fix transformers naming

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

docs/source/en/api/pipelines/stable_diffusion/attend_and_excite.mdx

@@ -32,7 +32,7 @@ Resources
 
 | Pipeline | Tasks | Colab | Demo
 |---|---|:---:|:---:|
-| [pipeline_semantic_stable_diffusion_attend_and_excite.py](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_semantic_stable_diffusion_attend_and_excite) | *Text-to-Image Generation* | - | https://huggingface.co/spaces/AttendAndExcite/Attend-and-Excite
+| [pipeline_semantic_stable_diffusion_attend_and_excite.py](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_semantic_stable_diffusion_attend_and_excite) | *Text-to-Image Generation* | - | -
 
 
 ### Usage example

docs/source/en/api/pipelines/stable_diffusion/img2img.mdx

@@ -20,9 +20,6 @@ The original codebase can be found here: [CampVis/stable-diffusion](https://gith
 
 [`StableDiffusionImg2ImgPipeline`] is compatible with all Stable Diffusion checkpoints for [Text-to-Image](./text2img) 
 
-The pipeline uses the diffusion-denoising mechanism proposed by SDEdit ([SDEdit: Guided Image Synthesis and Editing with Stochastic Differential Equations](https://arxiv.org/abs/2108.01073)
-proposed by Chenlin Meng, Yutong He, Yang Song, Jiaming Song, Jiajun Wu, Jun-Yan Zhu, Stefano Ermon).
-
 [[autodoc]] StableDiffusionImg2ImgPipeline
 	- all
 	- __call__

docs/source/en/api/pipelines/stable_diffusion/pix2pix.mdx

@@ -60,7 +60,7 @@ def download_image(url):
 image = download_image(url)
 
 prompt = "make the mountains snowy"
-images = pipe(prompt, image=image, num_inference_steps=20, image_guidance_scale=1.5, guidance_scale=7).images
+edit = pipe(prompt, image=image, num_inference_steps=20, image_guidance_scale=1.5, guidance_scale=7).images[0]
 images[0].save("snowy_mountains.png")
 ```
 

docs/source/en/using-diffusers/controlling_generation.mdx

@@ -34,7 +34,6 @@ Unless otherwise mentioned, these are techniques that work with existing models
 6. [Depth2image](#depth2image)
 7. [DreamBooth](#dreambooth)
 8. [Textual Inversion](#textual-inversion)
-10. [MultiDiffusion Panorama](#panorama)
 
 ## Instruct pix2pix
 
@@ -123,12 +122,3 @@ See [here](../training/dreambooth) for more information on how to use it.
 [Textual Inversion](../training/text_inversion) fine-tunes a model to teach it about a new concept. I.e. a few pictures of a style of artwork can be used to generate images in that style.
 
 See [here](../training/text_inversion) for more information on how to use it.
-
-## MultiDiffusion Panorama
-
-[Paper](https://multidiffusion.github.io/)
-[Demo](https://huggingface.co/spaces/weizmannscience/MultiDiffusion)
-MultiDiffusion defines a new generation process over a pre-trained diffusion model. This process binds together multiple diffusion generation processes can be readily applied to generate high quality and diverse images that adhere to user-provided controls, such as desired aspect ratio (e.g., panorama), and spatial guiding signals, ranging from tight segmentation masks to bounding boxes.
-[MultiDiffusion Panorama](../api/pipelines/stable_diffusion/panorama) allows to generate high-quality images at arbitrary aspect ratios (e.g., panoramas).
-
-See [here](../api/pipelines/stable_diffusion/panorama) for more information on how to use it to generate panoramic images.

examples/community/imagic_stable_diffusion.py

@@ -22,7 +22,7 @@ from diffusers.models import AutoencoderKL, UNet2DConditionModel
 from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput
 from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
 from diffusers.schedulers import DDIMScheduler, LMSDiscreteScheduler, PNDMScheduler
-from diffusers.utils import logging
+from diffusers.utils import deprecate, logging
 
 
 if version.parse(version.parse(PIL.__version__).base_version) >= version.parse("9.1.0"):
@@ -184,6 +184,10 @@ class ImagicStableDiffusionPipeline(DiffusionPipeline):
             list of `bool`s denoting whether the corresponding generated image likely represents "not-safe-for-work"
             (nsfw) content, according to the `safety_checker`.
         """
+        message = "Please use `image` instead of `init_image`."
+        init_image = deprecate("init_image", "0.14.0", message, take_from=kwargs)
+        image = init_image or image
+
         accelerator = Accelerator(
             gradient_accumulation_steps=1,
             mixed_precision="fp16",
@@ -342,6 +346,7 @@ class ImagicStableDiffusionPipeline(DiffusionPipeline):
         return_dict: bool = True,
         guidance_scale: float = 7.5,
         eta: float = 0.0,
+        **kwargs,
     ):
         r"""
         Function invoked when calling the pipeline for generation.

