Add AltDiffusion (#1299)

* add conversion script for vae

* up

* up

* some fixes

* add text model

* use the correct config

* add docs

* move model in it's own file

* move model in its own file

* pass attenion mask to text encoder

* pass attn mask to uncond inputs

* quality

* fix image2image

* add imag2image in init

* fix import

* fix one more import

* fix import, dummy objetcs

* fix copied from

* up

* finish
Co-authored-by: patil-suraj <surajp815@gmail.com>

docs/source/_toctree.yml

@@ -80,6 +80,8 @@
   - sections:
     - local: api/pipelines/overview
       title: "Overview"
+    - local: api/pipelines/alt_diffusion
+      title: "AltDiffusion"
     - local: api/pipelines/cycle_diffusion
       title: "Cycle Diffusion"
     - local: api/pipelines/ddim

docs/source/api/pipelines/alt_diffusion.mdx

+<!--Copyright 2022 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+-->
+
+# AltDiffusion
+
+AltDiffusion was proposed in [AltCLIP: Altering the Language Encoder in CLIP for Extended Language Capabilities](https://arxiv.org/abs/2211.06679) by Zhongzhi Chen, Guang Liu, Bo-Wen Zhang, Fulong Ye, Qinghong Yang, Ledell Wu
+
+The abstract of the paper is the following:
+
+*In this work, we present a conceptually simple and effective method to train a strong bilingual multimodal representation model. Starting from the pretrained multimodal representation model CLIP released by OpenAI, we switched its text encoder with a pretrained multilingual text encoder XLM-R, and aligned both languages and image representations by a two-stage training schema consisting of teacher learning and contrastive learning. We validate our method through evaluations of a wide range of tasks. We set new state-of-the-art performances on a bunch of tasks including ImageNet-CN, Flicker30k- CN, and COCO-CN. Further, we obtain very close performances with CLIP on almost all tasks, suggesting that one can simply alter the text encoder in CLIP for extended capabilities such as multilingual understanding.*
+
+
+*Overview*:
+
+| Pipeline | Tasks | Colab | Demo
+|---|---|:---:|:---:|
+| [pipeline_alt_diffusion.py](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/alt_diffusion/pipeline_alt_diffusion.py) | *Text-to-Image Generation* | - | -
+| [pipeline_alt_diffusion_img2img.py](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/alt_diffusion/pipeline_alt_diffusion_img2img.py) | *Image-to-Image Text-Guided Generation* | - |-
+
+## Tips
+
+- AltDiffusion is conceptually exaclty the same as [Stable Diffusion](./api/pipelines/stable_diffusion).
+
+- *Run AltDiffusion*
+
+AltDiffusion can be tested very easily with the [`AltDiffusionPipeline`], [`AltDiffusionImg2ImgPipeline`] and the `"BAAI/AltDiffusion"` checkpoint exactly in the same way it is shown in the [Conditional Image Generation Guide](./using-diffusers/conditional_image_generation) and the [Image-to-Image Generation Guide](./using-diffusers/img2img).
+
+- *How to load and use different schedulers.*
+
+The alt diffusion pipeline uses [`DDIMScheduler`] scheduler by default. But `diffusers` provides many other schedulers that can be used with the alt diffusion pipeline such as [`PNDMScheduler`], [`LMSDiscreteScheduler`], [`EulerDiscreteScheduler`], [`EulerAncestralDiscreteScheduler`] etc.
+To use a different scheduler, you can either change it via the [`ConfigMixin.from_config`] method or pass the `scheduler` argument to the `from_pretrained` method of the pipeline. For example, to use the [`EulerDiscreteScheduler`], you can do the following:
+
+```python
+>>> from diffusers import AltDiffusionPipeline, EulerDiscreteScheduler
+
+>>> pipeline = AltDiffusionPipeline.from_pretrained("BAAI/AltDiffusion")
+>>> pipeline.scheduler = EulerDiscreteScheduler.from_config(pipeline.scheduler.config)
+
+>>> # or
+>>> euler_scheduler = EulerDiscreteScheduler.from_pretrained("BAAI/AltDiffusion", subfolder="scheduler")
+>>> pipeline = AltDiffusionPipeline.from_pretrained("BAAI/AltDiffusion", scheduler=euler_scheduler)
+```
+
+
+- *How to conver all use cases with multiple or single pipeline*
+
+If you want to use all possible use cases in a single `DiffusionPipeline` we recommend using the `components` functionality to instantiate all components in the most memory-efficient way:
+
+```python
+>>> from diffusers import (
+...     AltDiffusionPipeline,
+...     AltDiffusionImg2ImgPipeline,
+... )
+
+>>> img2text = AltDiffusionPipeline.from_pretrained("BAAI/AltDiffusion")
+>>> img2img = AltDiffusionImg2ImgPipeline(**img2text.components)
+
+>>> # now you can use img2text(...) and img2img(...) just like the call methods of each respective pipeline
+```
+
+## AltDiffusionPipelineOutput
+[[autodoc]] pipelines.alt_diffusion.AltDiffusionPipelineOutput
+
+## AltDiffusionPipeline
+[[autodoc]] AltDiffusionPipeline
+	- __call__
+	- enable_attention_slicing
+	- disable_attention_slicing
+
+## AltDiffusionImg2ImgPipeline
+[[autodoc]] AltDiffusionImg2ImgPipeline
+	- __call__
+	- enable_attention_slicing
+	- disable_attention_slicing

docs/source/api/pipelines/overview.mdx

@@ -44,6 +44,7 @@ available a colab notebook to directly try them out.
 
 | Pipeline | Paper | Tasks | Colab
 |---|---|:---:|:---:|
+| [alt_diffusion](./api/pipelines/alt_diffusion) | [**AltDiffusion**](https://arxiv.org/abs/2211.06679) | Image-to-Image Text-Guided Generation | -
 | [cycle_diffusion](./api/pipelines/cycle_diffusion) | [**Cycle Diffusion**](https://arxiv.org/abs/2210.05559) | Image-to-Image Text-Guided Generation |
 | [dance_diffusion](./api/pipelines/dance_diffusion) | [**Dance Diffusion**](https://github.com/williamberman/diffusers.git) | Unconditional Audio Generation |
 | [ddpm](./api/pipelines/ddpm) | [**Denoising Diffusion Probabilistic Models**](https://arxiv.org/abs/2006.11239) | Unconditional Image Generation |

docs/source/index.mdx

@@ -34,6 +34,7 @@ available a colab notebook to directly try them out.
 
 | Pipeline | Paper | Tasks | Colab
 |---|---|:---:|:---:|
+| [alt_diffusion](./api/pipelines/alt_diffusion) | [**AltDiffusion**](https://arxiv.org/abs/2211.06679) | Image-to-Image Text-Guided Generation |
 | [cycle_diffusion](./api/pipelines/cycle_diffusion) | [**Cycle Diffusion**](https://arxiv.org/abs/2210.05559) | Image-to-Image Text-Guided Generation |
 | [dance_diffusion](./api/pipelines/dance_diffusion) | [**Dance Diffusion**](https://github.com/williamberman/diffusers.git) | Unconditional Audio Generation |
 | [ddpm](./api/pipelines/ddpm) | [**Denoising Diffusion Probabilistic Models**](https://arxiv.org/abs/2006.11239) | Unconditional Image Generation |

docs/source/using-diffusers/conditional_image_generation.mdx

@@ -44,5 +44,3 @@ You can save the image by simply calling:
 ```python
 >>> image.save("image_of_squirrel_painting.png")
 ```
-
-

src/diffusers/__init__.py

@@ -65,6 +65,8 @@ else:
 
 if is_torch_available() and is_transformers_available():
     from .pipelines import (
+        AltDiffusionImg2ImgPipeline,
+        AltDiffusionPipeline,
         CycleDiffusionPipeline,
         LDMTextToImagePipeline,
         StableDiffusionImg2ImgPipeline,

src/diffusers/pipelines/__init__.py

@@ -15,6 +15,7 @@ else:
     from ..utils.dummy_pt_objects import *  # noqa F403
 
 if is_torch_available() and is_transformers_available():
+    from .alt_diffusion import AltDiffusionImg2ImgPipeline, AltDiffusionPipeline
     from .latent_diffusion import LDMTextToImagePipeline
     from .stable_diffusion import (
         CycleDiffusionPipeline,

