def _clear_job_results(*, jobs, incomplete_only=True):
for job in jobs:
val = mt.getValue(key=job)
if val:
if (not incomplete_only) or (val.startswith('in-process')) or (
val.startswith('error')):
print('Clearing job: ' + job['label'])
mt.setValue(key=job, value=None)
def _download_recordings(*, jobs):
for _, job in enumerate(jobs):
val = mt.getValue(key=job)
if not val:
if 'recording' in job:
if 'directory' in job['recording']:
dsdir = job['recording']['directory']
fname = dsdir + '/raw.mda'
print('REALIZING FILE: ' + fname)
mt.realizeFile(path=fname)
def _run_job(job):
val = mt.getValue(key=job)
if val:
return
code = ''.join(random.choice(string.ascii_uppercase) for x in range(10))
if not mt.setValue(key=job, value='in-process-' + code, overwrite=False):
return
status = dict(time_started=_make_timestamp(), status='running')
_set_job_status(job, status)
print('Running job: ' + job['label'])
try:
result = _do_run_job(job)
except:
status['time_finished'] = _make_timestamp()
status['status'] = 'error'
status['error'] = 'Exception in _do_run_job'
val = mt.getValue(key=job)
if val == 'in-process-' + code:
_set_job_status(job, status)
raise
val = mt.getValue(key=job)
if val != 'in-process-' + code:
print(
'Not saving result because in-process code does not match {} <> {}.'
.format(val, 'in-process-' + code))
return
status['time_finished'] = _make_timestamp()
status['result'] = result
if 'error' in result:
print('Error running job: ' + result['error'])
status['status'] = 'error'
status['error'] = result['error']
_set_job_status(job, status)
mt.setValue(key=job, value='error-' + code)
return
status['status'] = 'finished'
mt.saveObject(
key=job, object=result
) # Not needed in future, because we should instead use the status object
def initialize(self):
#self._resource_names = ca.getSubKeys(key=dict(name='spikeforest_results'))
obj = mt.getValue(key=dict(name='compute_resources'),
subkey='-',
parse_json=True)
if obj:
self._resource_names = list(obj.keys())
else:
self._resource_names = ['none-found']
self._SEL_resource_name.setOptions([''] + self._resource_names)
self._on_resource_name_changed(value=self._SEL_resource_name.value())
def main():
parser = argparse.ArgumentParser(
description=
'Generate unit detail data (including spikesprays) for website')
parser.add_argument(
'--output_ids',
help=
'Comma-separated list of IDs of the analysis outputs to include in the website.',
required=False,
default=None)
args = parser.parse_args()
use_slurm = True
mt.configDownloadFrom(['spikeforest.kbucket'])
if args.output_ids is not None:
output_ids = args.output_ids.split(',')
else:
output_ids = [
'paired_boyden32c',
'paired_crcns',
'paired_mea64c',
'paired_kampff',
'paired_monotrode',
'synth_monotrode',
# 'synth_bionet',
'synth_magland',
'manual_franklab',
'synth_mearec_neuronexus',
'synth_mearec_tetrode',
'synth_visapy',
'hybrid_janelia'
]
print('Using output ids: ', output_ids)
print(
'******************************** LOADING ANALYSIS OUTPUT OBJECTS...')
for output_id in output_ids:
slurm_working_dir = 'tmp_slurm_job_handler_' + _random_string(5)
job_handler = mlpr.SlurmJobHandler(working_dir=slurm_working_dir)
if use_slurm:
job_handler.addBatchType(name='default',
num_workers_per_batch=20,
num_cores_per_job=1,
time_limit_per_batch=1800,
use_slurm=True,
additional_srun_opts=['-p ccm'])
else:
job_handler.addBatchType(
name='default',
num_workers_per_batch=multiprocessing.cpu_count(),
num_cores_per_job=1,
use_slurm=False)
with mlpr.JobQueue(job_handler=job_handler) as JQ:
print('=============================================', output_id)
print('Loading output object: {}'.format(output_id))
output_path = (
'key://pairio/spikeforest/spikeforest_analysis_results.{}.json'
).format(output_id)
obj = mt.loadObject(path=output_path)
# studies = obj['studies']
# study_sets = obj.get('study_sets', [])
# recordings = obj['recordings']
sorting_results = obj['sorting_results']
print(
'Determining sorting results to process ({} total)...'.format(
len(sorting_results)))
sorting_results_to_process = []
for sr in sorting_results:
key = dict(name='unit-details-v0.1.0',
recording_directory=sr['recording']['directory'],
firings_true=sr['firings_true'],
firings=sr['firings'])
val = mt.getValue(key=key, collection='spikeforest')
if not val:
sr['key'] = key
sorting_results_to_process.append(sr)
print('Need to process {} of {} sorting results'.format(
len(sorting_results_to_process), len(sorting_results)))
recording_directories_to_process = sorted(
list(
set([
sr['recording']['directory']
for sr in sorting_results_to_process
])))
print('{} recording directories to process'.format(
len(recording_directories_to_process)))
print('Filtering recordings...')
