def addModeledPrt():
for tname in [_.rstrip() for _ in file("../dat/casf_names.txt")]:
luigi.build([AddCenter2Modeled(tname,
version="0.7"),
AddCenter2Modeled(tname,
version="0.5")],
local_scheduler=True)
def test_row_overload(self):
"""Overload the rows method and we should be able to insert data into database"""
class SQLARowOverloadTest(sqla.CopyToTable):
columns = [(["item", sqlalchemy.String(64)], {}), (["property", sqlalchemy.String(64)], {})]
connection_string = CONNECTION_STRING
table = "item_property"
chunk_size = 1
def rows(self):
tasks = [
("item0", "property0"),
("item1", "property1"),
("item2", "property2"),
("item3", "property3"),
("item4", "property4"),
("item5", "property5"),
("item6", "property6"),
("item7", "property7"),
("item8", "property8"),
("item9", "property9"),
]
for row in tasks:
yield row
task = SQLARowOverloadTest()
luigi.build([task], local_scheduler=True, workers=self.NUM_WORKERS)
self._check_entries(self.engine)
def main(tname, subset):
luigi.build([Curate.PairWisePsScore(tname=tname,
subset=subset),
Curate.PairWiseTanimoto(tname=tname,
subset=subset)],
local_scheduler=True)
pass
def test_unified_snapshot(self):
task = UnifiedSnapshotMock()
luigi.build([task], local_scheduler=True)
print(task.output().path)
with task.output().open() as handle:
lines = handle.readlines()
self.assertEquals(186, len(lines))
def setUp(self):
super(TestVisualiser, self).setUp()
x = 'I scream for ice cream'
task = UberTask(base_task=FailingMergeSort, x=x, copies=4)
luigi.build([task], workers=1, scheduler_port=self.get_http_port())
self.done = threading.Event()
def _do_ioloop():
# Enter ioloop for maximum TEST_TIMEOUT. Check every 2s whether the test has finished.
print('Entering event loop in separate thread')
for i in range(TEST_TIMEOUT):
try:
self.wait(timeout=1)
except AssertionError:
pass
if self.done.is_set():
break
print('Exiting event loop thread')
self.iothread = threading.Thread(target=_do_ioloop)
self.iothread.start()
def test():
luigi.build(
[
Extract("1b9vA"),
PrepareLigStateEnsemble("1b9vA"),
],
local_scheduler=True, )
def closest(self):
task = VIAFLatestDate()
luigi.build([task], local_scheduler=True)
with task.output().open() as handle:
date, _ = handle.iter_tsv(cols=('date', 'url')).next()
dateobj = datetime.date(*map(int, date.split('-')))
return dateobj
def main():
# get the arguments from the command line
parser = build_arg_parser()
cmdline_args = parser.parse_args()
# get a named tuple of indexes v1...vN
indexes = Load.label_indices(
cmdline_args.backup_count + 1, cmdline_args.index)
# get the end class
task = Load(indexes=indexes,
mapping_file=cmdline_args.mapping_file,
settings_file=cmdline_args.settings_file,
docs_file=cmdline_args.docs_file,
table=cmdline_args.table,
sql_filter=cmdline_args.sql_filter,
marker_table=cmdline_args.marker_table,
es_timeout=cmdline_args.es_timeout)
if cmdline_args.clear:
clear(task)
else:
if cmdline_args.restart:
clear(task)
luigi.build([task], local_scheduler=True)
# luigi suppresses exceptions
if capture_task_exceptions:
raise Exception(
'A luigi.Task failed, see traceback in the Luigi Execution Summary')
def requires(self):
task = SWBOpenDataDates(date=self.date)
luigi.build([task], local_scheduler=True)
with task.output().open() as handle:
for row in handle.iter_tsv(cols=('date',)):
dateobj = datetime.date(*map(int, row.date.split('-')))
yield SWBOpenDataListified(date=dateobj)
def run(self):
output = shellout("cut -f2 {input} | LANG=C sort | LANG=C uniq > {output}", input=self.input().path)
with open(output) as handle:
dates = map(string.strip, handle.readlines())
with self.output().open('w') as output:
for date in dates:
dateobj = datetime.date(*map(int, date.split('-')))
marc = SWBOpenDataMarc(date=dateobj)
sdb = SWBOpenDataSeekMapDB(date=dateobj)
luigi.build([marc, sdb], local_scheduler=True)
with open(marc.output().path) as handle:
with sqlite3db(sdb.output().path) as cursor:
idset = df[df.date == date].id.values.tolist()
limit, offset = self.limit, 0
while True:
cursor.execute("""
SELECT offset, length
FROM seekmap WHERE id IN (%s)""" % (
','.join(("'%s'" % id for id in idset[offset:offset + limit]))))
rows = cursor.fetchall()
if not rows:
break
else:
copyregions(handle, output, rows)
offset += limit
def test():
task1 = VinaResultAccuracy("3ofl_JHM_B_1.pdb")
task2 = QueryVinaResultOnIdenticalTemplate("3ofl_JHM_B_1.pdb")
task3 = QueryVinaResultOnIdenticalTemplate("1yst_U10_L_1.pdb")
task4 = QueryVinaResultOnIdenticalTemplate("3mbm_717_C_1.pdb")
task5 = QueryVinaResultOnIdenticalTemplate("3jdw_ORN_A_1.pdb")
luigi.build([task1, task2, task3, task4, task5], local_scheduler=True)
def test_update_table_task(self, mock_config):
mock_config.get_config.return_value.get.return_value = AWS_ACCESS_KEY
t = TestUpdateDynamoDBTableTask()
# mock s3 location for writing output token
s3_client = S3Client(AWS_ACCESS_KEY, AWS_SECRET_KEY)
s3_client.s3.create_bucket("mybucket")
# create table
table_name = "dynamo_table1"
schema = [HashKey("my_hash", data_type=STRING)]
indexes = [
AllIndex(
"IndexName", parts=[HashKey("my_hash", data_type=STRING), RangeKey("range_index", data_type=NUMBER)]
)
]
throughput = {"read": 2, "write": 4}
client = DynamoDBClient(aws_access_key_id=AWS_ACCESS_KEY, aws_secret_access_key=AWS_SECRET_KEY)
client.create_table(table_name, schema, throughput, indexes=indexes)
luigi.build([t], local_scheduler=True)
table = client.get_table("dynamo_table1")
self.assertEquals(8, table.throughput["read"])
self.assertEquals(16, table.throughput["write"])
def test():
luigi.build(
[AddNativeCenter("3owjA00"), AddCenter2Modeled("3owjA00",
version="0.7"),
AddCenter2Modeled("3owjA00",
version="0.5")],
local_scheduler=True)
def test_column_row_separator(self):
"""
Test alternate column row separator works
:return:
"""
class ModBaseTask(luigi.Task):
def output(self):
return MockFile("ModBaseTask", mirror_on_stderr=True)
def run(self):
out = self.output().open("w")
tasks = ["item%d,property%d\n" % (i, i) for i in range(10)]
for task in tasks:
out.write(task)
out.close()
class ModSQLATask(sqla.CopyToTable):
columns = [(["item", sqlalchemy.String(64)], {}), (["property", sqlalchemy.String(64)], {})]
connection_string = CONNECTION_STRING
table = "item_property"
column_separator = ","
chunk_size = 1
def requires(self):
return ModBaseTask()
task1, task2 = ModBaseTask(), ModSQLATask()
luigi.build([task1, task2], local_scheduler=True, workers=self.NUM_WORKERS)
self._check_entries(self.engine)
def mariobros(
targets=('DEFAULT',), mariofile='mario.ini', print_ns=False, dry_run=False, workers=1,
port=8082,
**kwargs
):
"""Main mariobros entry point. Parse the configuration file and launch the build of targets.
:param sequence targets: List of targets.
:param unicode mariofile: MarioFile name.
:param bool print_ns: Flag to print namespace.
:param bool dry_run: Dry run flag.
:param int workers: Number of workers.
:param dict kwargs: Passed to the luigi.build function.
"""
assert all(isinstance(target, str) for target in targets)
if dry_run and workers > 1:
workers = 1
mario.LOGGER.warning('Dry run is incompatible with multiprocessing. Setting --workers=1')
section_namespaces = mariofile_.parse_mariofile(mariofile)
default_namespace, rendered_namespaces = mario.render_config(section_namespaces)
if print_ns:
namespaces = mario.print_namespaces(default_namespace, section_namespaces)
print(namespaces)
else:
target_tasks = mario.mario(
rendered_namespaces, default_namespace, targets=targets, dry_run=dry_run
)
luigi.build(target_tasks, workers=workers, scheduler_port=port, **kwargs)
def test_multiple_tasks(self):
"""
Test a case where there are multiple tasks
:return:
"""
class SmallSQLATask(sqla.CopyToTable):
item = luigi.Parameter()
property = luigi.Parameter()
columns = [(["item", sqlalchemy.String(64)], {}), (["property", sqlalchemy.String(64)], {})]
connection_string = CONNECTION_STRING
table = "item_property"
chunk_size = 1
def rows(self):
yield (self.item, self.property)
class ManyBaseTask(luigi.Task):
def requires(self):
for t in TASK_LIST:
item, property = t.strip().split("\t")
yield SmallSQLATask(item=item, property=property)
task2 = ManyBaseTask()
luigi.build([task2], local_scheduler=True, workers=self.NUM_WORKERS)
self._check_entries(self.engine)
def test_customized_worker(self):
a = DummyTask(3)
self.assertFalse(a.complete())
self.assertFalse(self.worker.complete())
luigi.build([a], scheduler_instance=self.scheduler, worker_instance=self.worker)
self.assertTrue(a.complete())
self.assertTrue(self.worker.complete())
def test_map_only(self):
luigi.build([MapOnlyJob()], local_scheduler=True)
c = []
for line in File("luigitest-3").open("r"):
c.append(line.strip())
self.assertEquals(c[0], "kj")
self.assertEquals(c[4], "ljoi")
def test_map_only(self):
luigi.build([MapOnlyJob()], local_scheduler=True)
c = []
for line in File('luigitest-3').open('r'):
c.append(line.strip())
self.assertEquals(c[0], 'kj')
self.assertEquals(c[4], 'ljoi')
def run(self):
self.setup_arguments()
self.process_arguments(self.parser.parse_args())
self.log_arguments()
luigi.build(self.create_tasks(), local_scheduler=self.local_scheduler, workers=self.workers)
def test_extract(self, mock_config):
"""
Tests that the mysql extraction runs correctly.