examples/community/lpw_stable_diffusion.py

@@ -12,7 +12,7 @@ import diffusers
 from diffusers import SchedulerMixin, StableDiffusionPipeline
 from diffusers.models import AutoencoderKL, UNet2DConditionModel
 from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput, StableDiffusionSafetyChecker
-from diffusers.utils import logging
+from diffusers.utils import deprecate, logging
 
 
 try:
@@ -252,6 +252,7 @@ def get_weighted_text_embeddings(
     no_boseos_middle: Optional[bool] = False,
     skip_parsing: Optional[bool] = False,
     skip_weighting: Optional[bool] = False,
+    **kwargs,
 ):
     r"""
     Prompts can be assigned with local weights using brackets. For example,
@@ -681,6 +682,7 @@ class StableDiffusionLongPromptWeightingPipeline(StableDiffusionPipeline):
         callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
         is_cancelled_callback: Optional[Callable[[], bool]] = None,
         callback_steps: int = 1,
+        **kwargs,
     ):
         r"""
         Function invoked when calling the pipeline for generation.
@@ -756,6 +758,10 @@ class StableDiffusionLongPromptWeightingPipeline(StableDiffusionPipeline):
             list of `bool`s denoting whether the corresponding generated image likely represents "not-safe-for-work"
             (nsfw) content, according to the `safety_checker`.
         """
+        message = "Please use `image` instead of `init_image`."
+        init_image = deprecate("init_image", "0.14.0", message, take_from=kwargs)
+        image = init_image or image
+
         # 0. Default height and width to unet
         height = height or self.unet.config.sample_size * self.vae_scale_factor
         width = width or self.unet.config.sample_size * self.vae_scale_factor
@@ -878,6 +884,7 @@ class StableDiffusionLongPromptWeightingPipeline(StableDiffusionPipeline):
         callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
         is_cancelled_callback: Optional[Callable[[], bool]] = None,
         callback_steps: int = 1,
+        **kwargs,
     ):
         r"""
         Function for text-to-image generation.
@@ -953,6 +960,7 @@ class StableDiffusionLongPromptWeightingPipeline(StableDiffusionPipeline):
             callback=callback,
             is_cancelled_callback=is_cancelled_callback,
             callback_steps=callback_steps,
+            **kwargs,
         )
 
     def img2img(
@@ -972,6 +980,7 @@ class StableDiffusionLongPromptWeightingPipeline(StableDiffusionPipeline):
         callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
         is_cancelled_callback: Optional[Callable[[], bool]] = None,
         callback_steps: int = 1,
+        **kwargs,
     ):
         r"""
         Function for image-to-image generation.
@@ -1047,6 +1056,7 @@ class StableDiffusionLongPromptWeightingPipeline(StableDiffusionPipeline):
             callback=callback,
             is_cancelled_callback=is_cancelled_callback,
             callback_steps=callback_steps,
+            **kwargs,
         )
 
     def inpaint(
@@ -1067,6 +1077,7 @@ class StableDiffusionLongPromptWeightingPipeline(StableDiffusionPipeline):
         callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
         is_cancelled_callback: Optional[Callable[[], bool]] = None,
         callback_steps: int = 1,
+        **kwargs,
     ):
         r"""
         Function for inpaint.
@@ -1147,4 +1158,5 @@ class StableDiffusionLongPromptWeightingPipeline(StableDiffusionPipeline):
             callback=callback,
             is_cancelled_callback=is_cancelled_callback,
             callback_steps=callback_steps,
+            **kwargs,
         )

examples/community/lpw_stable_diffusion_onnx.py

@@ -11,7 +11,7 @@ from transformers import CLIPFeatureExtractor, CLIPTokenizer
 import diffusers
 from diffusers import OnnxRuntimeModel, OnnxStableDiffusionPipeline, SchedulerMixin
 from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput
-from diffusers.utils import logging
+from diffusers.utils import deprecate, logging
 
 
 try:
@@ -744,6 +744,10 @@ class OnnxStableDiffusionLongPromptWeightingPipeline(OnnxStableDiffusionPipeline
             list of `bool`s denoting whether the corresponding generated image likely represents "not-safe-for-work"
             (nsfw) content, according to the `safety_checker`.
         """
+        message = "Please use `image` instead of `init_image`."
