Postprocessing refactor all others (#3337)

* add text2img

* fix-copies

* add

* add all other pipelines

* add

* add

* add

* add

* add

* make style

* style + fix copies

---------
Co-authored-by: yiyixuxu <yixu310@gmail,com>

tests/pipelines/stable_diffusion_2/test_stable_diffusion_attend_and_excite.py

@@ -29,16 +29,19 @@ from diffusers import (
 from diffusers.utils import load_numpy, skip_mps, slow
 from diffusers.utils.testing_utils import require_torch_gpu
 
-from ..pipeline_params import TEXT_TO_IMAGE_BATCH_PARAMS, TEXT_TO_IMAGE_PARAMS
-from ..test_pipelines_common import PipelineTesterMixin
+from ..pipeline_params import TEXT_TO_IMAGE_BATCH_PARAMS, TEXT_TO_IMAGE_IMAGE_PARAMS, TEXT_TO_IMAGE_PARAMS
+from ..test_pipelines_common import PipelineLatentTesterMixin, PipelineTesterMixin
 
 
 @skip_mps
-class StableDiffusionAttendAndExcitePipelineFastTests(PipelineTesterMixin, unittest.TestCase):
+class StableDiffusionAttendAndExcitePipelineFastTests(
+    PipelineLatentTesterMixin, PipelineTesterMixin, unittest.TestCase
+):
     pipeline_class = StableDiffusionAttendAndExcitePipeline
     test_attention_slicing = False
     params = TEXT_TO_IMAGE_PARAMS
     batch_params = TEXT_TO_IMAGE_BATCH_PARAMS.union({"token_indices"})
+    image_params = TEXT_TO_IMAGE_IMAGE_PARAMS
 
     def get_dummy_components(self):
         torch.manual_seed(0)

tests/pipelines/stable_diffusion_2/test_stable_diffusion_depth.py

@@ -52,19 +52,22 @@ from diffusers.utils import (
 from diffusers.utils.testing_utils import require_torch_gpu, skip_mps
 
 from ..pipeline_params import TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS, TEXT_GUIDED_IMAGE_VARIATION_PARAMS
-from ..test_pipelines_common import PipelineTesterMixin
+from ..test_pipelines_common import PipelineLatentTesterMixin, PipelineTesterMixin
 
 
 torch.backends.cuda.matmul.allow_tf32 = False
 
 
 @skip_mps
-class StableDiffusionDepth2ImgPipelineFastTests(PipelineTesterMixin, unittest.TestCase):
+class StableDiffusionDepth2ImgPipelineFastTests(PipelineLatentTesterMixin, PipelineTesterMixin, unittest.TestCase):
     pipeline_class = StableDiffusionDepth2ImgPipeline
     test_save_load_optional_components = False
     params = TEXT_GUIDED_IMAGE_VARIATION_PARAMS - {"height", "width"}
     required_optional_params = PipelineTesterMixin.required_optional_params - {"latents"}
     batch_params = TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS
+    image_params = frozenset(
+        []
+    )  # TO-DO: update image_params once pipeline is refactored with VaeImageProcessor.preprocess
 
     def get_dummy_components(self):
         torch.manual_seed(0)
@@ -132,7 +135,7 @@ class StableDiffusionDepth2ImgPipelineFastTests(PipelineTesterMixin, unittest.Te
             backbone_config=backbone_config,
             backbone_featmap_shape=[1, 384, 24, 24],
         )
-        depth_estimator = DPTForDepthEstimation(depth_estimator_config)
+        depth_estimator = DPTForDepthEstimation(depth_estimator_config).eval()
         feature_extractor = DPTFeatureExtractor.from_pretrained(
             "hf-internal-testing/tiny-random-DPTForDepthEstimation"
         )

tests/pipelines/stable_diffusion_2/test_stable_diffusion_diffedit.py

@@ -34,16 +34,19 @@ from diffusers.utils import load_image, slow
 from diffusers.utils.testing_utils import floats_tensor, require_torch_gpu, torch_device
 
