Make safety_checker optional in more pipelines (#1796)

* Make safety_checker optional in more pipelines.

* Remove inappropriate comment in inpaint pipeline.

* InPaint Test: set feature_extractor to None.

* Remove import

* img2img test: set feature_extractor to None.

* inpaint sd2 test: set feature_extractor to None.

Co-authored-by: Suraj Patil <surajp815@gmail.com>

src/diffusers/pipelines/alt_diffusion/pipeline_alt_diffusion_img2img.py

@@ -92,7 +92,7 @@ class AltDiffusionImg2ImgPipeline(DiffusionPipeline):
         feature_extractor ([`CLIPFeatureExtractor`]):
             Model that extracts features from generated images to be used as inputs for the `safety_checker`.
     """
-    _optional_components = ["safety_checker"]
+    _optional_components = ["safety_checker", "feature_extractor"]
 
     def __init__(
         self,

src/diffusers/pipelines/paint_by_example/pipeline_paint_by_example.py

@@ -161,6 +161,8 @@ class PaintByExamplePipeline(DiffusionPipeline):
         feature_extractor ([`CLIPFeatureExtractor`]):
             Model that extracts features from generated images to be used as inputs for the `safety_checker`.
     """
+    # TODO: feature_extractor is required to encode initial images (if they are in PIL format),
+    # we should give a descriptive message if the pipeline doesn't have one.
     _optional_components = ["safety_checker"]
 
     def __init__(

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_image_variation.py

@@ -65,6 +65,8 @@ class StableDiffusionImageVariationPipeline(DiffusionPipeline):
         feature_extractor ([`CLIPFeatureExtractor`]):
             Model that extracts features from generated images to be used as inputs for the `safety_checker`.
     """
+    # TODO: feature_extractor is required to encode images (if they are in PIL format),
+    # we should give a descriptive message if the pipeline doesn't have one.
     _optional_components = ["safety_checker"]
 
     def __init__(

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py

@@ -90,7 +90,7 @@ class StableDiffusionImg2ImgPipeline(DiffusionPipeline):
         feature_extractor ([`CLIPFeatureExtractor`]):
             Model that extracts features from generated images to be used as inputs for the `safety_checker`.
     """
-    _optional_components = ["safety_checker"]
+    _optional_components = ["safety_checker", "feature_extractor"]
 
     # Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.__init__
     def __init__(

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py

@@ -166,7 +166,7 @@ class StableDiffusionInpaintPipeline(DiffusionPipeline):
         feature_extractor ([`CLIPFeatureExtractor`]):
             Model that extracts features from generated images to be used as inputs for the `safety_checker`.
     """
-    _optional_components = ["safety_checker"]
+    _optional_components = ["safety_checker", "feature_extractor"]
 
     def __init__(
         self,

tests/pipelines/stable_diffusion/test_stable_diffusion_img2img.py

@@ -31,7 +31,7 @@ from diffusers import (
 )
 from diffusers.utils import floats_tensor, load_image, load_numpy, nightly, slow, torch_device
 from diffusers.utils.testing_utils import require_torch_gpu
-from transformers import CLIPImageProcessor, CLIPTextConfig, CLIPTextModel, CLIPTokenizer
+from transformers import CLIPTextConfig, CLIPTextModel, CLIPTokenizer
 
 from ...test_pipelines_common import PipelineTesterMixin
 
@@ -78,7 +78,6 @@ class StableDiffusionImg2ImgPipelineFastTests(PipelineTesterMixin, unittest.Test
         )
         text_encoder = CLIPTextModel(text_encoder_config)
         tokenizer = CLIPTokenizer.from_pretrained("hf-internal-testing/tiny-random-clip")
-        feature_extractor = CLIPImageProcessor(crop_size=32, size=32)
 
         components = {
             "unet": unet,
@@ -87,7 +86,7 @@ class StableDiffusionImg2ImgPipelineFastTests(PipelineTesterMixin, unittest.Test
             "text_encoder": text_encoder,
             "tokenizer": tokenizer,
             "safety_checker": None,
-            "feature_extractor": feature_extractor,
+            "feature_extractor": None,
         }
         return components
 

tests/pipelines/stable_diffusion/test_stable_diffusion_inpaint.py

@@ -32,7 +32,7 @@ from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion_inpaint impo
 from diffusers.utils import floats_tensor, load_image, load_numpy, nightly, slow, torch_device
 from diffusers.utils.testing_utils import require_torch_gpu
 from PIL import Image
-from transformers import CLIPImageProcessor, CLIPTextConfig, CLIPTextModel, CLIPTokenizer
+from transformers import CLIPTextConfig, CLIPTextModel, CLIPTokenizer
 
 from ...test_pipelines_common import PipelineTesterMixin
 
@@ -79,7 +79,6 @@ class StableDiffusionInpaintPipelineFastTests(PipelineTesterMixin, unittest.Test
         )
         text_encoder = CLIPTextModel(text_encoder_config)
         tokenizer = CLIPTokenizer.from_pretrained("hf-internal-testing/tiny-random-clip")
-        feature_extractor = CLIPImageProcessor(crop_size=32, size=32)
 
         components = {
             "unet": unet,
@@ -88,7 +87,7 @@ class StableDiffusionInpaintPipelineFastTests(PipelineTesterMixin, unittest.Test
             "text_encoder": text_encoder,
             "tokenizer": tokenizer,
             "safety_checker": None,
-            "feature_extractor": feature_extractor,
+            "feature_extractor": None,
         }
         return components
 

tests/pipelines/stable_diffusion_2/test_stable_diffusion_inpaint.py

@@ -24,7 +24,7 @@ from diffusers import AutoencoderKL, PNDMScheduler, StableDiffusionInpaintPipeli
 from diffusers.utils import floats_tensor, load_image, load_numpy, torch_device
 from diffusers.utils.testing_utils import require_torch_gpu, slow
 from PIL import Image
-from transformers import CLIPImageProcessor, CLIPTextConfig, CLIPTextModel, CLIPTokenizer
+from transformers import CLIPTextConfig, CLIPTextModel, CLIPTokenizer
 
 from ...test_pipelines_common import PipelineTesterMixin
 
@@ -78,7 +78,6 @@ class StableDiffusion2InpaintPipelineFastTests(PipelineTesterMixin, unittest.Tes
         )
         text_encoder = CLIPTextModel(text_encoder_config)
         tokenizer = CLIPTokenizer.from_pretrained("hf-internal-testing/tiny-random-clip")
-        feature_extractor = CLIPImageProcessor(crop_size=32, size=32)
 
         components = {
             "unet": unet,
@@ -87,7 +86,7 @@ class StableDiffusion2InpaintPipelineFastTests(PipelineTesterMixin, unittest.Tes
             "text_encoder": text_encoder,
             "tokenizer": tokenizer,
             "safety_checker": None,
-            "feature_extractor": feature_extractor,
+            "feature_extractor": None,
         }
         return components