Generate: Load generation config when `device_map` is passed (#25413)

src/transformers/modeling_utils.py

@@ -2849,9 +2849,9 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
                     "'sequential'."
                 )
 
-            kwargs = {"no_split_module_classes": no_split_modules}
+            device_map_kwargs = {"no_split_module_classes": no_split_modules}
             if "special_dtypes" in inspect.signature(infer_auto_device_map).parameters:
-                kwargs["special_dtypes"] = special_dtypes
+                device_map_kwargs["special_dtypes"] = special_dtypes
             elif len(special_dtypes) > 0:
                 logger.warning(
                     "This model has some weights that should be kept in higher precision, you need to upgrade "
@@ -2863,12 +2863,12 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
                     dtype=target_dtype,
                     low_zero=(device_map == "balanced_low_0"),
                     max_memory=max_memory,
-                    **kwargs,
+                    **device_map_kwargs,
                 )
-            kwargs["max_memory"] = max_memory
+            device_map_kwargs["max_memory"] = max_memory
             # Make sure tied weights are tied before creating the device map.
             model.tie_weights()
-            device_map = infer_auto_device_map(model, dtype=target_dtype, **kwargs)
+            device_map = infer_auto_device_map(model, dtype=target_dtype, **device_map_kwargs)
 
             if load_in_8bit or load_in_4bit:
                 # The LM head / tied weights or any last module can stay on disk / CPU
@@ -2966,7 +2966,7 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
         model.eval()
 
         # If it is a model with generation capabilities, attempt to load the generation config
-        if model.can_generate():
+        if model.can_generate() and pretrained_model_name_or_path is not None:
             try:
                 model.generation_config = GenerationConfig.from_pretrained(
                     pretrained_model_name_or_path,
@@ -2982,7 +2982,7 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
                     _from_pipeline=from_pipeline,
                     **kwargs,
                 )
-            except (OSError, TypeError):
+            except OSError:
                 logger.info(
                     "Generation config file not found, using a generation config created from the model config."
                 )
@@ -2990,10 +2990,14 @@ class PreTrainedModel(nn.Module, ModuleUtilsMixin, GenerationMixin, PushToHubMix
 
         # Dispatch model with hooks on all devices if necessary
         if device_map is not None:
-            kwargs = {"device_map": device_map, "offload_dir": offload_folder, "offload_index": offload_index}
+            device_map_kwargs = {
+                "device_map": device_map,
+                "offload_dir": offload_folder,
+                "offload_index": offload_index,
+            }
             if "skip_keys" in inspect.signature(dispatch_model).parameters:
-                kwargs["skip_keys"] = model._skip_keys_device_placement
-            dispatch_model(model, **kwargs)
+                device_map_kwargs["skip_keys"] = model._skip_keys_device_placement
+            dispatch_model(model, **device_map_kwargs)
 
         if output_loading_info:
             if loading_info is None:

tests/test_modeling_utils.py

@@ -1036,6 +1036,20 @@ class ModelUtilsTest(TestCasePlus):
 
             self.assertEqual(model.__class__.__name__, model_ref.__class__.__name__)
 
+    def test_generation_config_is_loaded_with_model(self):
+        # Note: `joaogante/tiny-random-gpt2-with-generation-config` has a `generation_config.json` containing a dummy
+        # `transformers_version` field set to `foo`. If loading the file fails, this test also fails.
+
+        # 1. Load without further parameters
+        model = AutoModelForCausalLM.from_pretrained("joaogante/tiny-random-gpt2-with-generation-config")
+        self.assertEqual(model.generation_config.transformers_version, "foo")
+
+        # 2. Load with `device_map`
+        model = AutoModelForCausalLM.from_pretrained(
+            "joaogante/tiny-random-gpt2-with-generation-config", device_map="auto"
+        )
+        self.assertEqual(model.generation_config.transformers_version, "foo")
+
 
 @require_torch
 @is_staging_test