 from ..pipeline_params import TEXT_GUIDED_IMAGE_INPAINTING_BATCH_PARAMS, TEXT_GUIDED_IMAGE_INPAINTING_PARAMS
-from ..test_pipelines_common import PipelineTesterMixin
+from ..test_pipelines_common import PipelineLatentTesterMixin, PipelineTesterMixin
 
 
 torch.backends.cuda.matmul.allow_tf32 = False
 
 
-class StableDiffusionDiffEditPipelineFastTests(PipelineTesterMixin, unittest.TestCase):
+class StableDiffusionDiffEditPipelineFastTests(PipelineLatentTesterMixin, PipelineTesterMixin, unittest.TestCase):
     pipeline_class = StableDiffusionDiffEditPipeline
     params = TEXT_GUIDED_IMAGE_INPAINTING_PARAMS - {"height", "width", "image"} | {"image_latents"}
     batch_params = TEXT_GUIDED_IMAGE_INPAINTING_BATCH_PARAMS - {"image"} | {"image_latents"}
+    image_params = frozenset(
+        []
+    )  # TO-DO: update image_params once pipeline is refactored with VaeImageProcessor.preprocess
 
     def get_dummy_components(self):
         torch.manual_seed(0)

tests/pipelines/stable_diffusion_2/test_stable_diffusion_inpaint.py

@@ -27,16 +27,19 @@ from diffusers.utils import floats_tensor, load_image, load_numpy, torch_device
 from diffusers.utils.testing_utils import require_torch_gpu, slow
 
 from ..pipeline_params import TEXT_GUIDED_IMAGE_INPAINTING_BATCH_PARAMS, TEXT_GUIDED_IMAGE_INPAINTING_PARAMS
-from ..test_pipelines_common import PipelineTesterMixin
+from ..test_pipelines_common import PipelineLatentTesterMixin, PipelineTesterMixin
 
 
 torch.backends.cuda.matmul.allow_tf32 = False
 
 
-class StableDiffusion2InpaintPipelineFastTests(PipelineTesterMixin, unittest.TestCase):
+class StableDiffusion2InpaintPipelineFastTests(PipelineLatentTesterMixin, PipelineTesterMixin, unittest.TestCase):
     pipeline_class = StableDiffusionInpaintPipeline
     params = TEXT_GUIDED_IMAGE_INPAINTING_PARAMS
     batch_params = TEXT_GUIDED_IMAGE_INPAINTING_BATCH_PARAMS
+    image_params = frozenset(
+        []
+    )  # TO-DO: update image_params once pipeline is refactored with VaeImageProcessor.preprocess
 
     def get_dummy_components(self):
         torch.manual_seed(0)

tests/pipelines/stable_diffusion_2/test_stable_diffusion_latent_upscale.py

@@ -32,13 +32,13 @@ from diffusers.utils import floats_tensor, load_image, load_numpy, slow, torch_d
 from diffusers.utils.testing_utils import require_torch_gpu
 
 from ..pipeline_params import TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS, TEXT_GUIDED_IMAGE_VARIATION_PARAMS
-from ..test_pipelines_common import PipelineTesterMixin
+from ..test_pipelines_common import PipelineLatentTesterMixin, PipelineTesterMixin
 
 
 torch.backends.cuda.matmul.allow_tf32 = False
 
 
-class StableDiffusionLatentUpscalePipelineFastTests(PipelineTesterMixin, unittest.TestCase):
+class StableDiffusionLatentUpscalePipelineFastTests(PipelineLatentTesterMixin, PipelineTesterMixin, unittest.TestCase):
     pipeline_class = StableDiffusionLatentUpscalePipeline
     params = TEXT_GUIDED_IMAGE_VARIATION_PARAMS - {
         "height",
@@ -49,6 +49,10 @@ class StableDiffusionLatentUpscalePipelineFastTests(PipelineTesterMixin, unittes
     }
     required_optional_params = PipelineTesterMixin.required_optional_params - {"num_images_per_prompt"}
     batch_params = TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS
+    image_params = frozenset(
+        []
+    )  # TO-DO: update image_params once pipeline is refactored with VaeImageProcessor.preprocess
+
     test_cpu_offload = True
 
