better comments

src/sparseml/transformers/sparsification/sparse_model.py

@@ -101,22 +101,29 @@ def skip(*args, **kwargs):
         )
 
         # instantiate compressor from model config
-        compressor = ModelCompressor.from_pretrained(pretrained_model_name_or_path, **kwargs)
+        compressor = ModelCompressor.from_pretrained(
+            pretrained_model_name_or_path, **kwargs
+        )
 
         # temporarily set the log level to error, to ignore printing out long missing
         # and unexpected key error messages (these are EXPECTED for quantized models)
         logger = logging.getLogger("transformers.modeling_utils")
         restore_log_level = logger.getEffectiveLevel()
         logger.setLevel(level=logging.ERROR)
-
-
-        if kwargs.get('trust_remote_code'):
+
+        if kwargs.get("trust_remote_code"):
+            # By artifically aliasing
+            # class name SparseAutoModelForCausallLM to
+            # AutoModelForCausalLM we can "trick" the
+            # `from_pretrained` method into properly
+            # resolving the logic when
+            # (has_remote_code and trust_remote_code) == True
             cls.__name__ = AutoModelForCausalLM.__name__
-            
+
         model = super(AutoModelForCausalLM, cls).from_pretrained(
             pretrained_model_name_or_path, *model_args, **kwargs
         )
-        
+
         if model.dtype != model.config.torch_dtype:
             _LOGGER.warning(
                 f"The dtype of the loaded model: {model.dtype} is different "