Raise relevent err when wrong type is passed in as the accelerator_config (#29997)

* Raise relevent err

* Use type instead

src/transformers/training_args.py

@@ -1815,6 +1815,13 @@ class TrainingArguments:
                     self.accelerator_config = AcceleratorConfig()
                 elif isinstance(self.accelerator_config, dict):
                     self.accelerator_config = AcceleratorConfig(**self.accelerator_config)
+                # Check that a user didn't pass in the class instantiator
+                # such as `accelerator_config = AcceleratorConfig`
+                elif isinstance(self.accelerator_config, type):
+                    raise NotImplementedError(
+                        "Tried passing in a callable to `accelerator_config`, but this is not supported. "
+                        "Please pass in a fully constructed `AcceleratorConfig` object instead."
+                    )
                 else:
                     self.accelerator_config = AcceleratorConfig.from_json_file(self.accelerator_config)
             if self.dispatch_batches is not None:

tests/trainer/test_trainer.py

@@ -3104,6 +3104,35 @@ class TrainerIntegrationTest(TestCasePlus, TrainerIntegrationCommon):
                 trainer = Trainer(model=model, args=args, eval_dataset=eval_dataset)
             self.assertTrue("The `AcceleratorConfig`'s `num_steps` is set but" in str(context.exception))
 
+    def test_accelerator_config_not_instantiated(self):
+        # Checks that accelerator kwargs can be passed through
+        # and the accelerator is initialized respectively
+        with tempfile.TemporaryDirectory() as tmp_dir:
+            with self.assertRaises(NotImplementedError) as context:
+                _ = RegressionTrainingArguments(
+                    output_dir=tmp_dir,
+                    accelerator_config=AcceleratorConfig,
+                )
+            self.assertTrue("Tried passing in a callable to `accelerator_config`" in str(context.exception))
+
+        # Now test with a custom subclass
+        @dataclasses.dataclass
+        class CustomAcceleratorConfig(AcceleratorConfig):
+            pass
+
+        @dataclasses.dataclass
+        class CustomTrainingArguments(TrainingArguments):
+            accelerator_config: dict = dataclasses.field(
+                default=CustomAcceleratorConfig,
+            )
+
+        with tempfile.TemporaryDirectory() as tmp_dir:
+            with self.assertRaises(NotImplementedError) as context:
+                _ = CustomTrainingArguments(
+                    output_dir=tmp_dir,
+                )
+            self.assertTrue("Tried passing in a callable to `accelerator_config`" in str(context.exception))
+
 
 @require_torch
 @is_staging_test