+        init_image = deprecate("init_image", "0.14.0", message, take_from=kwargs)
+        image = init_image or image
+
         # 0. Default height and width to unet
         height = height or self.unet.config.sample_size * self.vae_scale_factor
         width = width or self.unet.config.sample_size * self.vae_scale_factor

examples/dreambooth/train_dreambooth.py

@@ -47,7 +47,7 @@ from diffusers.utils.import_utils import is_xformers_available
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.14.0.dev0")
+check_min_version("0.13.0")
 
 logger = get_logger(__name__)
 

examples/dreambooth/train_dreambooth_flax.py

@@ -36,7 +36,7 @@ from diffusers.utils import check_min_version
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.14.0.dev0")
+check_min_version("0.13.0")
 
 # Cache compiled models across invocations of this script.
 cc.initialize_cache(os.path.expanduser("~/.cache/jax/compilation_cache"))

examples/dreambooth/train_dreambooth_lora.py

@@ -54,7 +54,7 @@ from diffusers.utils.import_utils import is_xformers_available
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.14.0.dev0")
+check_min_version("0.13.0")
 
 logger = get_logger(__name__)
 

examples/text_to_image/train_text_to_image.py

@@ -47,7 +47,7 @@ from diffusers.utils.import_utils import is_xformers_available
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.14.0.dev0")
+check_min_version("0.13.0")
 
 logger = get_logger(__name__, log_level="INFO")
 

examples/text_to_image/train_text_to_image_flax.py

@@ -34,7 +34,7 @@ from diffusers.utils import check_min_version
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.14.0.dev0")
+check_min_version("0.13.0")
 
 logger = logging.getLogger(__name__)
 

examples/text_to_image/train_text_to_image_lora.py

@@ -48,7 +48,7 @@ from diffusers.utils.import_utils import is_xformers_available
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.14.0.dev0")
+check_min_version("0.13.0")
 
 logger = get_logger(__name__, log_level="INFO")
 
@@ -448,6 +448,19 @@ def main():
     vae.to(accelerator.device, dtype=weight_dtype)
     text_encoder.to(accelerator.device, dtype=weight_dtype)
 
+    if args.enable_xformers_memory_efficient_attention:
+        if is_xformers_available():
+            import xformers
+
+            xformers_version = version.parse(xformers.__version__)
+            if xformers_version == version.parse("0.0.16"):
+                logger.warn(
+                    "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
+                )
+            unet.enable_xformers_memory_efficient_attention()
+        else:
+            raise ValueError("xformers is not available. Make sure it is installed correctly")
+
     # now we will add new LoRA weights to the attention layers
     # It's important to realize here how many attention weights will be added and of which sizes
     # The sizes of the attention layers consist only of two different variables:
@@ -479,20 +492,6 @@ def main():
         )
 
     unet.set_attn_processor(lora_attn_procs)
-
-    if args.enable_xformers_memory_efficient_attention:
-        if is_xformers_available():
-            import xformers
-
-            xformers_version = version.parse(xformers.__version__)
-            if xformers_version == version.parse("0.0.16"):
-                logger.warn(
-                    "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
-                )
-            unet.enable_xformers_memory_efficient_attention()
-        else:
-            raise ValueError("xformers is not available. Make sure it is installed correctly")
-
     lora_layers = AttnProcsLayers(unet.attn_processors)
 
     # Enable TF32 for faster training on Ampere GPUs,

examples/textual_inversion/textual_inversion.py

@@ -74,7 +74,7 @@ else:
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.14.0.dev0")
+check_min_version("0.13.0")
 
 logger = get_logger(__name__)
 

examples/textual_inversion/textual_inversion_flax.py

@@ -57,7 +57,7 @@ else:
 # ------------------------------------------------------------------------------
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.14.0.dev0")
+check_min_version("0.13.0")
 
 logger = logging.getLogger(__name__)
 

examples/unconditional_image_generation/train_unconditional.py

@@ -23,11 +23,11 @@ import diffusers
 from diffusers import DDPMPipeline, DDPMScheduler, UNet2DModel
 from diffusers.optimization import get_scheduler
 from diffusers.training_utils import EMAModel
-from diffusers.utils import check_min_version, is_accelerate_version, is_tensorboard_available, is_wandb_available
+from diffusers.utils import check_min_version, is_tensorboard_available, is_wandb_available
 
 
