Merge branch 'main' into let-server-decide-default-repo-visibility

Author: Wauplin

docs/source/en/api/pipelines/flux.md

@@ -148,7 +148,7 @@ image.save("output.png")
 **Note:** `black-forest-labs/Flux.1-Depth-dev` is _not_ a ControlNet model. [`ControlNetModel`] models are a separate component from the UNet/Transformer whose residuals are added to the actual underlying model. Depth Control is an alternate architecture that achieves effectively the same results as a ControlNet model would, by using channel-wise concatenation with input control condition and ensuring the transformer learns structure control by following the condition as closely as possible.
 
 ```python
-# !pip install git+https://github.com/asomoza/image_gen_aux.git
+# !pip install git+https://github.com/huggingface/image_gen_aux
 import torch
 from diffusers import FluxControlPipeline, FluxTransformer2DModel
 from diffusers.utils import load_image

tests/single_file/single_file_testing_utils.py

@@ -156,14 +156,14 @@ def test_single_file_components_with_original_config_local_files_only(
     def test_single_file_format_inference_is_same_as_pretrained(self, expected_max_diff=1e-4):
         sf_pipe = self.pipeline_class.from_single_file(self.ckpt_path, safety_checker=None)
         sf_pipe.unet.set_attn_processor(AttnProcessor())
-        sf_pipe.enable_model_cpu_offload()
+        sf_pipe.enable_model_cpu_offload(device=torch_device)
 
         inputs = self.get_inputs(torch_device)
         image_single_file = sf_pipe(**inputs).images[0]
 
         pipe = self.pipeline_class.from_pretrained(self.repo_id, safety_checker=None)
         pipe.unet.set_attn_processor(AttnProcessor())
-        pipe.enable_model_cpu_offload()
+        pipe.enable_model_cpu_offload(device=torch_device)
 
         inputs = self.get_inputs(torch_device)
         image = pipe(**inputs).images[0]

tests/single_file/test_model_controlnet_single_file.py

@@ -22,17 +22,19 @@
     ControlNetModel,
 )
 from diffusers.utils.testing_utils import (
+    backend_empty_cache,
     enable_full_determinism,
-    require_torch_gpu,
+    require_torch_accelerator,
     slow,
+    torch_device,
 )
 
 
 enable_full_determinism()
 
 
 @slow
-@require_torch_gpu
+@require_torch_accelerator
 class ControlNetModelSingleFileTests(unittest.TestCase):
     model_class = ControlNetModel
     ckpt_path = "https://huggingface.co/lllyasviel/ControlNet-v1-1/blob/main/control_v11p_sd15_canny.pth"
@@ -41,12 +43,12 @@ class ControlNetModelSingleFileTests(unittest.TestCase):
     def setUp(self):
         super().setUp()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def tearDown(self):
         super().tearDown()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def test_single_file_components(self):
         model = self.model_class.from_pretrained(self.repo_id)

tests/single_file/test_model_sd_cascade_unet_single_file.py

@@ -21,9 +21,11 @@
 from diffusers import StableCascadeUNet
 from diffusers.utils import logging
 from diffusers.utils.testing_utils import (
+    backend_empty_cache,
     enable_full_determinism,
-    require_torch_gpu,
+    require_torch_accelerator,
     slow,
+    torch_device,
 )
 
 
@@ -33,17 +35,17 @@
 
 
 @slow
-@require_torch_gpu
+@require_torch_accelerator
 class StableCascadeUNetSingleFileTest(unittest.TestCase):
     def setUp(self):
         super().setUp()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def tearDown(self):
         super().tearDown()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def test_single_file_components_stage_b(self):
         model_single_file = StableCascadeUNet.from_single_file(

tests/single_file/test_model_vae_single_file.py

@@ -22,10 +22,11 @@
     AutoencoderKL,
 )
 from diffusers.utils.testing_utils import (
+    backend_empty_cache,
     enable_full_determinism,
     load_hf_numpy,
     numpy_cosine_similarity_distance,
-    require_torch_gpu,
+    require_torch_accelerator,
     slow,
     torch_device,
 )
@@ -35,7 +36,7 @@
 
 
 @slow
-@require_torch_gpu