"""
output_file = os.path.join(tempfile.gettempdir(), 'extract-%s.txt' % uuid.uuid4().hex)
conf = ConfigParser.ConfigParser()
conf.add_section('mysql')
conf.set('mysql', 'dbname', 'mydb')
conf.set('mysql', 'host', 'myhost')
conf.set('mysql', 'port', '3306')
conf.set('mysql', 'user', 'myuser')
conf.set('mysql', 'password', 'my_password')
mock_config.get_config.return_value = conf
with mock.patch("mortar.luigi.dbms.subprocess") as subprocess:
subprocess_return = mock.Mock()
subprocess_return.stdout = StringIO.StringIO()
subprocess_return.stderr = StringIO.StringIO()
subprocess_return.communicate.return_value = (None, None)
subprocess_return.returncode = 0
subprocess.Popen.return_value = subprocess_return
t = dbms.ExtractFromMySQL(table='foo', output_path=output_file)
self.assertFalse(t.output()[0].exists())
luigi.build([t], local_scheduler=True)
self.assertTrue(t.output()[0].exists())
os.remove(output_file)
def main():
op = OptionParser()
op.add_option('--interval', '-i', default=1000)
op.add_option('--directory', '-d')
op.add_option('--config', '-c')
op.add_option('--start', '-s')
op.add_option('--end', '-e')
op.add_option('--workflow', '-w')
op.add_option('--filetype', '-f', default='json')
options, arguments = op.parse_args()
if not options.config:
op.error('No configuration YAML')
if not options.directory:
op.error('No input file directory')
if not options.workflow:
op.error('No workflow specified')
files = glob.glob(
os.path.join(
options.directory, '*.{0}'.format(options.filetype)
)
)
if not files:
op.error(
'Empty input file directory (no {0})'.format(options.filetype)
)
try:
interval = int(options.interval)
except:
op.error('Non-integer interval value')
try:
start_index = int(options.start)
except:
start_index = 0
try:
end_index = int(options.end)
except:
end_index = len(files)
# this only works for the workflows imported above (and they
# need to be imported, obv). things like the eda, which aren't
# working on a single file dependency tree, need something else
try:
workflow_class = getattr(sys.modules[__name__], options.workflow)
except AttributeError:
op.error('Unable to load workflow for {0}'.format(options.workflow))
for i in xrange(start_index, end_index, interval):
w = workflow_class(
doc_dir=options.directory,
yaml_file=options.config,
start_index=i,
end_index=(i + interval) if (i + interval) < end_index else end_index
)
luigi.build([w], local_scheduler=True)
def test_rows(self):
task, task0 = SQLATask(), BaseTask()
luigi.build([task, task0], local_scheduler=True, workers=self.NUM_WORKERS)
for i, row in enumerate(task.rows()):
given = TASK_LIST[i].strip("\n").split("\t")
self.assertEqual(row, given)
def test_customized_worker(self):
a = DummyTask(3)
self.assertFalse(a.complete())
self.assertFalse(self.worker_scheduler_factory.worker.complete())
luigi.build([a], worker_scheduler_factory=self.worker_scheduler_factory)
self.assertTrue(a.complete())
self.assertTrue(self.worker_scheduler_factory.worker.complete())
def test_reflect(self):
"""
If the table is setup already, then one can set reflect to True, and
completely skip the columns part. It is not even required at that point.