 # Will error if the minimal version of diffusers is not installed. Remove at your own risks.
-check_min_version("0.14.0.dev0")
+check_min_version("0.13.0")
 
 logger = get_logger(__name__, log_level="INFO")
 
@@ -628,13 +628,10 @@ def main(args):
                 images_processed = (images * 255).round().astype("uint8")
 
                 if args.logger == "tensorboard":
-                    if is_accelerate_version(">=", "0.17.0.dev0"):
-                        tracker = accelerator.get_tracker("tensorboard", unwrap=True)
-                    else:
-                        tracker = accelerator.get_tracker()
-                    tracker.add_images("test_samples", images_processed.transpose(0, 3, 1, 2), epoch)
+                    accelerator.get_tracker("tensorboard").add_images(
+                        "test_samples", images_processed.transpose(0, 3, 1, 2), epoch
+                    )
                 elif args.logger == "wandb":
-                    # Upcoming `log_images` helper coming in https://github.com/huggingface/accelerate/pull/962/files
                     accelerator.get_tracker("wandb").log(
                         {"test_samples": [wandb.Image(img) for img in images_processed], "epoch": epoch},
                         step=global_step,

setup.py

@@ -219,7 +219,7 @@ install_requires = [
 
 setup(
     name="diffusers",
-    version="0.14.0.dev0",  # expected format is one of x.y.z.dev0, or x.y.z.rc1 or x.y.z (no to dashes, yes to dots)
+    version="0.13.1",  # expected format is one of x.y.z.dev0, or x.y.z.rc1 or x.y.z (no to dashes, yes to dots)
     description="Diffusers",
     long_description=open("README.md", "r", encoding="utf-8").read(),
     long_description_content_type="text/markdown",

src/diffusers/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "0.14.0.dev0"
+__version__ = "0.13.1"
 
 from .configuration_utils import ConfigMixin
 from .utils import (

src/diffusers/models/attention.py

@@ -18,7 +18,6 @@ import torch
 import torch.nn.functional as F
 from torch import nn
 
-from ..utils import logging
 from ..utils.import_utils import is_xformers_available
 from .cross_attention import CrossAttention
 from .embeddings import CombinedTimestepLabelEmbeddings
@@ -30,8 +29,6 @@ if is_xformers_available():
 else:
     xformers = None
 
-logger = logging.get_logger(__name__)  # pylint: disable=invalid-name
-
 
 class AttentionBlock(nn.Module):
     """
@@ -211,7 +208,6 @@ class BasicTransformerBlock(nn.Module):
         final_dropout: bool = False,
     ):
         super().__init__()
-        print(f"Using {activation_fn} as activation_fn in BasicTransformerBlock.")
         self.only_cross_attention = only_cross_attention
 
         self.use_ada_layer_norm_zero = (num_embeds_ada_norm is not None) and norm_type == "ada_norm_zero"
@@ -357,22 +353,15 @@ class FeedForward(nn.Module):
         super().__init__()
         inner_dim = int(dim * mult)
         dim_out = dim_out if dim_out is not None else dim
-        use_bias = True
 
         if activation_fn == "gelu":
             act_fn = GELU(dim, inner_dim)
         if activation_fn == "gelu-approximate":
             act_fn = GELU(dim, inner_dim, approximate="tanh")
         elif activation_fn == "geglu":
-            print("Using GEGLU as the activation function in the FFN.")
             act_fn = GEGLU(dim, inner_dim)
         elif activation_fn == "geglu-approximate":
             act_fn = ApproximateGELU(dim, inner_dim)
-        elif activation_fn == "swiglu":
-            print("Using SwiGLU as the activation function in the FFN.")
-            inner_dim = int(2 * dim_out / 3)
-            act_fn = SwiGLU(dim, inner_dim)
-            use_bias = False
 
         self.net = nn.ModuleList([])
         # project in
@@ -380,7 +369,7 @@ class FeedForward(nn.Module):
         # project dropout
         self.net.append(nn.Dropout(dropout))
         # project out
-        self.net.append(nn.Linear(inner_dim, dim_out, bias=use_bias))
+        self.net.append(nn.Linear(inner_dim, dim_out))
         # FF as used in Vision Transformer, MLP-Mixer, etc. have a final dropout
         if final_dropout:
             self.net.append(nn.Dropout(dropout))
@@ -453,22 +442,6 @@ class ApproximateGELU(nn.Module):
         return x * torch.sigmoid(1.702 * x)
 
 
-class SwiGLU(nn.Module):
-    """
-    GEGLU-like that uses SiLU instead of GELU on the gates. SwiGLU is used in works like PaLM.
-
-    Reference: https://arxiv.org/abs/2002.05202
-    """
-
-    def __init__(self, dim_in: int, dim_out: int):
-        super().__init__()
-        self.w1 = nn.Linear(dim_in, dim_out, bias=False)
-        self.w3 = nn.Linear(dim_in, dim_out, bias=False)
-
-    def forward(self, x):
-        return F.silu(self.w1(x)) * self.w3(x)
-
-
 class AdaLayerNorm(nn.Module):
     """
     Norm layer modified to incorporate timestep embeddings.

src/diffusers/models/transformer_2d.py

@@ -104,7 +104,6 @@ class Transformer2DModel(ModelMixin, ConfigMixin):
         self.num_attention_heads = num_attention_heads
         self.attention_head_dim = attention_head_dim
         inner_dim = num_attention_heads * attention_head_dim
-        print(f"Using {activation_fn} as activation_fn in Transformer2DModel.")
 
         # 1. Transformer2DModel can process both standard continous images of shape `(batch_size, num_channels, width, height)` as well as quantized image embeddings of shape `(batch_size, num_image_vectors)`
         # Define whether input is continuous or discrete depending on configuration

src/diffusers/models/unet_2d_blocks.py

@@ -42,7 +42,6 @@ def get_down_block(
     only_cross_attention=False,
     upcast_attention=False,
     resnet_time_scale_shift="default",
-    ff_activation_fn="geglu",
 ):
     down_block_type = down_block_type[7:] if down_block_type.startswith("UNetRes") else down_block_type
     if down_block_type == "DownBlock2D":
@@ -104,7 +103,6 @@ def get_down_block(
             only_cross_attention=only_cross_attention,
             upcast_attention=upcast_attention,
             resnet_time_scale_shift=resnet_time_scale_shift,
-            ff_activation_fn=ff_activation_fn,
         )
     elif down_block_type == "SimpleCrossAttnDownBlock2D":
         if cross_attention_dim is None:
@@ -216,7 +214,6 @@ def get_up_block(
     only_cross_attention=False,
     upcast_attention=False,
     resnet_time_scale_shift="default",
-    ff_activation_fn="geglu",
 ):
     up_block_type = up_block_type[7:] if up_block_type.startswith("UNetRes") else up_block_type
     if up_block_type == "UpBlock2D":
@@ -265,7 +262,6 @@ def get_up_block(
             only_cross_attention=only_cross_attention,
             upcast_attention=upcast_attention,
             resnet_time_scale_shift=resnet_time_scale_shift,
-            ff_activation_fn=ff_activation_fn,
         )
     elif up_block_type == "SimpleCrossAttnUpBlock2D":
         if cross_attention_dim is None:
@@ -469,10 +465,8 @@ class UNetMidBlock2DCrossAttn(nn.Module):
         dual_cross_attention=False,
         use_linear_projection=False,
         upcast_attention=False,
-        ff_activation_fn="geglu",
     ):
         super().__init__()
-        print(f"Using {ff_activation_fn} as ff_activation_fn in UNetMidBlock2DCrossAttn")
 
         self.has_cross_attention = True
         self.attn_num_head_channels = attn_num_head_channels
@@ -507,7 +501,6 @@ class UNetMidBlock2DCrossAttn(nn.Module):
                         norm_num_groups=resnet_groups,
                         use_linear_projection=use_linear_projection,
                         upcast_attention=upcast_attention,
-                        activation_fn=ff_activation_fn,
                     )
                 )
             else:
@@ -519,7 +512,6 @@ class UNetMidBlock2DCrossAttn(nn.Module):
                         num_layers=1,
                         cross_attention_dim=cross_attention_dim,
                         norm_num_groups=resnet_groups,
-                        activation_fn=ff_activation_fn,
                     )
                 )
             resnets.append(
@@ -750,7 +742,6 @@ class CrossAttnDownBlock2D(nn.Module):
         use_linear_projection=False,
         only_cross_attention=False,
         upcast_attention=False,
-        ff_activation_fn="geglu",
     ):
         super().__init__()
         resnets = []
@@ -787,7 +778,6 @@ class CrossAttnDownBlock2D(nn.Module):
                         use_linear_projection=use_linear_projection,
                         only_cross_attention=only_cross_attention,
                         upcast_attention=upcast_attention,
-                        activation_fn=ff_activation_fn,
                     )
                 )
             else:
@@ -799,7 +789,6 @@ class CrossAttnDownBlock2D(nn.Module):
                         num_layers=1,
                         cross_attention_dim=cross_attention_dim,
                         norm_num_groups=resnet_groups,
-                        activation_fn=ff_activation_fn,
                     )
                 )
         self.attentions = nn.ModuleList(attentions)
@@ -1723,7 +1712,6 @@ class CrossAttnUpBlock2D(nn.Module):
         use_linear_projection=False,
         only_cross_attention=False,
         upcast_attention=False,
-        ff_activation_fn="geglu",
     ):
         super().__init__()
         resnets = []
@@ -1762,7 +1750,6 @@ class CrossAttnUpBlock2D(nn.Module):
                         use_linear_projection=use_linear_projection,
                         only_cross_attention=only_cross_attention,
                         upcast_attention=upcast_attention,
-                        activation_fn=ff_activation_fn,
                     )
                 )
             else:
@@ -1774,7 +1761,6 @@ class CrossAttnUpBlock2D(nn.Module):
                         num_layers=1,
                         cross_attention_dim=cross_attention_dim,
                         norm_num_groups=resnet_groups,
-                        activation_fn=ff_activation_fn,
                     )
                 )
         self.attentions = nn.ModuleList(attentions)

src/diffusers/models/unet_2d_condition.py

@@ -148,10 +148,8 @@ class UNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoadersMixin)
         conv_in_kernel: int = 3,
         conv_out_kernel: int = 3,
         projection_class_embeddings_input_dim: Optional[int] = None,
-        ff_activation_fn="geglu",
     ):
         super().__init__()
-        print(f"Using {ff_activation_fn} as ff_activation_fn in UNet2DConditionModel")
 
         self.sample_size = sample_size
 
@@ -266,7 +264,6 @@ class UNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoadersMixin)
                 only_cross_attention=only_cross_attention[i],
                 upcast_attention=upcast_attention,
                 resnet_time_scale_shift=resnet_time_scale_shift,
-                ff_activation_fn=ff_activation_fn,
             )
             self.down_blocks.append(down_block)
 
@@ -285,7 +282,6 @@ class UNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoadersMixin)
                 dual_cross_attention=dual_cross_attention,
                 use_linear_projection=use_linear_projection,
                 upcast_attention=upcast_attention,
-                ff_activation_fn=ff_activation_fn,
             )
         elif mid_block_type == "UNetMidBlock2DSimpleCrossAttn":
             self.mid_block = UNetMidBlock2DSimpleCrossAttn(
@@ -345,7 +341,6 @@ class UNet2DConditionModel(ModelMixin, ConfigMixin, UNet2DConditionLoadersMixin)
                 only_cross_attention=only_cross_attention[i],
                 upcast_attention=upcast_attention,
                 resnet_time_scale_shift=resnet_time_scale_shift,
-                ff_activation_fn=ff_activation_fn,
             )
             self.up_blocks.append(up_block)
             prev_output_channel = output_channel

src/diffusers/pipelines/alt_diffusion/pipeline_alt_diffusion_img2img.py

@@ -561,6 +561,7 @@ class AltDiffusionImg2ImgPipeline(DiffusionPipeline):
         return_dict: bool = True,
         callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
         callback_steps: int = 1,
+        **kwargs,
     ):
         r"""
         Function invoked when calling the pipeline for generation.
@@ -627,6 +628,10 @@ class AltDiffusionImg2ImgPipeline(DiffusionPipeline):
             list of `bool`s denoting whether the corresponding generated image likely represents "not-safe-for-work"
             (nsfw) content, according to the `safety_checker`.
         """
+        message = "Please use `image` instead of `init_image`."
+        init_image = deprecate("init_image", "0.14.0", message, take_from=kwargs)
+        image = init_image or image
+
         # 1. Check inputs. Raise error if not correct
         self.check_inputs(prompt, strength, callback_steps, negative_prompt, prompt_embeds, negative_prompt_embeds)
 

src/diffusers/pipelines/latent_diffusion/pipeline_latent_diffusion_superresolution.py

@@ -15,7 +15,7 @@ from ...schedulers import (
     LMSDiscreteScheduler,
     PNDMScheduler,
 )
-from ...utils import PIL_INTERPOLATION, randn_tensor
+from ...utils import PIL_INTERPOLATION, deprecate, randn_tensor
 from ..pipeline_utils import DiffusionPipeline, ImagePipelineOutput
 
 
@@ -72,6 +72,7 @@ class LDMSuperResolutionPipeline(DiffusionPipeline):
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
         output_type: Optional[str] = "pil",
         return_dict: bool = True,
+        **kwargs,
     ) -> Union[Tuple, ImagePipelineOutput]:
         r"""
         Args:
@@ -99,6 +100,10 @@ class LDMSuperResolutionPipeline(DiffusionPipeline):
             [`~pipelines.ImagePipelineOutput`] or `tuple`: [`~pipelines.utils.ImagePipelineOutput`] if `return_dict` is
             True, otherwise a `tuple. When returning a tuple, the first element is a list with the generated images.
         """
+        message = "Please use `image` instead of `init_image`."
+        init_image = deprecate("init_image", "0.14.0", message, take_from=kwargs)
+        image = init_image or image
+
         if isinstance(image, PIL.Image.Image):
             batch_size = 1
         elif isinstance(image, torch.Tensor):

src/diffusers/pipelines/stable_diffusion/pipeline_cycle_diffusion.py

@@ -582,6 +582,7 @@ class CycleDiffusionPipeline(DiffusionPipeline):
         return_dict: bool = True,
         callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
         callback_steps: int = 1,
+        **kwargs,
     ):
         r"""
         Function invoked when calling the pipeline for generation.
