Merge pull request #134 from gabinguo/main

issue#126: torch.load device issue.

src/peft/peft_model.py

@@ -155,7 +155,8 @@ class PeftModel(PushToHubMixin, torch.nn.Module):
                     f"Please check that the file {WEIGHTS_NAME} is present at {model_id}."
                 )
 
-        adapters_weights = torch.load(filename)
+        adapters_weights = torch.load(
+            filename, map_location=torch.device("cuda" if torch.cuda.is_available() else "cpu"))
         # load the weights into the model
         model = set_peft_model_state_dict(model, adapters_weights)
         if getattr(model, "hf_device_map", None) is not None: