torch setuptools bugfix (#593)

* Remove: Ignore tensorboard logger if attribute error due to torch and setuptools

* Remove: Ignore tensorboard logger if attribute error due to torch and setuptools

* Use warning instead of printing
Style Fixes

Author: rahul-tuli

src/sparseml/pytorch/image_classification/utils/helpers.py

@@ -16,6 +16,7 @@
 Helper methods for image classification/detection based tasks
 """
 import os
+import warnings
 from enum import Enum, auto, unique
 from typing import Any, List, Optional, Tuple, Union
 
@@ -99,7 +100,14 @@ def get_save_dir_and_loggers(
         if task == Tasks.TRAIN:
             logs_dir = os.path.join(logs_dir, model_id)
             create_dirs(logs_dir)
-            loggers.append(TensorBoardLogger(log_path=logs_dir))
+            try:
+                loggers.append(TensorBoardLogger(log_path=logs_dir))
+            except AttributeError:
+                warnings.warn(
+                    "Failed to initialize TensorBoard logger, "
+                    "it will not be used for logging",
+                )
+
         print(f"Model id is set to {model_id}")
     else:
         # do not log for non main processes