Fix WoQ loading for Large Model (#2200)

* fix save/load

Change-Id: I36d1ca10426b65abfe2093091f41906cecf2fcaf
Signed-off-by: Yi Liu <yiliu4@habana.ai>

* fix check

Change-Id: I2cc647ed3ecc7ff499206a3d79e0f71e25605f7e
Signed-off-by: Yi Liu <yiliu4@habana.ai>

---------

Signed-off-by: Yi Liu <yiliu4@habana.ai>
Co-authored-by: Yi Liu <yiliu4@habana.ai>

Author: yiliu30

neural_compressor/torch/algorithms/weight_only/save_load.py

@@ -478,7 +478,11 @@ def _get_model_class_and_config(self):
 
         has_remote_code = hasattr(config, "auto_map") and AutoModelForCausalLM.__name__ in config.auto_map
 
-        has_local_code = type(config) in AutoModelForCausalLM._model_mapping.keys()
+        has_local_code = (
+            hasattr(AutoModelForCausalLM, "_model_mapping")
+            and type(config) in AutoModelForCausalLM._model_mapping.keys()
+        )
+
         trust_remote_code = resolve_trust_remote_code(
             trust_remote_code,
             self.model_name_or_path,
@@ -601,6 +605,7 @@ def _get_loaded_state_dict(self, config):
         # if hpu format tensor can be used directly, then update resolved_archive_file to the hpu format tensor file
         if self._use_hpu_module():
             resolved_archive_file = os.path.join(self._model_local_dir, HPU_SAFE_WEIGHTS_NAME)
+            is_sharded = False
 
         logger.info(f"Find weight file {resolved_archive_file}")