src/diffusers/pipelines/alt_diffusion/__init__.py

+from dataclasses import dataclass
+from typing import List, Optional, Union
+
+import numpy as np
+
+import PIL
+from PIL import Image
+
+from ...utils import BaseOutput, is_torch_available, is_transformers_available
+
+
+@dataclass
+# Copied from diffusers.pipelines.stable_diffusion.__init__.StableDiffusionPipelineOutput with Stable->Alt
+class AltDiffusionPipelineOutput(BaseOutput):
+    """
+    Output class for Alt Diffusion pipelines.
+
+    Args:
+        images (`List[PIL.Image.Image]` or `np.ndarray`)
+            List of denoised PIL images of length `batch_size` or numpy array of shape `(batch_size, height, width,
+            num_channels)`. PIL images or numpy array present the denoised images of the diffusion pipeline.
+        nsfw_content_detected (`List[bool]`)
+            List of flags denoting whether the corresponding generated image likely represents "not-safe-for-work"
+            (nsfw) content, or `None` if safety checking could not be performed.
+    """
+
+    images: Union[List[PIL.Image.Image], np.ndarray]
+    nsfw_content_detected: Optional[List[bool]]
+
+
+if is_transformers_available() and is_torch_available():
+    from .modeling_roberta_series import RobertaSeriesModelWithTransformation
+    from .pipeline_alt_diffusion import AltDiffusionPipeline
+    from .pipeline_alt_diffusion_img2img import AltDiffusionImg2ImgPipeline

src/diffusers/pipelines/alt_diffusion/modeling_roberta_series.py

+from dataclasses import dataclass
+from typing import Optional, Tuple
+
+import torch
+from torch import nn
+
+from transformers import RobertaPreTrainedModel, XLMRobertaConfig, XLMRobertaModel
+from transformers.utils import ModelOutput
+
+
+@dataclass
+class TransformationModelOutput(ModelOutput):
+    """
+    Base class for text model's outputs that also contains a pooling of the last hidden states.
+
+    Args:
+        text_embeds (`torch.FloatTensor` of shape `(batch_size, output_dim)` *optional* returned when model is initialized with `with_projection=True`):
+            The text embeddings obtained by applying the projection layer to the pooler_output.
+        last_hidden_state (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`):
+            Sequence of hidden-states at the output of the last layer of the model.
+        hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`):
+            Tuple of `torch.FloatTensor` (one for the output of the embeddings, if the model has an embedding layer, +
+            one for the output of each layer) of shape `(batch_size, sequence_length, hidden_size)`.
+
+            Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.
+        attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`):
+            Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length,
+            sequence_length)`.
+
+            Attentions weights after the attention softmax, used to compute the weighted average in the self-attention
+            heads.
+    """
+
+    projection_state: Optional[torch.FloatTensor] = None
+    last_hidden_state: torch.FloatTensor = None
+    hidden_states: Optional[Tuple[torch.FloatTensor]] = None
+    attentions: Optional[Tuple[torch.FloatTensor]] = None
+
+
+class RobertaSeriesConfig(XLMRobertaConfig):
+    def __init__(
+        self,
+        pad_token_id=1,
+        bos_token_id=0,
+        eos_token_id=2,
+        project_dim=512,
+        pooler_fn="cls",
+        learn_encoder=False,
+        use_attention_mask=True,
+        **kwargs,
+    ):
+        super().__init__(pad_token_id=pad_token_id, bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs)
+        self.project_dim = project_dim
+        self.pooler_fn = pooler_fn
+        self.learn_encoder = learn_encoder
+        self.use_attention_mask = use_attention_mask
+
+
+class RobertaSeriesModelWithTransformation(RobertaPreTrainedModel):
+    _keys_to_ignore_on_load_unexpected = [r"pooler"]
+    _keys_to_ignore_on_load_missing = [r"position_ids", r"predictions.decoder.bias"]
+    base_model_prefix = "roberta"
+    config_class = RobertaSeriesConfig
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.roberta = XLMRobertaModel(config)
+        self.transformation = nn.Linear(config.hidden_size, config.project_dim)
+        self.post_init()
+
+    def forward(
+        self,
+        input_ids: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        token_type_ids: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        encoder_hidden_states: Optional[torch.Tensor] = None,
+        encoder_attention_mask: Optional[torch.Tensor] = None,
+        output_attentions: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+    ):
+        r""" """
+
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.base_model(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_attention_mask,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        projection_state = self.transformation(outputs.last_hidden_state)
+
+        return TransformationModelOutput(
+            projection_state=projection_state,
+            last_hidden_state=outputs.last_hidden_state,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )

src/diffusers/pipelines/stable_diffusion/pipeline_cycle_diffusion.py

@@ -301,7 +301,17 @@ class CycleDiffusionPipeline(DiffusionPipeline):
                 "The following part of your input was truncated because CLIP can only handle sequences up to"
                 f" {self.tokenizer.model_max_length} tokens: {removed_text}"
             )
-        text_embeddings = self.text_encoder(text_input_ids.to(device))[0]
+
+        if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
+            attention_mask = text_inputs.attention_mask.to(device)
+        else:
+            attention_mask = None
+
+        text_embeddings = self.text_encoder(
+            text_input_ids.to(device),
+            attention_mask=attention_mask,
+        )
+        text_embeddings = text_embeddings[0]
 
         # duplicate text embeddings for each generation per prompt, using mps friendly method
         bs_embed, seq_len, _ = text_embeddings.shape
@@ -337,7 +347,17 @@ class CycleDiffusionPipeline(DiffusionPipeline):
                 truncation=True,
                 return_tensors="pt",
             )
-            uncond_embeddings = self.text_encoder(uncond_input.input_ids.to(device))[0]
+
+            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
+                attention_mask = uncond_input.attention_mask.to(device)
+            else:
+                attention_mask = None
+
+            uncond_embeddings = self.text_encoder(
+                uncond_input.input_ids.to(device),
+                attention_mask=attention_mask,
+            )
+            uncond_embeddings = uncond_embeddings[0]
 
             # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
             seq_len = uncond_embeddings.shape[1]

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py

@@ -248,7 +248,17 @@ class StableDiffusionPipeline(DiffusionPipeline):
                 "The following part of your input was truncated because CLIP can only handle sequences up to"
                 f" {self.tokenizer.model_max_length} tokens: {removed_text}"
             )
-        text_embeddings = self.text_encoder(text_input_ids.to(device))[0]
+
+        if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
+            attention_mask = text_inputs.attention_mask.to(device)
+        else:
+            attention_mask = None
+
+        text_embeddings = self.text_encoder(
+            text_input_ids.to(device),
+            attention_mask=attention_mask,
+        )
+        text_embeddings = text_embeddings[0]
 
         # duplicate text embeddings for each generation per prompt, using mps friendly method
         bs_embed, seq_len, _ = text_embeddings.shape
@@ -284,7 +294,17 @@ class StableDiffusionPipeline(DiffusionPipeline):
                 truncation=True,
                 return_tensors="pt",
             )
-            uncond_embeddings = self.text_encoder(uncond_input.input_ids.to(device))[0]
+
+            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
+                attention_mask = uncond_input.attention_mask.to(device)
+            else:
+                attention_mask = None
+
+            uncond_embeddings = self.text_encoder(
+                uncond_input.input_ids.to(device),
+                attention_mask=attention_mask,
+            )
+            uncond_embeddings = uncond_embeddings[0]
 
             # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
             seq_len = uncond_embeddings.shape[1]

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py

@@ -268,7 +268,17 @@ class StableDiffusionImg2ImgPipeline(DiffusionPipeline):
                 "The following part of your input was truncated because CLIP can only handle sequences up to"
                 f" {self.tokenizer.model_max_length} tokens: {removed_text}"
             )
-        text_embeddings = self.text_encoder(text_input_ids.to(device))[0]
+
+        if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
+            attention_mask = text_inputs.attention_mask.to(device)
+        else:
+            attention_mask = None
+
+        text_embeddings = self.text_encoder(
+            text_input_ids.to(device),
+            attention_mask=attention_mask,
+        )
+        text_embeddings = text_embeddings[0]
 
         # duplicate text embeddings for each generation per prompt, using mps friendly method
         bs_embed, seq_len, _ = text_embeddings.shape
@@ -304,7 +314,17 @@ class StableDiffusionImg2ImgPipeline(DiffusionPipeline):
                 truncation=True,
                 return_tensors="pt",
             )
-            uncond_embeddings = self.text_encoder(uncond_input.input_ids.to(device))[0]
+
+            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
+                attention_mask = uncond_input.attention_mask.to(device)
+            else:
+                attention_mask = None
+
+            uncond_embeddings = self.text_encoder(
+                uncond_input.input_ids.to(device),
+                attention_mask=attention_mask,
+            )
+            uncond_embeddings = uncond_embeddings[0]
 
             # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
             seq_len = uncond_embeddings.shape[1]

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py

@@ -261,7 +261,17 @@ class StableDiffusionInpaintPipeline(DiffusionPipeline):
                 "The following part of your input was truncated because CLIP can only handle sequences up to"
                 f" {self.tokenizer.model_max_length} tokens: {removed_text}"
             )
-        text_embeddings = self.text_encoder(text_input_ids.to(device))[0]
+
+        if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
+            attention_mask = text_inputs.attention_mask.to(device)
+        else:
+            attention_mask = None
+
+        text_embeddings = self.text_encoder(
+            text_input_ids.to(device),
+            attention_mask=attention_mask,
+        )
+        text_embeddings = text_embeddings[0]
 
         # duplicate text embeddings for each generation per prompt, using mps friendly method
         bs_embed, seq_len, _ = text_embeddings.shape
@@ -297,7 +307,17 @@ class StableDiffusionInpaintPipeline(DiffusionPipeline):
                 truncation=True,
                 return_tensors="pt",
             )
-            uncond_embeddings = self.text_encoder(uncond_input.input_ids.to(device))[0]
+
+            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
+                attention_mask = uncond_input.attention_mask.to(device)
+            else:
+                attention_mask = None
+
+            uncond_embeddings = self.text_encoder(
+                uncond_input.input_ids.to(device),
+                attention_mask=attention_mask,
+            )
+            uncond_embeddings = uncond_embeddings[0]
 
             # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
             seq_len = uncond_embeddings.shape[1]

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint_legacy.py

@@ -281,7 +281,17 @@ class StableDiffusionInpaintPipelineLegacy(DiffusionPipeline):
                 "The following part of your input was truncated because CLIP can only handle sequences up to"
                 f" {self.tokenizer.model_max_length} tokens: {removed_text}"
             )
-        text_embeddings = self.text_encoder(text_input_ids.to(device))[0]
+
+        if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
+            attention_mask = text_inputs.attention_mask.to(device)
+        else:
+            attention_mask = None
+
+        text_embeddings = self.text_encoder(
+            text_input_ids.to(device),
+            attention_mask=attention_mask,
+        )
+        text_embeddings = text_embeddings[0]
 
         # duplicate text embeddings for each generation per prompt, using mps friendly method
         bs_embed, seq_len, _ = text_embeddings.shape
@@ -317,7 +327,17 @@ class StableDiffusionInpaintPipelineLegacy(DiffusionPipeline):
                 truncation=True,
                 return_tensors="pt",
             )
-            uncond_embeddings = self.text_encoder(uncond_input.input_ids.to(device))[0]
+
+            if hasattr(self.text_encoder.config, "use_attention_mask") and self.text_encoder.config.use_attention_mask:
+                attention_mask = uncond_input.attention_mask.to(device)
+            else:
+                attention_mask = None
+
+            uncond_embeddings = self.text_encoder(
+                uncond_input.input_ids.to(device),
+                attention_mask=attention_mask,
+            )
+            uncond_embeddings = uncond_embeddings[0]
 
             # duplicate unconditional embeddings for each generation per prompt, using mps friendly method
             seq_len = uncond_embeddings.shape[1]

src/diffusers/utils/dummy_torch_and_transformers_objects.py

@@ -4,6 +4,36 @@
 from ..utils import DummyObject, requires_backends
 
 
+class AltDiffusionImg2ImgPipeline(metaclass=DummyObject):
+    _backends = ["torch", "transformers"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch", "transformers"])
+
+    @classmethod
+    def from_config(cls, *args, **kwargs):
+        requires_backends(cls, ["torch", "transformers"])
+
+    @classmethod
+    def from_pretrained(cls, *args, **kwargs):
+        requires_backends(cls, ["torch", "transformers"])
+
+
+class AltDiffusionPipeline(metaclass=DummyObject):
+    _backends = ["torch", "transformers"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch", "transformers"])
+
+    @classmethod
+    def from_config(cls, *args, **kwargs):
+        requires_backends(cls, ["torch", "transformers"])
+
+    @classmethod
+    def from_pretrained(cls, *args, **kwargs):
+        requires_backends(cls, ["torch", "transformers"])
+
+
 class CycleDiffusionPipeline(metaclass=DummyObject):
     _backends = ["torch", "transformers"]