Add not_params to ModularPipelineTesterMixin

tests/modular_pipelines/flux2/test_modular_pipeline_flux2_klein.py

@@ -49,6 +49,7 @@ class TestFlux2KleinModularPipelineFast(ModularPipelineTesterMixin):
 
     params = frozenset(["prompt", "height", "width"])
     batch_params = frozenset(["prompt"])
+    not_params = frozenset(["negative_prompt"])
     expected_workflow_blocks = FLUX2_KLEIN_WORKFLOWS
 
     def get_dummy_inputs(self, seed=0):
@@ -94,6 +95,7 @@ class TestFlux2KleinImageConditionedModularPipelineFast(ModularPipelineTesterMix
 
     params = frozenset(["prompt", "height", "width", "image"])
     batch_params = frozenset(["prompt", "image"])
+    not_params = frozenset(["negative_prompt"])
     expected_workflow_blocks = FLUX2_KLEIN_IMAGE_CONDITIONED_WORKFLOWS
 
     def get_dummy_inputs(self, seed=0):

tests/modular_pipelines/test_modular_pipelines_common.py

@@ -69,6 +69,10 @@ class ModularPipelineTesterMixin:
     # of the type of pipeline. They are always optional and have common
     # sense default values.
     optional_params = frozenset(["num_inference_steps", "num_images_per_prompt", "latents", "output_type"])
+    # Parameters the pipeline deliberately does NOT accept — e.g. `negative_prompt` on a
+    # guidance-distilled pipeline. `test_pipeline_call_signature` asserts they are absent,
+    # so accidentally (re)introducing one fails the test.
+    not_params = frozenset()
     # this is modular specific: generator needs to be a intermediate input because it's mutable
     intermediate_params = frozenset(["generator"])
     # Output type for the pipeline (e.g., "images" for image pipelines, "videos" for video pipelines)
@@ -176,6 +180,11 @@ def _check_for_parameters(parameters, expected_parameters, param_type):
         _check_for_parameters(self.params, input_parameters, "input")
         _check_for_parameters(self.optional_params, optional_parameters, "optional")
 
+        unsupported_parameters = {param for param in self.not_params if param in input_parameters}
+        assert len(unsupported_parameters) == 0, (
+            f"Parameters declared in `not_params` unexpectedly present in the pipeline inputs: {unsupported_parameters}"
+        )
+
     def test_inference_batch_consistent(self, batch_sizes=[2], batch_generator=True):
         pipe = self.get_pipeline().to(torch_device)