filter_jobs = FilterTimeseries.createJobs([
dict(recording_directory=recdir,
timeseries_out={'ext': '.mda'},
_timeout=600)
for recdir in recording_directories_to_process
])
filter_results = [job.execute() for job in filter_jobs]
JQ.wait()
filtered_timeseries_by_recdir = dict()
for i, recdir in enumerate(recording_directories_to_process):
result0 = filter_results[i]
if result0.retcode != 0:
raise Exception(
'Problem computing filtered timeseries for recording: {}'
.format(recdir))
filtered_timeseries_by_recdir[recdir] = result0.outputs[
'timeseries_out']
print('Creating spike sprays...')
for sr in sorting_results_to_process:
rec = sr['recording']
study_name = rec['study']
rec_name = rec['name']
sorter_name = sr['sorter']['name']
print('====== COMPUTING {}/{}/{}'.format(
study_name, rec_name, sorter_name))
if sr.get('comparison_with_truth', None) is not None:
cwt = mt.loadObject(
path=sr['comparison_with_truth']['json'])
filtered_timeseries = filtered_timeseries_by_recdir[
rec['directory']]
spike_spray_job_objects = []
list0 = list(cwt.values())
for _, unit in enumerate(list0):
# print('')
# print('=========================== {}/{}/{} unit {} of {}'.format(study_name, rec_name, sorter_name, ii + 1, len(list0)))
# ssobj = create_spikesprays(rx=rx, sx_true=sx_true, sx_sorted=sx, neighborhood_size=neighborhood_size, num_spikes=num_spikes, unit_id_true=unit['unit_id'], unit_id_sorted=unit['best_unit'])
spike_spray_job_objects.append(
dict(args=dict(
recording_directory=rec['directory'],
filtered_timeseries=filtered_timeseries,
firings_true=os.path.join(
rec['directory'], 'firings_true.mda'),
firings_sorted=sr['firings'],
unit_id_true=unit['unit_id'],
unit_id_sorted=unit['best_unit'],
json_out={'ext': '.json'},
_container='default',
_timeout=180),
study_name=study_name,
rec_name=rec_name,
sorter_name=sorter_name,
unit=unit))
spike_spray_jobs = CreateSpikeSprays.createJobs(
[obj['args'] for obj in spike_spray_job_objects])
spike_spray_results = [
job.execute() for job in spike_spray_jobs
]
sr['spike_spray_job_objects'] = spike_spray_job_objects
sr['spike_spray_results'] = spike_spray_results
JQ.wait()
for sr in sorting_results_to_process:
rec = sr['recording']
study_name = rec['study']
rec_name = rec['name']
sorter_name = sr['sorter']['name']
print('====== SAVING {}/{}/{}'.format(study_name, rec_name,
sorter_name))
if sr.get('comparison_with_truth', None) is not None:
spike_spray_job_objects = sr['spike_spray_job_objects']
spike_spray_results = sr['spike_spray_results']
unit_details = []
ok = True
for i, result in enumerate(spike_spray_results):
obj0 = spike_spray_job_objects[i]
if result.retcode != 0:
print('WARNING: Error creating spike sprays for job:')
print(spike_spray_job_objects[i])
ok = False
break
ssobj = mt.loadObject(path=result.outputs['json_out'])
if ssobj is None:
raise Exception(
'Problem loading spikespray object output.')