     @property

tests/pipelines/stable_unclip/test_stable_unclip.py

@@ -15,14 +15,15 @@ from diffusers import (
 from diffusers.pipelines.stable_diffusion.stable_unclip_image_normalizer import StableUnCLIPImageNormalizer
 from diffusers.utils.testing_utils import load_numpy, require_torch_gpu, slow, torch_device
 
-from ..pipeline_params import TEXT_TO_IMAGE_BATCH_PARAMS, TEXT_TO_IMAGE_PARAMS
-from ..test_pipelines_common import PipelineTesterMixin, assert_mean_pixel_difference
+from ..pipeline_params import TEXT_TO_IMAGE_BATCH_PARAMS, TEXT_TO_IMAGE_IMAGE_PARAMS, TEXT_TO_IMAGE_PARAMS
+from ..test_pipelines_common import PipelineLatentTesterMixin, PipelineTesterMixin, assert_mean_pixel_difference
 
 
-class StableUnCLIPPipelineFastTests(PipelineTesterMixin, unittest.TestCase):
+class StableUnCLIPPipelineFastTests(PipelineLatentTesterMixin, PipelineTesterMixin, unittest.TestCase):
     pipeline_class = StableUnCLIPPipeline
     params = TEXT_TO_IMAGE_PARAMS
     batch_params = TEXT_TO_IMAGE_BATCH_PARAMS
+    image_params = TEXT_TO_IMAGE_IMAGE_PARAMS
 
     # TODO(will) Expected attn_bias.stride(1) == 0 to be true, but got false
     test_xformers_attention = False

tests/pipelines/stable_unclip/test_stable_unclip_img2img.py

@@ -29,15 +29,19 @@ from diffusers.utils.testing_utils import (
 
 from ..pipeline_params import TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS, TEXT_GUIDED_IMAGE_VARIATION_PARAMS
 from ..test_pipelines_common import (
+    PipelineLatentTesterMixin,
     PipelineTesterMixin,
     assert_mean_pixel_difference,
 )
 
 
-class StableUnCLIPImg2ImgPipelineFastTests(PipelineTesterMixin, unittest.TestCase):
+class StableUnCLIPImg2ImgPipelineFastTests(PipelineLatentTesterMixin, PipelineTesterMixin, unittest.TestCase):
     pipeline_class = StableUnCLIPImg2ImgPipeline
     params = TEXT_GUIDED_IMAGE_VARIATION_PARAMS
     batch_params = TEXT_GUIDED_IMAGE_VARIATION_BATCH_PARAMS
+    image_params = frozenset(
+        []
+    )  # TO-DO: update image_params once pipeline is refactored with VaeImageProcessor.preprocess
 
     def get_dummy_components(self):
         embedder_hidden_size = 32

tests/pipelines/test_pipelines_common.py

@@ -79,7 +79,7 @@ class PipelineLatentTesterMixin:
         self.assertLess(max_diff, 1e-4, "`output_type=='pt'` generate different results from `output_type=='np'`")
 
         max_diff = np.abs(np.array(output_pil[0]) - (output_np * 255).round()).max()
-        self.assertLess(max_diff, 1e-4, "`output_type=='pil'` generate different results from `output_type=='np'`")
+        self.assertLess(max_diff, 2.0, "`output_type=='pil'` generate different results from `output_type=='np'`")
 
     def test_pt_np_pil_inputs_equivalent(self):
         if len(self.image_params) == 0: