set random seed by default in batch processing (preproc + source_recon)

osl/preprocessing/batch.py

@@ -33,6 +33,7 @@
 from ..utils import logger as osl_logger
 from ..utils.parallel import dask_parallel_bag
 from ..utils.version_utils import check_version
+from ..utils.misc import set_random_seed
 
 logger = logging.getLogger(__name__)
 
@@ -641,6 +642,7 @@ def run_proc_chain(
     gen_report=None,
     overwrite=False,
     extra_funcs=None,
+    random_seed='auto',
     verbose="INFO",
     mneverbose="WARNING",
 ):
@@ -670,6 +672,9 @@ def run_proc_chain(
         Should we overwrite the output file if it already exists?
     extra_funcs : list
         User-defined functions.
+    random_seed : 'auto' (default), int or None
+        Random seed to set. If 'auto', a random seed will be generated. Random seeds are set for both Python and NumPy.
+        If None, no random seed is set.
     verbose : str
         Level of info to print.
         Can be: ``'CRITICAL'``, ``'ERROR'``, ``'WARNING'``, ``'INFO'``, ``'DEBUG'`` or ``'NOTSET'``.
@@ -740,6 +745,14 @@ def run_proc_chain(
     logger.info("{0} : Starting OSL Processing".format(now))
     logger.info("input : {0}".format(infile))
 
+    # Set random seed
+    if random_seed == 'auto':
+        set_random_seed()
+    elif random_seed is None:
+        pass
+    else:
+        set_random_seed(random_seed)
+
     # Write preprocessed data to output directory
     if outdir is not None:
         # Check for existing outputs - should be a .fif at least
@@ -867,6 +880,7 @@ def run_proc_batch(
     gen_report=True,
     overwrite=False,
     extra_funcs=None,
+    random_seed='auto',
     verbose="INFO",
     mneverbose="WARNING",
     strictrun=False,
@@ -898,6 +912,9 @@ def run_proc_batch(
         Should we generate a report?
     overwrite : bool
         Should we overwrite the output file if it exists?
+    random_seed : 'auto' (default), int or None
+        Random seed to set. If 'auto', a random seed will be generated. Random seeds are set for both Python and NumPy.
+        If None, no random seed is set.
     extra_funcs : list
         User-defined functions.
     verbose : str
@@ -924,7 +941,7 @@ def run_proc_batch(
     >>> from dask.distributed import Client
     >>> client = Client(threads_per_worker=1, n_workers=4)
     """
-
+   
     if outdir is None:
         # Use the current working directory
         outdir = os.getcwd()
@@ -944,6 +961,14 @@ def run_proc_batch(
 
     logger.info('Starting OSL Batch Processing')
 
+    # Set random seed
+    if random_seed == 'auto':
+        random_seed = set_random_seed()
+    elif random_seed is None:
+        pass
+    else:
+        set_random_seed(random_seed)
+
     # Check through inputs and parameters
     infiles, good_files_outnames, good_files = process_file_inputs(files)
 
@@ -994,6 +1019,7 @@ def run_proc_batch(
         gen_report=gen_report,
         overwrite=overwrite,
         extra_funcs=extra_funcs,
+        random_seed=random_seed,
     )
 
     # Loop through input files to generate arguments for run_proc_chain

osl/source_recon/batch.py

@@ -21,6 +21,7 @@
 from ..report import src_report
 from ..utils import logger as osl_logger
 from ..utils import validate_outdir, find_run_id, parallel
+from ..utils.misc import set_random_seed
 
 import logging
 logger = logging.getLogger(__name__)
@@ -101,6 +102,7 @@ def run_src_chain(
     verbose="INFO",
     mneverbose="WARNING",
     extra_funcs=None,
+    random_seed='auto',
 ):
     """Source reconstruction.
 
@@ -130,6 +132,9 @@ def run_src_chain(
         Level of MNE verbose.
     extra_funcs : list of functions
         Custom functions.
+    random_seed : 'auto' (default), int or None
+        Random seed to set. If 'auto', a random seed will be generated. Random seeds are set for both Python and NumPy.
+        If None, no random seed is set.
 
     Returns
     -------
@@ -160,6 +165,14 @@ def run_src_chain(
     logger.info("{0} : Starting OSL Processing".format(now))
     logger.info("input : {0}".format(outdir / subject))
 
+    # Set random seed
+    if random_seed == 'auto':
+        set_random_seed()
+    elif random_seed is None:
+        pass
+    else:
+        set_random_seed(random_seed)
+
     # Load config
     if not isinstance(config, dict):
         config = load_config(config)
@@ -251,6 +264,7 @@ def run_src_batch(
     mneverbose="WARNING",
     extra_funcs=None,
     dask_client=False,
+    random_seed='auto',
 ):
     """Batch source reconstruction.
 
@@ -283,6 +297,9 @@ def run_src_batch(
         Custom functions.
     dask_client : bool
         Are we using a dask client?
+    random_seed : 'auto' (default), int or None
+        Random seed to set. If 'auto', a random seed will be generated. Random seeds are set for both Python and NumPy.
+        If None, no random seed is set.
 
     Returns
     -------
@@ -302,6 +319,14 @@ def run_src_batch(
     osl_logger.set_up(log_file=logfile, level=verbose, startup=False)
     logger.info('Starting OSL Batch Source Reconstruction')
 
+    # Set random seed
+    if random_seed == 'auto':
+        random_seed = set_random_seed()
+    elif random_seed is None:
+        pass
+    else:
+        set_random_seed(random_seed)
+
     # Load config
     config = load_config(config)
     config_str = pprint.PrettyPrinter().pformat(config)
@@ -377,6 +402,7 @@ def run_src_batch(
         verbose=verbose,
         mneverbose=mneverbose,
         extra_funcs=extra_funcs,
+        random_seed=random_seed,
     )
 
     # Loop through input files to generate arguments for run_coreg_chain

osl/utils/misc.py

@@ -0,0 +1,30 @@
+"""Miscellaneous utility classes and functions.
+
+"""
+
+import logging
+import random
+import numpy as np
+
+
+logger = logging.getLogger(__name__)
+
+
+def set_random_seed(seed=None):
+    """Set all random seeds.
+
+    This includes Python's random module and NumPy.
+
+    Parameters
+    ----------
+    seed : int
+        Random seed.
+    """
+    if seed is None:
+        seed = random.randint(0, 2**32 - 1)
+
+    logger.info(f"Setting random seed to {seed}")
+
+    random.seed(seed)
+    np.random.seed(seed)
+    return seed