def submit_benchmark(benchmark_id,
num_tasks,
fast=False,
use_database=False,
complete_run=False,
keep_old_data=False,
use_native_structure=False):
'''Submit the benchmark to the cluster.'''
qsub_command = 'qsub',
benchmark_command = (
'loop_benchmark.py', str(benchmark_id),
'--use-database' if use_database else '--not-use-database',
'--complete-run' if complete_run else '--not-complete_run',
'--use-native-structure'
if use_native_structure else '--not-use-native-structure')
if fast:
qsub_command += '-t', '1-{0}'.format(num_tasks)
qsub_command += '-l', 'h_rt=0:30:00'
else:
qsub_command += '-t', '1-{0}'.format(num_tasks)
qsub_command += '-l', 'h_rt=24:00:00'
if not keep_old_data:
utilities.clear_directory('job_output')
qsub_command += '-o', 'job_output', '-e', 'job_output'
qsub_cwd = os.path.dirname(__file__)
subprocess.call(qsub_command + benchmark_command, cwd=qsub_cwd)
def resume_benchmark(benchmark_id, nstruct=None):
qsub_command = 'qsub',
benchmark_command = 'loop_benchmark.py', benchmark_id
# You get weird errors if you forget to cast nstruct from string to int.
if nstruct is not None: nstruct = int(nstruct)
# Read the job parameters from the database.
with database.connect() as session:
benchmark = session.query(database.Benchmarks).get(benchmark_id)
num_pdbs = len(benchmark.input_pdbs)
# Make sure the right version of rosetta is being used.
git_commit = subprocess.check_output(
shlex.split('git rev-parse HEAD'),
cwd=settings.rosetta).strip()
git_diff = subprocess.check_output(
shlex.split('git diff'),
cwd=settings.rosetta).strip()
if benchmark.git_commit != git_commit:
message = "Benchmark \"{0}\" was run with rosetta commit #{1}, but commit #{2} is currently checked out. Press [Ctrl-C] to abort or [Enter] to continue."
message = textwrap.fill(message.format(benchmark.id, benchmark.git_commit[:8], git_commit[:8]))
raw_input(message)
elif benchmark.git_diff != git_diff:
message = "Uncommitted changes have been made to rosetta since benchmark \"{0}\" was run. Press [Ctrl-C] to abort or [Enter] to continue."
message = textwrap.fill(message.format(benchmark.id))
raw_input(message)
# Build the qsub command.
if benchmark.fast:
qsub_command += '-t', '1-{0}'.format((nstruct or 10) * num_pdbs)
qsub_command += '-l', 'h_rt=0:30:00'
else:
qsub_command += '-t', '1-{0}'.format((nstruct or 500) * num_pdbs)
qsub_command += '-l', 'h_rt=4:00:00'
print "Your benchmark \"{0}\" (id={1}) is being resumed".format(
benchmark.name, benchmark_id)
# Submit the job.
utilities.clear_directory('job_output')
qsub_command += '-o', 'job_output', '-e', 'job_output'
subprocess.call(qsub_command + benchmark_command)
def setup_latex_dir(self):
"""
Create the directory where all the *.tex and *.pdf files generated for
this report will be stored. This method is meant to be used in a
'with' statement, so that it can automatically clean up the directory
once the program terminates (either normally or via an exception).
"""
import shutil
import tempfile
try:
# Create the root latex directory.
if self.latex_dir is not None:
utilities.clear_directory(self.latex_dir)
else:
self.latex_dir = tempfile.mkdtemp('.loop_benchmark')
self.latex_dir = os.path.abspath(self.latex_dir)
# Create subdirectories for every benchmark and loop.
for benchmark in self:
benchmark.latex_dir = os.path.join(
self.latex_dir,
benchmark.name + '_analysis',
)
os.mkdir(benchmark.latex_dir)
for loop in benchmark:
loop.latex_dir = os.path.join(
benchmark.latex_dir,
loop.pdb_id,
)
os.mkdir(loop.latex_dir)
# Yield control to the rest of the program.
yield
finally:
# Remove the latex directory if the user doesn't want to keep it.
if not self.keep_latex and self.latex_dir is not None:
shutil.rmtree(self.latex_dir)
def setup_latex_dir(self):
"""
Create the directory where all the *.tex and *.pdf files generated for
this report will be stored. This method is meant to be used in a
'with' statement, so that it can automatically clean up the directory
once the program terminates (either normally or via an exception).