address = mt.saveObject(object=ssobj,
upload_to='spikeforest.kbucket')
unit = obj0['unit']
unit_details.append(
dict(studyName=obj0['study_name'],
recordingName=obj0['rec_name'],
sorterName=obj0['sorter_name'],
trueUnitId=unit['unit_id'],
sortedUnitId=unit['best_unit'],
spikeSprayUrl=mt.findFile(
path=address,
remote_only=True,
download_from='spikeforest.kbucket'),
_container='default'))
if ok:
mt.saveObject(collection='spikeforest',
key=sr['key'],
object=unit_details,
upload_to='spikeforest.public')
def main():
from mountaintools import client as mt
parser = argparse.ArgumentParser(
description=
'Generate unit detail data (including spikesprays) for website')
parser.add_argument('analysis_path',
help='assembled analysis file (output.json)')
parser.add_argument(
'--studysets',
help='Comma-separated list of study set names to include',
required=False,
default=None)
parser.add_argument('--force-run',
help='Force rerunning of processing',
action='store_true')
parser.add_argument(
'--force-run-all',
help='Force rerunning of processing including filtering',
action='store_true')
parser.add_argument('--parallel',
help='Optional number of parallel jobs',
required=False,
default='0')
parser.add_argument('--slurm',
help='Path to slurm config file',
required=False,
default=None)
parser.add_argument('--cache',
help='The cache database to use',
required=False,
default=None)
parser.add_argument('--job-timeout',
help='Timeout for processing jobs',
required=False,
default=600)
parser.add_argument('--log-file',
help='Log file for analysis progress',
required=False,
default=None)
parser.add_argument(
'--force-regenerate',
help=
'Whether to force regenerating spike sprays (for when code has changed)',
action='store_true')
parser.add_argument('--test',
help='Whether to just test by running only 1',
action='store_true')
args = parser.parse_args()
mt.configDownloadFrom(['spikeforest.kbucket'])
with open(args.analysis_path, 'r') as f:
analysis = json.load(f)
if args.studysets is not None:
studyset_names = args.studysets.split(',')
print('Using study sets: ', studyset_names)
else:
studyset_names = None
study_sets = analysis['StudySets']
sorting_results = analysis['SortingResults']
studies_to_include = []
for ss in study_sets:
if (studyset_names is None) or (ss['name'] in studyset_names):
for study in ss['studies']:
studies_to_include.append(study['name'])
print('Including studies:', studies_to_include)
print('Determining sorting results to process ({} total)...'.format(
len(sorting_results)))
sorting_results_to_process = []
sorting_results_to_consider = []
for sr in sorting_results:
study_name = sr['studyName']
if study_name in studies_to_include:
if 'firings' in sr:
if sr.get('comparisonWithTruth', None) is not None:
sorting_results_to_consider.append(sr)
key = dict(name='unit-details-v0.1.0',
recording_directory=sr['recordingDirectory'],
firings_true=sr['firingsTrue'],
firings=sr['firings'])
val = mt.getValue(key=key, collection='spikeforest')
if (not val) or (args.force_regenerate):
sr['key'] = key
sorting_results_to_process.append(sr)
if args.test and len(sorting_results_to_process) > 0:
sorting_results_to_process = [sorting_results_to_process[0]]
print('Need to process {} of {} sorting results'.format(
len(sorting_results_to_process), len(sorting_results_to_consider)))
recording_directories_to_process = sorted(
list(
set([
sr['recordingDirectory'] for sr in sorting_results_to_process
])))
print('{} recording directories to process'.format(
len(recording_directories_to_process)))
if int(args.parallel) > 0:
job_handler = hither.ParallelJobHandler(int(args.parallel))
elif args.slurm:
with open(args.slurm, 'r') as f:
slurm_config = json.load(f)
job_handler = hither.SlurmJobHandler(working_dir='tmp_slurm',
**slurm_config['cpu'])
else:
job_handler = None
print('Filtering recordings...')
