Fix quant model re-load bug (#978) (#1027)

Author: bfineran

src/sparseml/pytorch/utils/model.py

@@ -80,7 +80,6 @@ def load_model(
     if path.startswith("zoo:"):
         path = download_framework_model_by_recipe_type(Model(path))
     model_dict = torch.load(path, map_location="cpu")
-    current_dict = model.state_dict()
     recipe = model_dict.get("recipe")
 
     if recipe:
@@ -90,6 +89,7 @@ def load_model(
         checkpoint_manager = ScheduledModifierManager.from_yaml(recipe)
         checkpoint_manager.apply_structure(module=model, epoch=epoch)
 
+    current_dict = model.state_dict()
     if "state_dict" in model_dict:
         model_dict = model_dict["state_dict"]