def set_up_dask(n_jobs=2):
MPControl.set_multiprocess_engine("dask-cluster")
MPControl.client.use_default_configuration("rusty_ccb", n_jobs=n_jobs)
MPControl.client.add_worker_conda(
"source ~/.local/anaconda3/bin/activate inferelator")
MPControl.client.add_slurm_command_line("--constraint=broadwell")
MPControl.connect()
def setUpClass(cls):
if not MPControl.is_initialized:
MPControl.set_multiprocess_engine("local")
MPControl.connect()
DotProduct.set_mkl(True)
def initialize_multiprocessing(self):
"""
Register the multiprocessing controller if set and run .connect()
"""
if self.multiprocessing_controller is not None:
MPControl.set_multiprocess_engine(self.multiprocessing_controller)
MPControl.connect()
def setUpClass(cls):
cls.tempdir = tempfile.mkdtemp()
MPControl.shutdown()
MPControl.set_multiprocess_engine(cls.name)
# Create a wrapper for LocalCluster so that the HPC controller can be tested locally
# And then bind it so that it works in py27 right
def fake_cluster(*args, **kwargs):
replace_args = dict()
replace_args["n_workers"] = kwargs.pop("n_workers", 0)
replace_args["threads_per_worker"] = kwargs.pop(
"threads_per_worker", 1)
replace_args["processes"] = kwargs.pop("processes", True)
replace_args["local_dir"] = kwargs.pop("local_directory", None)
clust = distributed.LocalCluster(**replace_args)
clust._active_worker_n = 0
def _count_active_workers(self):
val = self._active_worker_n
self._active_worker_n += 1
return val
clust._count_active_workers = types.MethodType(
_count_active_workers, clust)
return clust
MPControl.client._cluster_controller_class = types.MethodType(
fake_cluster, MPControl.client)
MPControl.client.set_job_size_params(n_jobs=1,
n_cores_per_job=1,
mem_per_job="1gb")
MPControl.client._local_directory = cls.tempdir
MPControl.connect()
def start_mpcontrol_dask(n_cores=N_CORES):
utils.Debug.set_verbose_level(1)
MPControl.set_multiprocess_engine("dask-cluster")
MPControl.client.minimum_cores = n_cores
MPControl.client.maximum_cores = n_cores
MPControl.client.walltime = '48:00:00'
MPControl.client.add_worker_env_line('module load slurm')
MPControl.client.add_worker_env_line('module load gcc/8.3.0')
MPControl.client.add_worker_env_line('source ' + CONDA_ACTIVATE_PATH)
MPControl.client.cluster_controller_options.append("-p ccb")
MPControl.connect()
def test_cant_change(self):
with self.assertRaises(RuntimeError):
MPControl.set_multiprocess_engine("local")
DATA_FILE = "YPD_LOCAL_OPT_PV_DEXFORM.h5ad"
CONDA_ACTIVATE_PATH = '~/.local/anaconda3/bin/activate'
EXPRESSION_MATRIX_METADATA = ['Genotype', 'Genotype_Group', 'Replicate', 'Condition', 'tenXBarcode']
YEASTRACT_PRIOR = "YEASTRACT_20190713_BOTH.tsv"
TF_NAMES = "tf_names_gold_standard.txt"
YEASTRACT_TF_NAMES = "tf_names_yeastract.txt"
INPUT_DIR = '/mnt/ceph/users/cjackson/inferelator/data/yeast'
OUTPUT_PATH = '/mnt/ceph/users/cjackson/jackson_2019_inferelator_v050/'
if __name__ == '__main__':
MPControl.set_multiprocess_engine("dask-cluster")
MPControl.client.use_default_configuration("rusty_preempt")
MPControl.client.set_job_size_params(n_jobs=1)
MPControl.client.add_worker_conda("source ~/.local/anaconda3/bin/activate inferelator")
MPControl.connect()
wkf = inferelator_workflow("stars", VelocityWorkflow)
wkf.set_file_paths(input_dir=INPUT_DIR,
output_dir=os.path.join(OUTPUT_PATH, "tau43"),
gold_standard_file='gold_standard.tsv',
priors_file=YEASTRACT_PRIOR,
tf_names_file=YEASTRACT_TF_NAMES)
wkf.set_expression_file(h5ad=DATA_FILE, h5_layer="smooth_count")
wkf.set_velocity_parameters(velocity_file_name=DATA_FILE, velocity_file_type="h5ad", velocity_file_layer="pv")
wkf.set_crossvalidation_parameters(split_gold_standard_for_crossvalidation=True,
cv_split_ratio=0.5)
def tearDownClass(cls):
MPControl.shutdown()
MPControl.set_multiprocess_engine("local")
MPControl.connect()
if cls.tempdir is not None:
shutil.rmtree(cls.tempdir)
def setUpClass(cls):
cls.tempdir = tempfile.mkdtemp()
MPControl.shutdown()
MPControl.set_multiprocess_engine("dask-local")
MPControl.connect(local_dir=cls.tempdir, n_workers=1, processes=False)
def set_up_dask(n_jobs=1):
MPControl.set_multiprocess_engine("dask-cluster")
MPControl.client.use_default_configuration("rusty_rome", n_jobs=n_jobs)
MPControl.client.add_worker_conda("source ~/.local/anaconda3/bin/activate inferelator")
MPControl.client.set_job_size_params(walltime="168:00:00")
MPControl.connect()
def setUpClass(cls):
cls.temp_dir = tempfile.mkdtemp()
cls.server = kvsstcp.KVSServer("", 0)
MPControl.shutdown()
MPControl.set_multiprocess_engine(cls.name)
MPControl.connect(host=cls.server.cinfo[0], port=cls.server.cinfo[1])
PRIORS_FILE_NAME = 'gold_standard.tsv.gz'
GOLD_STANDARD_FILE_NAME = 'gold_standard.tsv.gz'
META_DATA_FILE_NAME = 'meta_data.tsv'
TF_LIST_FILE_NAME = 'tf_names.tsv'
CV_SEEDS = list(range(42, 52))
# Multiprocessing uses the pathos implementation of multiprocessing (with dill instead of cPickle)
# This is suited for a single computer but will not work on a distributed cluster
n_cores_local = 10
local_engine = True
# Multiprocessing needs to be protected with the if __name__ == 'main' pragma
if __name__ == '__main__' and local_engine:
MPControl.set_multiprocess_engine("multiprocessing")
MPControl.client.set_processes(n_cores_local)
MPControl.connect()
# Define the general run parameters
def set_up_workflow(wkf):
wkf.set_file_paths(input_dir=DATA_DIR,
output_dir=OUTPUT_DIR,
tf_names_file=TF_LIST_FILE_NAME,
meta_data_file=META_DATA_FILE_NAME,
priors_file=PRIORS_FILE_NAME,
gold_standard_file=GOLD_STANDARD_FILE_NAME)
wkf.set_expression_file(tsv=EXPRESSION_FILE_NAME)
wkf.set_file_properties(expression_matrix_columns_are_genes=False)
wkf.set_run_parameters(num_bootstraps=5)
def test_bad_engine_II(self):
with self.assertRaises(ValueError):
MPControl.set_multiprocess_engine(unittest.TestCase)
def test_bad_engine(self):
with self.assertRaises(ValueError):
MPControl.set_multiprocess_engine("V8")
def tearDownClass(cls):
MPControl.shutdown()
MPControl.set_multiprocess_engine("local")
MPControl.connect()
def setUpClass(cls):
MPControl.shutdown()
MPControl.set_multiprocess_engine(cls.name)
MPControl.connect()
def setUpClass(cls):
cls.tempdir = tempfile.mkdtemp()
MPControl.shutdown()
MPControl.set_multiprocess_engine(cls.name)
MPControl.connect(local_dir=cls.tempdir, n_workers=1)
from inferelator.distributed.inferelator_mp import MPControl
MPControl.set_multiprocess_engine("local")
MPControl.connect()
def start_mpcontrol_dask(n_cores=N_CORES):
MPControl.set_multiprocess_engine("dask-cluster")
MPControl.client.use_default_configuration("rusty_ccb", n_jobs=2)
MPControl.client.add_worker_conda(
"source /mnt/ceph/users/sysbio/.anaconda3/bin/activate inferelator")
MPControl.connect()