clearml offline mode (#363)

Co-authored-by: Moshe Raboh [email protected] <[email protected]>

fuse/data/ops/ops_read.py

@@ -136,6 +136,7 @@ def __init__(
         :param key_index: name of value in sample_dict which will be used as the key/index
         :param key_column: name of the column which use as key/index. In case of None, the original dataframe index will be used to extract the values for a single sample.
         """
+        super().__init__()
         # store input
         self._data_filename = data_filename
         self._columns_to_extract = columns_to_extract
@@ -146,7 +147,7 @@ def __init__(
         self._h5 = h5py.File(self._data_filename, "r")
 
         if self._columns_to_extract is None:
-            self._columns_to_extract = self._h5.keys()
+            self._columns_to_extract = list(self._h5.keys())
 
         self._num_samples = len(self._h5[self._columns_to_extract[0]])
 

fuse/dl/lightning/pl_funcs.py

@@ -55,6 +55,7 @@ def start_clearml_logger(
     auto_resource_monitoring: bool = True,
     auto_connect_streams: Union[bool, Mapping[str, bool]] = True,
     deferred_init: bool = False,
+    offline_mode: bool = False,
 ) -> TaskInstance:
     """
     Just a fuse function to quickly start the clearml logger. It sets up patches to pytorch lightning logging hooks so it doesn't need to be passed to any lightning logger.
@@ -86,6 +87,10 @@ def start_clearml_logger(
         bool_start_logger = True
 
     if bool_start_logger:
+        if offline_mode:  # Use the set_offline class method before initializing a Task
+            Task.set_offline(offline_mode=True)
+            os.environ["CLEARML_OFFLINE_MODE"] = "1"
+
         task = Task.init(
             project_name=project_name,
             task_name=task_name,

fuse/dl/lightning/pl_module.py

@@ -176,6 +176,9 @@ def __init__(
 
     ## forward
     def forward(self, batch_dict: NDict) -> NDict:
+        # workaround for fsdp
+        if not isinstance(batch_dict, NDict):
+            batch_dict = NDict(batch_dict)
         return self._model(batch_dict)
 
     ## Step