@@ -645,6 +646,10 @@ class CycleDiffusionPipeline(DiffusionPipeline):
             list of `bool`s denoting whether the corresponding generated image likely represents "not-safe-for-work"
             (nsfw) content, according to the `safety_checker`.
         """
+        message = "Please use `image` instead of `init_image`."
+        init_image = deprecate("init_image", "0.14.0", message, take_from=kwargs)
+        image = init_image or image
+
         # 1. Check inputs
         self.check_inputs(prompt, strength, callback_steps)
 

src/diffusers/pipelines/stable_diffusion/pipeline_onnx_stable_diffusion_img2img.py

@@ -253,6 +253,7 @@ class OnnxStableDiffusionImg2ImgPipeline(DiffusionPipeline):
         return_dict: bool = True,
         callback: Optional[Callable[[int, int, np.ndarray], None]] = None,
         callback_steps: int = 1,
+        **kwargs,
     ):
         r"""
         Function invoked when calling the pipeline for generation.
@@ -308,6 +309,10 @@ class OnnxStableDiffusionImg2ImgPipeline(DiffusionPipeline):
             list of `bool`s denoting whether the corresponding generated image likely represents "not-safe-for-work"
             (nsfw) content, according to the `safety_checker`.
         """
+        message = "Please use `image` instead of `init_image`."
+        init_image = deprecate("init_image", "0.14.0", message, take_from=kwargs)
+        image = init_image or image
+
         if isinstance(prompt, str):
             batch_size = 1
         elif isinstance(prompt, list):

src/diffusers/pipelines/stable_diffusion/pipeline_onnx_stable_diffusion_inpaint_legacy.py

@@ -240,6 +240,7 @@ class OnnxStableDiffusionInpaintPipelineLegacy(DiffusionPipeline):
         return_dict: bool = True,
         callback: Optional[Callable[[int, int, np.ndarray], None]] = None,
         callback_steps: int = 1,
+        **kwargs,
     ):
         r"""
         Function invoked when calling the pipeline for generation.
@@ -300,6 +301,10 @@ class OnnxStableDiffusionInpaintPipelineLegacy(DiffusionPipeline):
             list of `bool`s denoting whether the corresponding generated image likely represents "not-safe-for-work"
             (nsfw) content, according to the `safety_checker`.
         """
+        message = "Please use `image` instead of `init_image`."
+        init_image = deprecate("init_image", "0.14.0", message, take_from=kwargs)
+        image = init_image or image
+
         if isinstance(prompt, str):
             batch_size = 1
         elif isinstance(prompt, list):

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_depth2img.py

@@ -587,26 +587,13 @@ class StableDiffusionDepth2ImgPipeline(DiffusionPipeline):
             (nsfw) content, according to the `safety_checker`.
         """
         # 1. Check inputs
-        self.check_inputs(
-            prompt,
-            strength,
-            callback_steps,
-            negative_prompt=negative_prompt,
-            prompt_embeds=prompt_embeds,
-            negative_prompt_embeds=negative_prompt_embeds,
-        )
+        self.check_inputs(prompt, strength, callback_steps)
 
         if image is None:
             raise ValueError("`image` input cannot be undefined.")
 
         # 2. Define call parameters
-        if prompt is not None and isinstance(prompt, str):
-            batch_size = 1
-        elif prompt is not None and isinstance(prompt, list):
-            batch_size = len(prompt)
-        else:
-            batch_size = prompt_embeds.shape[0]
-
+        batch_size = 1 if isinstance(prompt, str) else len(prompt)
         device = self._execution_device
         # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
         # of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1`

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py

@@ -572,6 +572,7 @@ class StableDiffusionImg2ImgPipeline(DiffusionPipeline):
         return_dict: bool = True,
         callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
         callback_steps: int = 1,
+        **kwargs,
     ):
         r"""
         Function invoked when calling the pipeline for generation.
@@ -638,6 +639,10 @@ class StableDiffusionImg2ImgPipeline(DiffusionPipeline):
             list of `bool`s denoting whether the corresponding generated image likely represents "not-safe-for-work"
             (nsfw) content, according to the `safety_checker`.
         """
+        message = "Please use `image` instead of `init_image`."
+        init_image = deprecate("init_image", "0.14.0", message, take_from=kwargs)
+        image = init_image or image
+
         # 1. Check inputs. Raise error if not correct
         self.check_inputs(prompt, strength, callback_steps, negative_prompt, prompt_embeds, negative_prompt_embeds)
 

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint_legacy.py

@@ -530,6 +530,7 @@ class StableDiffusionInpaintPipelineLegacy(DiffusionPipeline):
         return_dict: bool = True,
         callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
         callback_steps: int = 1,
+        **kwargs,
     ):
         r"""
         Function invoked when calling the pipeline for generation.
