallowed:
events.append(create_event('communityFeatureDisallowed', [ET.PERMANENT]))
# Only allow engagement with brake pressed when stopped behind another stopped car
if CS.brakePressed and sm['plan'].vTargetFuture >= STARTING_TARGET_SPEED and not CP.radarOffCan and CS.vEgo < 0.3:
events.append(create_event('noTarget', [ET.NO_ENTRY, ET.IMMEDIATE_DISABLE]))
if not read_only:
# update control state
state, soft_disable_timer, v_cruise_kph, v_cruise_kph_last = \
state_transition(sm.frame, CS, CP, state, events, soft_disable_timer, v_cruise_kph, AM)
prof.checkpoint("State transition")
# Compute actuators (runs PID loops and lateral MPC)
actuators, v_cruise_kph, driver_status, v_acc, a_acc, lac_log, last_blinker_frame = \
state_control(sm.frame, sm.rcv_frame, sm['plan'], sm['pathPlan'], CS, CP, state, events, v_cruise_kph, v_cruise_kph_last, AM, rk,
driver_status, LaC, LoC, read_only, is_metric, cal_perc, last_blinker_frame)
prof.checkpoint("State Control")
# Publish data
CC, events_prev = data_send(sm, pm, CS, CI, CP, VM, state, events, actuators, v_cruise_kph, rk, AM, driver_status, LaC,
LoC, read_only, start_time, v_acc, a_acc, lac_log, events_prev, last_blinker_frame, is_ldw_enabled)
prof.checkpoint("Sent")
rk.monitor_time()
prof.display()