"""
import shutil
import tempfile
try:
# Create the root latex directory.
if self.latex_dir is not None:
utilities.clear_directory(self.latex_dir)
else:
self.latex_dir = tempfile.mkdtemp('.loop_benchmark')
self.latex_dir = os.path.abspath(self.latex_dir)
# Create subdirectories for every benchmark and loop.
for benchmark in self:
benchmark.latex_dir = os.path.join(
self.latex_dir,
benchmark.name + '_analysis',
)
os.mkdir(benchmark.latex_dir)
for loop in benchmark:
loop.latex_dir = os.path.join(
benchmark.latex_dir,
loop.pdb_id,
)
os.mkdir(loop.latex_dir)
# Yield control to the rest of the program.
yield
finally:
# Remove the latex directory if the user doesn't want to keep it.
if not self.keep_latex and self.latex_dir is not None:
shutil.rmtree(self.latex_dir)
def resume_benchmark(benchmark_id, nstruct=None, use_database=False):
# This feature has not been implemented on the disk platfrom
if not use_database:
raise Exception(
"resume benchmark has not been implemented on the disk platform!")
qsub_command = 'qsub',
benchmark_command = ('loop_benchmark.py', benchmark_id, '--use-database'
if use_database else '--not-use-database',
'--not-complete-run')
# You get weird errors if you forget to cast nstruct from string to int.
if nstruct is not None: nstruct = int(nstruct)
# Read the job parameters from the saved data
data_controller = DataController(
'database') if use_database else DataController('disk')
benchmark_define_dict = data_controller.get_benchmark_define_dict(
benchmark_id)
num_pdbs = len(benchmark_define_dict['input_pdbs'])
# Make sure the right version of rosetta is being used.
git_commit = subprocess.check_output(shlex.split('git rev-parse HEAD'),
cwd=settings.rosetta).strip()
git_diff = subprocess.check_output(shlex.split('git diff'),
cwd=settings.rosetta).strip()
if benchmark_define_dict['git_commit'] != git_commit:
message = "Benchmark \"{0}\" was run with rosetta commit #{1}, but commit #{2} is currently checked out. Press [Ctrl-C] to abort or [Enter] to continue."
message = textwrap.fill(
message.format(benchmark_define_dict['id'],
benchmark_define_dict['git_commit'][:8],
git_commit[:8]))
raw_input(message)
elif benchmark_define_dict['git_diff'] != git_diff:
message = "Uncommitted changes have been made to rosetta since benchmark \"{0}\" was run. Press [Ctrl-C] to abort or [Enter] to continue."
message = textwrap.fill(message.format(benchmark_define_dict['id']))
raw_input(message)
# Build the qsub command.
if benchmark_define_dict['fast']:
qsub_command += '-t', '1-{0}'.format((nstruct or 10) * num_pdbs)
qsub_command += '-l', 'h_rt=24:00:00'
else:
qsub_command += '-t', '1-{0}'.format((nstruct or 500) * num_pdbs)
qsub_command += '-l', 'h_rt=24:00:00'
print "Your benchmark \"{0}\" (id={1}) is being resumed".format(
benchmark_define_dict['name'], benchmark_id)
# Submit the job.
utilities.clear_directory('job_output')
qsub_command += '-o', 'job_output', '-e', 'job_output'
subprocess.call(qsub_command + benchmark_command)
# Submit the benchmark to the cluster.
qsub_command = 'qsub',
benchmark_command = 'loop_benchmark.py', benchmark_id
if fast:
qsub_command += '-t', '1-{0}'.format(nstruct * len(pdbs))
qsub_command += '-l', 'h_rt=0:30:00'
else:
qsub_command += '-t', '1-{0}'.format(nstruct * len(pdbs))
qsub_command += '-l', 'h_rt=6:00:00'
if not arguments['--keep-old-data']:
utilities.clear_directory('job_output')
qsub_command += '-o', 'job_output', '-e', 'job_output'
qsub_cwd = os.path.dirname(__file__)
subprocess.call(qsub_command + benchmark_command, cwd=qsub_cwd)
def resume_benchmark(benchmark_id, nstruct=None):
qsub_command = 'qsub',
benchmark_command = 'loop_benchmark.py', benchmark_id
# You get weird errors if you forget to cast nstruct from string to int.
if nstruct is not None: nstruct = int(nstruct)
# Read the job parameters from the database.