@@ -602,6 +603,10 @@ class StableDiffusionInpaintPipelineLegacy(DiffusionPipeline):
             list of `bool`s denoting whether the corresponding generated image likely represents "not-safe-for-work"
             (nsfw) content, according to the `safety_checker`.
         """
+        message = "Please use `image` instead of `init_image`."
+        init_image = deprecate("init_image", "0.14.0", message, take_from=kwargs)
+        image = init_image or image
+
         # 1. Check inputs
         self.check_inputs(prompt, strength, callback_steps)
 

tests/models/test_models_unet_1d.py

@@ -66,10 +66,6 @@ class UNet1DModelTests(ModelTesterMixin, unittest.TestCase):
     def test_from_save_pretrained(self):
         super().test_from_save_pretrained()
 
-    @unittest.skipIf(torch_device == "mps", "mish op not supported in MPS")
-    def test_from_save_pretrained_variant(self):
-        super().test_from_save_pretrained_variant()
-
     @unittest.skipIf(torch_device == "mps", "mish op not supported in MPS")
     def test_model_from_pretrained(self):
         super().test_model_from_pretrained()
@@ -190,10 +186,6 @@ class UNetRLModelTests(ModelTesterMixin, unittest.TestCase):
     def test_from_save_pretrained(self):
         super().test_from_save_pretrained()
 
-    @unittest.skipIf(torch_device == "mps", "mish op not supported in MPS")
-    def test_from_save_pretrained_variant(self):
-        super().test_from_save_pretrained_variant()
-
     @unittest.skipIf(torch_device == "mps", "mish op not supported in MPS")
     def test_model_from_pretrained(self):
         super().test_model_from_pretrained()

tests/pipelines/stable_diffusion/test_stable_diffusion_panorama.py

@@ -30,7 +30,7 @@ from diffusers import (
     UNet2DConditionModel,
 )
 from diffusers.utils import slow, torch_device
-from diffusers.utils.testing_utils import require_torch_gpu, skip_mps
+from diffusers.utils.testing_utils import require_torch_gpu
 
 from ...test_pipelines_common import PipelineTesterMixin
 
@@ -38,7 +38,6 @@ from ...test_pipelines_common import PipelineTesterMixin
 torch.backends.cuda.matmul.allow_tf32 = False
 
 
-@skip_mps
 class StableDiffusionPanoramaPipelineFastTests(PipelineTesterMixin, unittest.TestCase):
     pipeline_class = StableDiffusionPanoramaPipeline
 

tests/pipelines/stable_unclip/test_stable_unclip.py

@@ -189,10 +189,6 @@ class StableUnCLIPPipelineIntegrationTests(unittest.TestCase):
         pipe = StableUnCLIPPipeline.from_pretrained("fusing/stable-unclip-2-1-l", torch_dtype=torch.float16)
         pipe.to(torch_device)
         pipe.set_progress_bar_config(disable=None)
-        # stable unclip will oom when integration tests are run on a V100,
-        # so turn on memory savings
-        pipe.enable_attention_slicing()
-        pipe.enable_sequential_cpu_offload()
 
         generator = torch.Generator(device="cpu").manual_seed(0)
         output = pipe("anime turle", generator=generator, output_type="np")

tests/pipelines/stable_unclip/test_stable_unclip_img2img.py

@@ -185,10 +185,6 @@ class StableUnCLIPImg2ImgPipelineIntegrationTests(unittest.TestCase):
         )
         pipe.to(torch_device)
         pipe.set_progress_bar_config(disable=None)
-        # stable unclip will oom when integration tests are run on a V100,
-        # so turn on memory savings
-        pipe.enable_attention_slicing()
-        pipe.enable_sequential_cpu_offload()
 
         generator = torch.Generator(device="cpu").manual_seed(0)
         output = pipe("anime turle", image=input_image, generator=generator, output_type="np")
@@ -213,10 +209,6 @@ class StableUnCLIPImg2ImgPipelineIntegrationTests(unittest.TestCase):
         )
         pipe.to(torch_device)
         pipe.set_progress_bar_config(disable=None)
-        # stable unclip will oom when integration tests are run on a V100,
-        # so turn on memory savings
-        pipe.enable_attention_slicing()
-        pipe.enable_sequential_cpu_offload()
 
         generator = torch.Generator(device="cpu").manual_seed(0)
         output = pipe("anime turle", image=input_image, generator=generator, output_type="np")