+@require_torch_accelerator
 class AutoencoderKLSingleFileTests(unittest.TestCase):
     model_class = AutoencoderKL
     ckpt_path = (
@@ -48,12 +49,12 @@ class AutoencoderKLSingleFileTests(unittest.TestCase):
     def setUp(self):
         super().setUp()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def tearDown(self):
         super().tearDown()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def get_file_format(self, seed, shape):
         return f"gaussian_noise_s={seed}_shape={'_'.join([str(s) for s in shape])}.npy"

tests/single_file/test_stable_diffusion_controlnet_img2img_single_file.py

@@ -8,9 +8,10 @@
 from diffusers.loaders.single_file_utils import _extract_repo_id_and_weights_name
 from diffusers.utils import load_image
 from diffusers.utils.testing_utils import (
+    backend_empty_cache,
     enable_full_determinism,
     numpy_cosine_similarity_distance,
-    require_torch_gpu,
+    require_torch_accelerator,
     slow,
     torch_device,
 )
@@ -27,7 +28,7 @@
 
 
 @slow
-@require_torch_gpu
+@require_torch_accelerator
 class StableDiffusionControlNetPipelineSingleFileSlowTests(unittest.TestCase, SDSingleFileTesterMixin):
     pipeline_class = StableDiffusionControlNetPipeline
     ckpt_path = (
@@ -41,12 +42,12 @@ class StableDiffusionControlNetPipelineSingleFileSlowTests(unittest.TestCase, SD
     def setUp(self):
         super().setUp()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def tearDown(self):
         super().tearDown()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def get_inputs(self, device, generator_device="cpu", dtype=torch.float32, seed=0):
         generator = torch.Generator(device=generator_device).manual_seed(seed)

tests/single_file/test_stable_diffusion_controlnet_inpaint_single_file.py

@@ -8,10 +8,12 @@
 from diffusers.loaders.single_file_utils import _extract_repo_id_and_weights_name
 from diffusers.utils import load_image
 from diffusers.utils.testing_utils import (
+    backend_empty_cache,
     enable_full_determinism,
     numpy_cosine_similarity_distance,
-    require_torch_gpu,
+    require_torch_accelerator,
     slow,
+    torch_device,
 )
 
 from .single_file_testing_utils import (
@@ -26,7 +28,7 @@
 
 
 @slow
-@require_torch_gpu
+@require_torch_accelerator
 class StableDiffusionControlNetInpaintPipelineSingleFileSlowTests(unittest.TestCase, SDSingleFileTesterMixin):
     pipeline_class = StableDiffusionControlNetInpaintPipeline
     ckpt_path = "https://huggingface.co/botp/stable-diffusion-v1-5-inpainting/blob/main/sd-v1-5-inpainting.ckpt"
@@ -36,12 +38,12 @@ class StableDiffusionControlNetInpaintPipelineSingleFileSlowTests(unittest.TestC
     def setUp(self):
         super().setUp()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def tearDown(self):
         super().tearDown()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def get_inputs(self):
         control_image = load_image(

tests/single_file/test_stable_diffusion_controlnet_single_file.py

@@ -8,10 +8,12 @@
 from diffusers.loaders.single_file_utils import _extract_repo_id_and_weights_name
 from diffusers.utils import load_image
 from diffusers.utils.testing_utils import (
+    backend_empty_cache,
     enable_full_determinism,
     numpy_cosine_similarity_distance,
-    require_torch_gpu,
+    require_torch_accelerator,
     slow,
+    torch_device,
 )
 
 from .single_file_testing_utils import (
@@ -26,7 +28,7 @@
 
 
 @slow
-@require_torch_gpu
+@require_torch_accelerator
 class StableDiffusionControlNetPipelineSingleFileSlowTests(unittest.TestCase, SDSingleFileTesterMixin):
     pipeline_class = StableDiffusionControlNetPipeline
     ckpt_path = (
@@ -40,12 +42,12 @@ class StableDiffusionControlNetPipelineSingleFileSlowTests(unittest.TestCase, SD
     def setUp(self):
         super().setUp()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def tearDown(self):
         super().tearDown()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def get_inputs(self):
         control_image = load_image(

tests/single_file/test_stable_diffusion_img2img_single_file.py

@@ -8,9 +8,11 @@
 )
 from diffusers.utils import load_image
 from diffusers.utils.testing_utils import (
+    backend_empty_cache,
     enable_full_determinism,
-    require_torch_gpu,
+    require_torch_accelerator,
     slow,
+    torch_device,
 )
 
 from .single_file_testing_utils import SDSingleFileTesterMixin
@@ -20,7 +22,7 @@
 
 
 @slow
-@require_torch_gpu
+@require_torch_accelerator
 class StableDiffusionImg2ImgPipelineSingleFileSlowTests(unittest.TestCase, SDSingleFileTesterMixin):
     pipeline_class = StableDiffusionImg2ImgPipeline
     ckpt_path = (
@@ -34,12 +36,12 @@ class StableDiffusionImg2ImgPipelineSingleFileSlowTests(unittest.TestCase, SDSin
     def setUp(self):
         super().setUp()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def tearDown(self):
         super().tearDown()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def get_inputs(self, device, generator_device="cpu", dtype=torch.float32, seed=0):
         generator = torch.Generator(device=generator_device).manual_seed(seed)
@@ -63,7 +65,7 @@ def test_single_file_format_inference_is_same_as_pretrained(self):
 
 
 @slow
-@require_torch_gpu
+@require_torch_accelerator
 class StableDiffusion21Img2ImgPipelineSingleFileSlowTests(unittest.TestCase, SDSingleFileTesterMixin):
     pipeline_class = StableDiffusionImg2ImgPipeline
     ckpt_path = "https://huggingface.co/stabilityai/stable-diffusion-2-1/blob/main/v2-1_768-ema-pruned.safetensors"
@@ -73,12 +75,12 @@ class StableDiffusion21Img2ImgPipelineSingleFileSlowTests(unittest.TestCase, SDS
     def setUp(self):
         super().setUp()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def tearDown(self):
         super().tearDown()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def get_inputs(self, device, generator_device="cpu", dtype=torch.float32, seed=0):
         generator = torch.Generator(device=generator_device).manual_seed(seed)

tests/single_file/test_stable_diffusion_inpaint_single_file.py

@@ -8,9 +8,11 @@
 )
 from diffusers.utils import load_image
 from diffusers.utils.testing_utils import (
+    backend_empty_cache,
     enable_full_determinism,
-    require_torch_gpu,
+    require_torch_accelerator,
     slow,
+    torch_device,
 )
 
 from .single_file_testing_utils import SDSingleFileTesterMixin
@@ -20,7 +22,7 @@
 
 
 @slow
-@require_torch_gpu
+@require_torch_accelerator
 class StableDiffusionInpaintPipelineSingleFileSlowTests(unittest.TestCase, SDSingleFileTesterMixin):
     pipeline_class = StableDiffusionInpaintPipeline
     ckpt_path = "https://huggingface.co/botp/stable-diffusion-v1-5-inpainting/blob/main/sd-v1-5-inpainting.ckpt"
@@ -30,12 +32,12 @@ class StableDiffusionInpaintPipelineSingleFileSlowTests(unittest.TestCase, SDSin
     def setUp(self):
         super().setUp()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def tearDown(self):
         super().tearDown()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def get_inputs(self, device, generator_device="cpu", dtype=torch.float32, seed=0):
         generator = torch.Generator(device=generator_device).manual_seed(seed)
@@ -78,7 +80,7 @@ def test_single_file_components_with_original_config_local_files_only(self):
 
 
 @slow
-@require_torch_gpu
+@require_torch_accelerator
 class StableDiffusion21InpaintPipelineSingleFileSlowTests(unittest.TestCase, SDSingleFileTesterMixin):
     pipeline_class = StableDiffusionInpaintPipeline
     ckpt_path = (
@@ -90,12 +92,12 @@ class StableDiffusion21InpaintPipelineSingleFileSlowTests(unittest.TestCase, SDS
     def setUp(self):
         super().setUp()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def tearDown(self):
         super().tearDown()
         gc.collect()
-        torch.cuda.empty_cache()
+        backend_empty_cache(torch_device)
 
     def get_inputs(self, device, generator_device="cpu", dtype=torch.float32, seed=0):
         generator = torch.Generator(device=generator_device).manual_seed(seed)