:return:
"""
class AnotherSQLATask(sqla.CopyToTable):
connection_string = CONNECTION_STRING
table = "item_property"
reflect = True
chunk_size = 1
def requires(self):
return SQLATask()
def copy(self, conn, ins_rows, table_bound):
ins = (
table_bound.update()
.where(table_bound.c.property == sqlalchemy.bindparam("_property"))
.values({table_bound.c.item: sqlalchemy.bindparam("_item")})
)
conn.execute(ins, ins_rows)
def rows(self):
for line in TASK_LIST:
yield line.strip("\n").split("\t")
task0, task1, task2 = AnotherSQLATask(), SQLATask(), BaseTask()
luigi.build([task0, task1, task2], local_scheduler=True, workers=self.NUM_WORKERS)
self._check_entries(self.engine)
def test_use_json_as_data_interchange_format_job(test_case):
job = UseJsonAsDataInteterchangeFormatJob(use_hdfs=test_case.use_hdfs)
luigi.build([job], local_scheduler=True)
c = []
for line in job.output().open("r"):
c.append(line)
test_case.assertEqual(c, ['{"data type": "json"}\n'])
def main(sdf):
luigi.build([
PairwiseCms(sdf, maximum_radius=8.0),
PairwiseCmsEvenlyDistributed(sdf, maximum_radius=8.0),
PairwiseCmsEvenlyGryration(sdf)
],
local_scheduler=True)
def test_run_job_with_dump(self, mock_check_output):
mock_check_output.side_effect = [
'Your job 12345 ("test_job") has been submitted',
''
]
task = TestJobTask(i=1, n_cpu=1, shared_tmp_dir='/tmp')
luigi.build([task], local_scheduler=True)
self.assertEqual(mock_check_output.call_count, 2)
def test_output_token(self):
"""
Output should be the given token path and the class name
"""
t = TestShellScriptTask(self.token_path)
self.t = t
luigi.build([t], local_scheduler=True)
self.assertEquals('%s/%s' % (self.token_path, t.__class__.__name__), t.output()[0].path)
def test_map_only(test_case):
job = MapOnlyJob(use_hdfs=test_case.use_hdfs)
luigi.build([job], local_scheduler=True)
c = []
for line in job.output().open("r"):
c.append(line.strip())
test_case.assertEqual(c[0], "kj")
test_case.assertEqual(c[4], "ljoi")
def test_inheritance(self):
t = PowerSum(lo=42, hi=45, p=2)
luigi.build([t], local_scheduler=True)
self.assertEqual(t.s, 42**2 + 43**2 + 44**2)
yield bss2_handle
if __name__ == "__main__":
# Set up the command line parameters
PARSER = argparse.ArgumentParser(
description="BS Seeker 2 Peak Caller Pipeline Wrapper")
PARSER.add_argument("--genome_fa", help="")
PARSER.add_argument("--genome_idx", help="")
PARSER.add_argument("--aligner", help="")
PARSER.add_argument("--aligner_path", help="")
PARSER.add_argument("--bss_path", help="")
PARSER.add_argument("--shared_tmp_dir", help="")
PARSER.add_argument("--python_path", default=sys.executable, help="")
# Get the matching parameters from the command line
ARGS = PARSER.parse_args()
SHARED_TMP_DIR = ARGS.shared_tmp_dir
luigi.build([
BSseeker2Indexer(genome_fa=ARGS.genome_fa,
genome_idx=ARGS.genome_idx,
aligner=ARGS.aligner,
aligner_path=ARGS.aligner_path,
bss_path=ARGS.bss_path,
user_python_path=ARGS.python_path)
],
local_scheduler=True,
workers=5)
import state_to_state_transitions2 as sst
import pandas as pd
import numpy as np
import luigi
class wrapper(luigi.WrapperTask):
def requires(self):
files = ['Session.csv', 'lead.csv', 'opportunity.csv', 'complete.csv']
task_list = []
for i in range(1, len(files)):
path = '/Users/emmanuels/Documents/AttributionData/Data/'
one = path + str(files[i - 1])
two = path + str(files[i])
task_list.append(
sst.state_to_state(first_file=one, second_file=two))
return task_list
def run(self):
print('Wrapper ran')
pd.DataFrame().to_csv(
'/Users/emmanuels/Documents/AttributionData/Data/wrangler1.csv')
def output(self):
return luigi.LocalTarget(
'/Users/emmanuels/Documents/AttributionData/Data/wrangler1.csv')
if __name__ == '__main__':
luigi.build([wrapper()], workers=8, local_scheduler=True)
def test_priority_w_dep(self):
x, y, z = PrioTask(25), PrioTask(15), PrioTask(5)
a, b, c = PrioTask(24), PrioTask(14), PrioTask(4)
luigi.build([a, b, c, x, y, z], local_scheduler=True)
self.assertTrue(z.t < y.t < x.t < c.t < b.t < a.t)
def run_tasks_for_dry_run(tasks_to_run):
for type in [
"failure",
"success",
"timeout",
"process_failure",
"processing_time",
"broken_task",
]:
os.makedirs(Path(constants.RESULTS_DIRECTORY) / type)
run_result = luigi.build(
tasks_to_run,
local_scheduler=True,
detailed_summary=True,
workers=10,
log_level='INFO',
)
for filename in glob('results/failure/*.json'):
result = json.loads(open(filename, 'r').read())
click.echo(
colorclass.Color("{red}" + result.get('task_type') +
" failed{/red}"))
click.echo(
f"{yaml.safe_dump({'parameters':result.get('task_params')})}")
click.echo("\n".join(result.get('exception_stack_trace')))
click.echo('')
exit_status_codes = {
LuigiStatusCode.SUCCESS: 0,
LuigiStatusCode.SUCCESS_WITH_RETRY: 0,
LuigiStatusCode.FAILED: 1,
LuigiStatusCode.FAILED_AND_SCHEDULING_FAILED: 2,
LuigiStatusCode.SCHEDULING_FAILED: 3,
LuigiStatusCode.NOT_RUN: 4,
LuigiStatusCode.MISSING_EXT: 5,
}
click.echo("Dry run results")
table_data = [
[
'Result', 'Launch', 'Account', 'Region', 'Current Version',
'New Version', 'Notes'
],
]
table = terminaltables.AsciiTable(table_data)
for filename in glob('output/TerminateProductDryRunTask/*.json'):
result = json.loads(open(filename, 'r').read())
table_data.append([
result.get('effect'),
result.get('params').get('launch_name'),
result.get('params').get('account_id'),
result.get('params').get('region'),
result.get('current_version'),
result.get('new_version'),
result.get('notes'),
])
for filename in glob('output/ProvisionProductDryRunTask/*.json'):
result = json.loads(open(filename, 'r').read())
table_data.append([
result.get('effect'),
result.get('params').get('launch_name'),
result.get('params').get('account_id'),
result.get('params').get('region'),
result.get('current_version'),
result.get('new_version'),
result.get('notes'),
])
click.echo(table.table)
sys.exit(exit_status_codes.get(run_result.status))
def run(cls):
cls.logger.info('Running Systems Checks...')
return luigi.build([SystemCheckTask()], local_scheduler=True)
def main(config):
luigi.build([groupEmbTags(config=config)])
combinations = {
'experiment': 'FF',
'setups': [setup],
'iterations': [50000, 100000, 150000],
'samples': [validation],
'thresholds': [0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9],
'merge_functions': ['mean_aff', 'max_aff'],
'init_with_maxs': [False],
'custom_fragments': True,
'histogram_quantiless': [False],
'discrete_queues': [True],
'keep_segmentation': True,
'dilate_mask': 0,
'mask_fragments': True
}
range_keys = [
'setups', 'iterations', 'samples', 'merge_functions',
'init_with_maxs', 'histogram_quantiless', 'discrete_queues'
]
jobs.append(EvaluateCombinations(combinations, range_keys))
set_base_dir(os.path.abspath('..'))
luigi.build(
jobs,
workers=50,
scheduler_host='slowpoke1.int.janelia.org',
logging_conf_file='/groups/saalfeld/home/funkej/.luigi/logging.conf')
def test_Ok(self):
task = TestTask()
luigi.build([task], local_scheduler=True)
from luigi import build
import bqe_search
from bqe_search import combined_search
build(
[
combined_search(gene_list=["p53", "CCND1"],
combination_terms=["apoptosis", "cell cycle"])
],
local_scheduler=True,
)
def run(tasks,
forced=None,
forced_all=False,
forced_all_upstream=False,
confirm=True,
workers=1,
abort=True,
**kwargs):
"""
Run tasks locally. See luigi.build for additional details
Args:
tasks (obj, list): task or list of tasks
forced (list): list of forced tasks
forced_all (bool): force all tasks
forced_all_upstream (bool): force all tasks including upstream
confirm (list): confirm invalidating tasks
workers (int): number of workers
abort (bool): on errors raise exception
kwargs: keywords to pass to luigi.build
"""
if not isinstance(tasks, (list, )):
tasks = [tasks]
if forced_all:
forced = tasks
if forced_all_upstream:
for t in tasks:
invalidate_upstream(t, confirm=confirm)
if forced is not None:
if not isinstance(forced, (list, )):
forced = [forced]
invalidate = []
for tf in forced:
for tup in tasks:
invalidate.append(d6tflow.taskflow_downstream(tf, tup))
invalidate = set().union(*invalidate)
invalidate = {t for t in invalidate if t.complete()}
if len(invalidate) > 0:
if confirm:
print('Forced tasks', invalidate)
c = input('Confirm invalidating forced tasks (y/n)')
else:
c = 'y'
if c == 'y':
[t.invalidate(confirm=False) for t in invalidate]
else:
return None
opts = {
**{
'workers': workers,
'local_scheduler': True,
'log_level': d6tflow.settings.log_level
},
**kwargs
}
result = luigi.build(tasks, **opts)
if abort and not result:
raise RuntimeError('Exception found running flow, check trace')
return result
'Args': [
"spark-submit",
"s3a://tar7/agg.py",
]
}
}])
#falta un paso para que espere a que terminen los steps
#client.terminate_job_flows(JobFlowIds=[clusters['Clusters'][0]['Id']])
while (time.time() - start_time < 360):
flag = 1
clusters = client.list_clusters()
if (clusters['Clusters'][0]['Status']['State'] in ['TERMINATING']):
#client.terminate_job_flows(JobFlowIds=[clusters['Clusters'][0]['Id']])
print('cluster termino')
else:
sys.exit('tiempo mayor')
os.system("aws s3 cp --recursive s3://tar7/promedio.parquet ./")
# AQUI ADENTRO VA EL CODIGO -------------------------------------------------
luigi.build([StartPipeline()])
def predict(device_mapping, target):
max_jobs = len(device_mapping)
tmp_folder = './tmp_inference'
input_path = '/g/kreshuk/data/FIB25/data.n5'
output_path = input_path
roi_begin = roi_end = None
in_key = 'volumes/raw/s0'
out_key = {'volumes/affinities/s0': (0, 3)}
mask_path = input_path
mask_key = 'volumes/masks/minfilter/s5'
config_folder = './config_inference'
if not os.path.exists(config_folder):
os.mkdir(config_folder)
input_blocks = (192, ) * 3
# remove (16, 16, 16) pixels from each side in the output
output_blocks = (160, ) * 3
halo = [(ib - ob) // 2 for ib, ob in zip(input_blocks, output_blocks)]
print("Found halo", halo)
shebang = "#! /g/kreshuk/pape/Work/software/conda/miniconda3/envs/torch/bin/python"
global_config = InferenceLocal.default_global_config()
global_config.update({
'shebang': shebang,
'block_shape': output_blocks,
'roi_begin': roi_begin,
'roi_end': roi_end
})
with open(os.path.join(config_folder, 'global.config'), 'w') as f:
json.dump(global_config, f)
config = InferenceLocal.default_task_config()
config.update({
'chunks': [ob // 2 for ob in output_blocks],
'mem_limit': 32,
'time_limit': 720,
'threads_per_job': 4,
'set_visible_device': False
})
with open(os.path.join(config_folder, 'inference.config'), 'w') as f:
json.dump(config, f)
ckpt = '/g/kreshuk/matskevych/boundary_map_prediction/project_folder_old/Weights'
task = InferenceLocal if target == 'local' else InferenceSlurm
t = task(tmp_folder=tmp_folder,
max_jobs=max_jobs,
config_dir=config_folder,
input_path=input_path,
input_key=in_key,
output_path=output_path,
output_key=out_key,
mask_path=mask_path,
mask_key=mask_key,
checkpoint_path=ckpt,
framework='inferno',
halo=halo)
ret = luigi.build([t], local_scheduler=True)
assert ret, "Failure"
import luigi
import os
from OrthoEvol.Tools.sge import SGEPipelineTask
# TIP Works on linux
logger = logging.getLogger('luigi-interface')
SGEPipelineTask.shared_tmp_dir = os.getcwd()
SGEPipelineTask.parallel_env = None
class TestPipelineTask(SGEPipelineTask):
"""Example pipeline task."""
i = luigi.Parameter()
def work(self): # Use work instead of run to DEBUG
logger.info('Running test job...')
with open(self.output().path, 'w') as f:
f.write('This is a test job.')
f.close()
def output(self):
return luigi.LocalTarget(
path=os.path.join(os.getcwd(), 'testjob_' + str(self.i) + '.txt'))
if __name__ == '__main__':
tasks = [TestPipelineTask(i=str(i), select=i + 1) for i in range(3)]
luigi.build(tasks, local_scheduler=True, workers=3)
def test_priority(self):
p, q, r = PrioTask(1), PrioTask(2), PrioTask(3)
luigi.build([p, q, r], local_scheduler=True)
self.assertTrue(r.t < q.t < p.t)
def test_invoke(self):
luigi.build([Popularity(datetime.date(2009, 1, 5))],
local_scheduler=True)
self.assertEqual(
MockTarget.fs.get_data('/tmp/popularity/2009-01-05.txt'), b'4\n')
from sqlalchemy import String
import luigi
from luigi.contrib import sqla
class SQLATask(sqla.CopyToTable):
# columns defines the table schema, with each element corresponding
# to a column in the format (args, kwargs) which will be sent to
# the sqlalchemy.Column(*args, **kwargs)
reflect = True
connection_string = "sqlite://" # in memory SQLite database
table = "Author" # name of the table to store data
columns = [(["First_Name", String(64)], {}), (["Last_Name",
String(64)], {})]
def rows(self):
for row in [("First_Name", "Babacar"), ("Last_Name", "Kane")]:
yield row
if __name__ == '__main__':
task = SQLATask()
luigi.build([task], local_scheduler=True)
import luigi
import os
class FileTarget(luigi.Target):
def __init__(self, filename):
self.filename = filename
def exists(self):
return os.path.isfile(self.filename)
class HelloTask(luigi.Task):
def run(self):
with open("hello.txt", "w") as f:
f.write("Hello Mario!")
def output(self):
return FileTarget("hello.txt")
if __name__ == "__main__":
luigi.build([HelloTask()])
def downscale_seg(sample, max_jobs=8, target='local'):
input_path = '/g/kreshuk/data/cremi/realigned/sample%s_small.n5' % sample
input_key = 'segmentation/multicut/s0'
config_dir = './config_ds_seg'
tmp_folder = './tmp_ds_seg_%s' % sample
try:
os.mkdir(config_dir)
except OSError:
pass
config = DownscalingWorkflow.get_config()
global_config = config['global']
shebang = '#! /g/kreshuk/pape/Work/software/conda/miniconda3/envs/cluster_env/bin/python'
global_config.update({'shebang': shebang})
with open(os.path.join(config_dir, 'global.config'), 'w') as f:
json.dump(global_config, f)
ds_config = config['downscaling']
# FIXME majority vote downscaling is broken
# ds_config.update({'library': 'skimage', 'threads_per_job': 8})
ds_config.update({
'library': 'vigra',
'library_kwargs': {
'order': 0
},
'threads_per_job': 8
})
with open(os.path.join(config_dir, 'downscaling.config'), 'w') as f:
json.dump(ds_config, f)
scale_factors = [[1, 2, 2], [1, 2, 2], [1, 2, 2], 2]
halos = [[0, 10, 10], [0, 10, 10], [0, 10, 10], [10, 10, 10]]
task = DownscalingWorkflow(tmp_folder=tmp_folder,
max_jobs=1,
config_dir=config_dir,
target='local',
input_path=input_path,
input_key='segmentation/multicut/s0',
output_key_prefix='segmentation/multicut',
scale_factors=scale_factors,
halos=halos)
success = luigi.build([task], local_scheduler=True)
#
if success and target == 'local':
with z5py.File(input_path) as f:
#
ds = f['raw/s2']
ds.n_threads = 8
raw = ds[:]
rshape = raw.shape
#
ds = f['segmentation/multicut/s2']
ds.n_threads = 8
seg = ds[:]
mshape = seg.shape
assert mshape == rshape, "%s %s" % (str(mshape), str(rshape))
view([raw, seg])
def test_run(self, os_mock):
t = MortarSqoopTaskTest(path=S3_PATH)
luigi.build([t], local_scheduler=True)
self.assertEquals(EXPECTED_ARGV, t.argv)
self.assertEquals(os.environ['AWS_ACCESS_KEY'], AWS_ACCESS_KEY)
self.assertEquals(os.environ['AWS_SECRET_KEY'], AWS_SECRET_KEY)
class HelloTask(luigi.Task):
to_whom = luigi.Parameter()
def run(self):
print("thinking about what to say...")
time.sleep(5)
with open("hello_%s.txt"%self.to_whom, "w") as f:
f.write("Hello %s!"%self.to_whom)
def output(self):
return FileTarget("hello_%s.txt"%self.to_whom)
class ReplyTask(luigi.Task):
def run(self):
print("thinking about what to say...")
time.sleep(5)
with open("reply.txt", "w") as f:
f.write("Hello Luigi!")
def output(self):
return FileTarget("reply.txt")
def requires(self):
return HelloTask(to_whom="Luigi")
if __name__ == "__main__":
luigi.build([ReplyTask()])
def test_run_options_with_jdbc_jar(self, os_mock):
t = MortarSqoopTaskTest(path=S3_PATH, jdbc_driver='some/path')
luigi.build([t], local_scheduler=True)
option_string = EXPECTED_ARGV + ['-j', 'some/path']
self.assertEqual(option_string, t.argv)
def test_recursion(self):
t = LinearSum(lo=42, hi=45)
luigi.build([t], local_scheduler=True)
self.assertEqual(t.s, 42 + 43 + 44)
def test_run_options_with_direct(self, os_mock):
t = MortarSqoopTaskTest(path=S3_PATH, direct=True)
luigi.build([t], local_scheduler=True)
option_string = EXPECTED_ARGV + ['--direct']
self.assertEqual(option_string, t.argv)
def test_run_job(self, mock_open, mock_communicate):
if on_lsf_master():
outfile = os.path.join(DEFAULT_HOME, 'testfile_1')
tasks = [TestJobTask(i=str(i), n_cpu_flag=1) for i in range(3)]
luigi.build(tasks, local_scheduler=True, workers=3)
self.assertTrue(os.path.exists(outfile))
def run_full_pipeline(project_abs_path, config_dir, config_file_name, local_scheduler_=True):
# Run pipeline this way
luigi.build([BuildSolution(project_abs_path, config_dir, config_file_name)], local_scheduler=local_scheduler_)
def test_build_internal(self):
luigi.build([Fib(100)], local_scheduler=True)
self.assertEqual(MockTarget.fs.get_data('/tmp/fib_10'), b'55\n')
self.assertEqual(MockTarget.fs.get_data('/tmp/fib_100'),
b'354224848179261915075\n')
def run_stacking_only(project_abs_path, config_dir, config_file_name, local_scheduler_=True):
# Run pipeline this way
luigi.build([MakeStackingPredictions(project_abs_path, config_dir, config_file_name)],
local_scheduler=local_scheduler_)
def handle(self, *args, **options):
build([LineDimLoad(subset=options["full"])], local_scheduler=True)
df_LineDim = pd.read_parquet("./data/linedim/part.0.parquet",
engine="fastparquet")
build([AccountDimLoad(subset=options["full"])], local_scheduler=True)
df_AccountDim = pd.read_parquet("./data/accountdim/part.0.parquet",
engine="fastparquet")
build([LimitFactLoad(subset=options["full"])], local_scheduler=True)
df_LimitFact = pd.read_parquet("./data/limitfact/part.0.parquet",
engine="fastparquet")
build([ByMdn()], local_scheduler=True)
local_root = "file://data/amplitude/by_mdn/"
target = ParquetTarget(local_root, flag=False, glob="*.parquet")
df_ActivityFact = target.read_dask(columns=["mdn", "event_time"],
parse_dates=["event_time"
]).compute()
LineDim.objects.all().delete()
with transaction.atomic():
df_LineDim_objs = [
LineDim(
MTN=line["MTN"],
Device_Grouping=line["DEVICE_GROUPING"],
Sales_Channel=line["SALES_CHANNEL"],
SVC_ACT_DT=line["SVC_ACT_DT"],
) for idx, line in df_LineDim.iterrows()
]
LineDim.objects.bulk_create(df_LineDim_objs)
AccountDim.objects.all().delete()
with transaction.atomic():
df_AcctDim_objs = [
AccountDim(
Cust_Acct=acct["CUST_ACCT"],
Segment_Name=acct["SEGMENT_NAME"],
SVC_Plan=acct["SVC_PLAN"],
) for idx, acct in df_AccountDim.iterrows()
]
AccountDim.objects.bulk_create(df_AcctDim_objs)
LimitFact.objects.all().delete()
with transaction.atomic():
df_LimitFact_objs = [
LimitFact(
MTN=LineDim.objects.get_or_create(MTN=limit["MTN"])[0],
Cust_Acct=AccountDim.objects.get_or_create(
Cust_Acct=limit["CUST_ACCT"])[0],
LIMIT_DT=limit["LIMITING_DT"],
LIMIT_TYPE=limit["LIMIT_TYPE"],
) for idx, limit in df_LimitFact.iterrows()
]
LimitFact.objects.bulk_create(df_LimitFact_objs)
ActivityFact.objects.all().delete()
with transaction.atomic():
df_ActivityFact_objs = [
ActivityFact(
MTN=LineDim.objects.get_or_create(MTN=activity["mdn"])[0],
EVENT_DT=activity["event_time"],
) for idx, activity in df_ActivityFact.iterrows()
]
ActivityFact.objects.bulk_create(df_ActivityFact_objs)
def mws_segmentation(offsets, path, input_key, fg_mask_key, output_key,
tmp_folder, target, max_jobs, stitch_mode):
task = MwsWorkflow
qos = 'normal'
config_folder = os.path.join(tmp_folder, 'configs')
os.makedirs(config_folder, exist_ok=True)
configs = task.get_config()
# we use a smaller block shape to speed up MWS
block_shape = [64, 256, 256]
conf = configs['global']
shebang = get_default_shebang()
block_shape = block_shape
conf.update({'shebang': shebang, 'block_shape': block_shape})
with open(os.path.join(config_folder, 'global.config'), 'w') as f:
json.dump(conf, f)
# write config for mws block task
strides = [4, 4, 4]
conf = configs['mws_blocks']
conf.update({
'randomize_strides': True,
'strides': strides,
'mem_limit': 12,
'time_limit': 900
})
with open(os.path.join(config_folder, 'mws_blocks.config'), 'w') as f:
json.dump(conf, f)
# determine config for the given stitching mode
if stitch_mode == '':
stitch_mc = False
elif stitch_mode == 'biased':
stitch_mc = True
beta1, beta2 = 0.5, 0.75
elif stitch_mode == 'unbiased':
stitch_mc = True
beta1 = beta2 = 0.5
if stitch_mc:
# write config for stitching multicut
conf = configs['stitching_multicut']
conf.update({'beta1': beta1, 'beta2': beta2, 'qos': qos})
with open(os.path.join(config_folder, 'stitching_multicut.config'),
'w') as f:
json.dump(conf, f)
conf = configs['write']
conf.update({'mem_limit': 8, 'time_limit': 120, 'qos': qos})
with open(os.path.join(config_folder, 'write.config'), 'w') as f:
json.dump(conf, f)
# write config for edge feature task
conf = configs['block_edge_features']
conf.update({'offsets': offsets, 'mem_limit': 4, 'qos': qos})
with open(os.path.join(config_folder, 'block_edge_features.config'),
'w') as f:
json.dump(conf, f)
conf_names = [
'merge_edge_features', 'merge_sub_graphs', 'map_edge_ids',
'simple_stitch_assignments'
]
for name in conf_names:
conf = configs[name]
conf.update({
'mem_limit': 128,
'time_limit': 240,
'threads_per_job': 16,
'qos': qos
})
with open(os.path.join(config_folder, '%s.config' % name),
'w') as f:
json.dump(conf, f)
conf = configs['stitching_multicut']
# set time limit for the multicut task to 18 hours (in minutes)
tlim_task = 18 * 60
# set time limit for the solver to 16 hours (in seconds)
tlim_solver = 16 * 60 * 60
conf.update({
'mem_limit': 256,
'time_limit': tlim_task,
'threads_per_job': 16,
'qos': qos,
'agglomerator': 'greedy-additive',
'time_limit_solver': tlim_solver
})
with open(os.path.join(config_folder, 'stitching_multicut.config'),
'w') as f:
json.dump(conf, f)
t = task(tmp_folder=tmp_folder,
config_dir=config_folder,
max_jobs=max_jobs,
target=target,
input_path=path,
input_key=input_key,
output_path=path,
output_key=output_key,
mask_path=path,
mask_key=fg_mask_key,
stitch_via_mc=stitch_mc,
offsets=offsets)
ret = luigi.build([t], local_scheduler=True)
if not ret:
raise RuntimeError("Mws segmentation failed")
