setting default values of num_workers to 1

infocusp · Jan 6, 2025 · df6a8db · df6a8db
1 parent 14627fb
commit df6a8db
Show file tree

Hide file tree

Showing 4 changed files with 6 additions and 6 deletions.
diff --git a/scalr/data/preprocess/_preprocess.py b/scalr/data/preprocess/_preprocess.py
@@ -54,7 +54,7 @@ def process_data(self,
                      full_data: Union[AnnData, AnnCollection],
                      sample_chunksize: int,
                      dirpath: str,
-                     num_workers: int = None):
+                     num_workers: int = 1):
         """A function to process the entire data chunkwise and write the processed data
         to disk.
 

diff --git a/scalr/feature/feature_subsetting.py b/scalr/feature/feature_subsetting.py
@@ -35,7 +35,7 @@ def __init__(self,
                  mappings: dict,
                  dirpath: str = None,
                  device: str = 'cpu',
-                 num_workers: int = None,
+                 num_workers: int = 1,
                  sample_chunksize: int = None):
         """Initialize required parameters for feature subset training.
 
@@ -52,7 +52,8 @@ def __init__(self,
             num_workers (int, optional): Number of parallel processes to launch to train multiple
                                          feature subsets simultaneously. Defaults to using single
                                          process.
-            sample_chunksize (int): Chunks of samples to be loaded in memory at once.
+            sample_chunksize (int, optional): Chunks of samples to be loaded in memory at once. 
+                                              Required when `num_workers` > 1.
         """
         self.feature_subsetsize = feature_subsetsize
         self.chunk_model_config = chunk_model_config

diff --git a/scalr/feature_extraction_pipeline.py b/scalr/feature_extraction_pipeline.py
@@ -85,7 +85,7 @@ def feature_subsetted_model_training(self) -> list[nn.Module]:
 
         self.feature_subsetsize = self.feature_selection_config.get(
             'feature_subsetsize', len(self.val_data.var_names))
-        self.num_workers = self.feature_selection_config.get('num_workers')
+        self.num_workers = self.feature_selection_config.get('num_workers', 1)
 
         chunk_model_config = self.feature_selection_config.get('model')
         chunk_model_train_config = self.feature_selection_config.get(

diff --git a/scalr/utils/file_utils.py b/scalr/utils/file_utils.py
@@ -83,7 +83,7 @@ def write_chunkwise_data(full_data: Union[AnnData, AnnCollection],
                          sample_inds: list[int] = None,
                          feature_inds: list[int] = None,
                          transform: callable = None,
-                         num_workers: int = None):
+                         num_workers: int = 1):
     """This function writes data subsets iteratively in a chunkwise manner, to ensure
     only at most `sample_chunksize` samples are loaded at a time.
 
@@ -152,7 +152,6 @@ def transform_and_write_data(data: AnnData, chunk_number: int):
                  num_workers)) if num_workers else sample_chunksize
 
         # Execute parallel jobs for transformation and witing of data.
-        # In case of `num_workers = None`, single process is used.
         parallel = Parallel(n_jobs=num_workers)
         parallel(
             delayed(transform_and_write_data)(