filter_results = []
with hither.config(container='default',
cache=args.cache,
force_run=args.force_run_all,
job_handler=job_handler,
log_path=args.log_file,
exception_on_fail=True,
cache_failing=False,
rerun_failing=True,
job_timeout=args.job_timeout), hither.job_queue():
for recdir in recording_directories_to_process:
result = filter_recording.run(recording_directory=recdir,
timeseries_out=hither.File())
filter_results.append(result)
filtered_timeseries_by_recdir = dict()
for i, recdir in enumerate(recording_directories_to_process):
result0 = filter_results[i]
if not result0.success:
raise Exception(
'Problem computing filtered timeseries for recording: {}'.
format(recdir))
filtered_timeseries_by_recdir[
recdir] = result0.outputs.timeseries_out._path
print('Creating spike sprays...')
with hither.config(container='default',
cache=args.cache,
force_run=args.force_run or args.force_run_all,
job_handler=job_handler,
log_path=args.log_file,
exception_on_fail=True,
cache_failing=False,
rerun_failing=True,
job_timeout=args.job_timeout), hither.job_queue():
for sr in sorting_results_to_process:
recdir = sr['recordingDirectory']
study_name = sr['studyName']
rec_name = sr['recordingName']
sorter_name = sr['sorterName']
print('====== COMPUTING {}/{}/{}'.format(study_name, rec_name,
sorter_name))
cwt = ka.load_object(path=sr['comparisonWithTruth']['json'])
filtered_timeseries = filtered_timeseries_by_recdir[recdir]
spike_spray_results = []
list0 = list(cwt.values())
for _, unit in enumerate(list0):
result = create_spike_sprays.run(
recording_directory=recdir,
filtered_timeseries=filtered_timeseries,
firings_true=os.path.join(recdir, 'firings_true.mda'),
firings_sorted=sr['firings'],
unit_id_true=unit['unit_id'],
unit_id_sorted=unit['best_unit'],
json_out=hither.File())
setattr(result, 'unit', unit)
spike_spray_results.append(result)
sr['spike_spray_results'] = spike_spray_results
for sr in sorting_results_to_process:
recdir = sr['recordingDirectory']
study_name = sr['studyName']
rec_name = sr['recordingName']
sorter_name = sr['sorterName']
print('====== SAVING {}/{}/{}'.format(study_name, rec_name,
sorter_name))
spike_spray_results = sr['spike_spray_results']
key = sr['key']
unit_details = []
ok = True
for i, result in enumerate(spike_spray_results):
if not result.success:
print(
'WARNING: Error creating spike sprays for {}/{}/{}'.format(
study_name, rec_name, sorter_name))
ok = False
break
ssobj = ka.load_object(result.outputs.json_out._path)
if ssobj is None:
raise Exception('Problem loading spikespray object output.')
address = mt.saveObject(object=ssobj,
upload_to='spikeforest.kbucket')
unit = getattr(result, 'unit')
unit_details.append(
dict(studyName=study_name,
recordingName=rec_name,
sorterName=sorter_name,
trueUnitId=unit['unit_id'],
sortedUnitId=unit['best_unit'],
spikeSprayUrl=mt.findFile(
path=address,
remote_only=True,
download_from='spikeforest.kbucket')))
if ok:
mt.saveObject(collection='spikeforest',
key=key,
object=unit_details,
upload_to='spikeforest.public')
def _test1(ii):
key = dict(test='key3')
val0 = '{}'.format(ii)
mt.setValue(key=key, value=val0)
val1 = mt.getValue(key=key)
return val1
from mountaintools import client as mt
print('===================')
# Local key/value store for associating relatively short strings (<=80 characters) with arbitrary keys (strings or dicts)
# Setting values (these should be short strings, <=80 characters)
mt.setValue(key='some-key1', value='hello 1')
mt.setValue(key=dict(name='some_name', number=2), value='hello 2')
# Getting values
val1 = mt.getValue(key='some-key1')
val2 = mt.getValue(key=dict(name='some_name', number=2))
print(val1)
print(val2)
print('===================')
# Setting password-protected values
mt.setValue(key='some_key2', password='******', value='the-secret-*y$#a')
# Retrieving password-protected values
print(mt.getValue(key='some_key2', password='******'))
print('===================')
# Local storage of data and files, retrievable by SHA-1 hash
path = mt.saveText('This is some text', basename='test.txt')
print(path)
# Output: sha1://482cb0cfcbed6740a2bcb659c9ccc22a4d27b369/test.txt
