def graph():
layout_path = join(get_test_data_path(), "ds005")
layout = BIDSLayout(layout_path)
json_file = join(layout_path, "models", "ds-005_type-test_model.json")
graph = BIDSStatsModelsGraph(layout, json_file)
graph.load_collections(scan_length=480, subject=["01", "02"])
return graph
def test_incremental_data_loading():
layout_path = join(get_test_data_path(), "ds005")
layout = BIDSLayout(layout_path)
json_file = join(layout_path, "models", "ds-005_type-test_model.json")
graph = BIDSStatsModelsGraph(layout, json_file)
graph.load_collections(scan_length=480, subject=["01"], run=[1])
graph.load_collections(scan_length=480, subject=["02"], run=[2])
assert len(graph["run"].get_collections()) == 2
def _run_interface(self, runtime):
from bids.layout import BIDSLayout
from bids.modeling import BIDSStatsModelsGraph
layout = BIDSLayout.load(database_path=self.inputs.database_path)
selectors = self.inputs.selectors
graph = BIDSStatsModelsGraph(layout, self.inputs.model)
graph.load_collections(**selectors)
self._results['all_specs'] = self._load_graph(runtime, graph)
return runtime
def test_automodel_runs(model):
layout_path = join(get_test_data_path(), 'ds005')
layout = BIDSLayout(layout_path)
# Test to make sure an analaysis can be setup from the generated model
graph = BIDSStatsModelsGraph(layout, model)
graph.load_collections(scan_length=480, subject=["01", "02"])
outputs = graph["Run"].run()
assert len(outputs) == 6
cis = list(chain(*[op.contrasts for op in outputs]))
assert len(cis) == 6
outputs = graph["Subject"].run(cis)
# 2 subjects x 1 contrast
assert len(outputs) == 2
cis = list(chain(*[op.contrasts for op in outputs]))
assert len(cis) == 2
def test_auto_model_graph(model):
layout_path = join(get_test_data_path(), 'ds005')
layout = BIDSLayout(layout_path)
# Test to make sure an analaysis can be setup from the generated model
graph = BIDSStatsModelsGraph(layout, model)
graph.load_collections(scan_length=480)
assert model['Name'] == 'ds005_mixedgamblestask'
# run level
block = model['Nodes'][0]
assert block['Name'] == 'Run'
assert block['Level'] == 'Run'
assert block['Transformations'][0]['Name'] == 'Factor'
assert block['Contrasts'][0]['Name'] == 'run_parametric gain'
assert block['Contrasts'][0]['Weights'] == [1]
assert block['Contrasts'][0]['Test'] == 't'
# subject level
block = model['Nodes'][1]
assert block['Name'] == 'Subject'
assert block['Level'] == 'Subject'
assert block['Model']['X'][0] == 'run_parametric gain'
assert block['Contrasts'][0]['Name'] == 'subject_run_parametric gain'
assert block['Contrasts'][0]['Test'] == 'FEMA'
# dataset level
block = model['Nodes'][2]
assert block['Name'] == 'Dataset'
assert block['Level'] == 'Dataset'
assert block['Model']['X'][0] == 'subject_run_parametric gain'
assert block['Contrasts'][0][
'Name'] == 'dataset_subject_run_parametric gain'
assert block['Contrasts'][0]['Test'] == 't'
def run_fitlins(argv=None):
import re
from nipype import logging as nlogging
warnings.showwarning = _warn_redirect
opts = get_parser().parse_args(argv)
force_index = [
# If entry looks like `/<pattern>/`, treat `<pattern>` as a regex
re.compile(ign[1:-1]) if (ign[0], ign[-1]) == ('/', '/') else ign
# Iterate over empty tuple if undefined
for ign in opts.force_index or ()
]
ignore = [
# If entry looks like `/<pattern>/`, treat `<pattern>` as a regex
re.compile(ign[1:-1]) if (ign[0], ign[-1]) == ('/', '/') else ign
# Iterate over empty tuple if undefined
for ign in opts.ignore or ()
]
log_level = 25 + 5 * (opts.quiet - opts.verbose)
logger.setLevel(log_level)
nlogging.getLogger('nipype.workflow').setLevel(log_level)
nlogging.getLogger('nipype.interface').setLevel(log_level)
nlogging.getLogger('nipype.utils').setLevel(log_level)
if not opts.space:
# make it an explicit None
opts.space = None
if not opts.desc_label:
# make it an explicit None
opts.desc_label = None
ncpus = opts.n_cpus
if ncpus < 1:
ncpus = cpu_count()
plugin_settings = {
'plugin': 'MultiProc',
'plugin_args': {
'n_procs': ncpus,
'raise_insufficient': False,
'maxtasksperchild': 1,
},
}
if opts.mem_gb:
plugin_settings['plugin_args']['memory_gb'] = opts.mem_gb
model = default_path(opts.model, opts.bids_dir, 'model-default_smdl.json')
if opts.model in (None, 'default') and not op.exists(model):
model = 'default'
derivatives = True if not opts.derivatives else opts.derivatives
# Need this when specifying args directly (i.e. neuroscout)
# god bless neuroscout, but let's make it work for others!
if isinstance(derivatives, list) and len(derivatives) == 1:
# WRONG AND EVIL to those who have spaces in their paths... bad bad practice
# TODO - fix neuroscout
derivatives = derivatives[0].split(" ")
if opts.estimator != 'afni':
if opts.error_ts:
raise NotImplementedError(
"Saving the error time series is only implemented for"
" the afni estimator. If this is a feature you want"
f" for {opts.estimator} please let us know on github.")
if opts.derivative_label:
logger.warning('--derivative-label no longer has any effect; '
'set output directory name directly')
os.makedirs(opts.output_dir, exist_ok=True)
fub.write_derivative_description(opts.bids_dir, opts.output_dir,
vars(opts))
work_dir = mkdtemp() if opts.work_dir is None else opts.work_dir
# Go ahead and initialize the layout database
if opts.database_path is None:
database_path = Path(work_dir) / 'dbcache'
reset_database = True
else:
database_path = opts.database_path
reset_database = False
indexer = bids.BIDSLayoutIndexer(ignore=ignore, force_index=force_index)
layout = bids.BIDSLayout(
opts.bids_dir,
derivatives=derivatives,
database_path=database_path,
reset_database=reset_database,
indexer=indexer,
)
subject_list = None
if opts.participant_label is not None:
subject_list = fub.collect_participants(
layout, participant_label=opts.participant_label)
# Build main workflow
logger.log(25, INIT_MSG(version=__version__, subject_list=subject_list))
# TODO: Fix AUTO_MODEL
# if model == 'default':
# models = auto_model(layout)
# else:
# import json
# if op.exists(model):
# model_dict = json.loads(Path(model).read_text())
# models = [model_dict]
model_dict = None
if model == 'default':
retcode = 1
raise NotImplementedError(
"The default model has not been implemented yet.")
else:
import json
if op.exists(model):
model_dict = json.loads(Path(model).read_text())
if not model_dict:
raise ValueError(
f'model_dict cannot be empty. Invalid model filepath {model}.')
graph = BIDSStatsModelsGraph(layout, model_dict)
fitlins_wf = init_fitlins_wf(
database_path,
opts.output_dir,
graph=graph,
analysis_level=opts.analysis_level,
model=model,
space=opts.space,
desc=opts.desc_label,
participants=subject_list,
base_dir=work_dir,
smoothing=opts.smoothing,
drop_missing=opts.drop_missing,
drift_model=opts.drift_model,
estimator=opts.estimator,
errorts=opts.error_ts,
)
fitlins_wf.config = deepcopy(config.get_fitlins_config()._sections)
if opts.work_dir:
# dump crashes in working directory (non /tmp)
fitlins_wf.config['execution']['crashdump_dir'] = opts.work_dir
retcode = 0
if not opts.reports_only:
try:
fitlins_wf.run(**plugin_settings)
except Exception as e:
logger.critical(f"FitLins failed: {e}")
raise
run_context = {
'version': __version__,
'command': ' '.join(sys.argv),
'timestamp': time.strftime('%Y-%m-%d %H:%M:%S %z'),
}
selectors = {'desc': opts.desc_label, 'space': opts.space}
if subject_list is not None:
selectors['subject'] = subject_list
graph.load_collections(**selectors)
report_dict = build_report_dict(opts.output_dir, work_dir, graph)
write_full_report(report_dict, run_context, opts.output_dir)
return retcode
json_file = './model-narps_smdl.json'
spec = json.loads(Path(json_file).read_text())
spec
# %%
graph = BIDSStatsModelsGraph(layout, spec)
# %%
graph.write_graph(format='svg')
# %%
root_node = graph.root_node
# %%
try:
graph.load_collections()
except ValueError:
graph.load_collections(scan_length=453) # TR = 1, nvols = 453
# %%
specs = root_node.run(group_by=root_node.group_by, force_dense=False)
len(specs)
# %%
api(specs[0])
# %%
specs[0]
# %%
plot_design_matrix(specs[0].X)