UniDiffuser
The UniDiffuser model was proposed in One Transformer Fits All Distributions in Multi-Modal Diffusion at Scale by Fan Bao, Shen Nie, Kaiwen Xue, Chongxuan Li, Shi Pu, Yaole Wang, Gang Yue, Yue Cao, Hang Su, Jun Zhu.
The abstract of the paper is the following:
This paper proposes a unified diffusion framework (dubbed UniDiffuser) to fit all distributions relevant to a set of multi-modal data in one model. Our key insight is — learning diffusion models for marginal, conditional, and joint distributions can be unified as predicting the noise in the perturbed data, where the perturbation levels (i.e. timesteps) can be different for different modalities. Inspired by the unified view, UniDiffuser learns all distributions simultaneously with a minimal modification to the original diffusion model — perturbs data in all modalities instead of a single modality, inputs individual timesteps in different modalities, and predicts the noise of all modalities instead of a single modality. UniDiffuser is parameterized by a transformer for diffusion models to handle input types of different modalities. Implemented on large-scale paired image-text data, UniDiffuser is able to perform image, text, text-to-image, image-to-text, and image-text pair generation by setting proper timesteps without additional overhead. In particular, UniDiffuser is able to produce perceptually realistic samples in all tasks and its quantitative results (e.g., the FID and CLIP score) are not only superior to existing general-purpose models but also comparable to the bespoken models (e.g., Stable Diffusion and DALL-E 2) in representative tasks (e.g., text-to-image generation).
Resources:
Available Checkpoints are:
- UniDiffuser-v0 (512x512 resolution) thu-ml/unidiffuser-v0
- UniDiffuser-v1 (512x512 resolution) thu-ml/unidiffuser-v1
This pipeline was contributed by our community member dg845.
Available Pipelines:
| Pipeline | Tasks | Demo | Colab |
|---|---|---|---|
| UniDiffuserPipeline | Joint Image-Text Gen, Text-to-Image, Image-to-Text, Image Gen, Text Gen, Image Variation, Text Variation |
🤗 Spaces |  |
Usage Examples
Because the UniDiffuser model is trained to model the joint distribution of (image, text) pairs, it is capable of performing a diverse range of generation tasks.
Unconditional Image and Text Generation
Unconditional generation (where we start from only latents sampled from a standard Gaussian prior) from a UniDiffuserPipeline will produce a (image, text) pair:
import torch
from diffusers import UniDiffuserPipeline
device = "cuda"
model_id_or_path = "thu-ml/unidiffuser-v1"
pipe = UniDiffuserPipeline.from_pretrained(model_id_or_path, torch_dtype=torch.float16)
pipe.to(device)
# Unconditional image and text generation. The generation task is automatically inferred.
sample = pipe(num_inference_steps=20, guidance_scale=8.0)
image = sample.images[0]
text = sample.text[0]
image.save("unidiffuser_joint_sample_image.png")
print(text)This is also called “joint” generation in the UniDiffusers paper, since we are sampling from the joint image-text distribution.
Note that the generation task is inferred from the inputs used when calling the pipeline. It is also possible to manually specify the unconditional generation task (“mode”) manually with UniDiffuserPipeline.set_joint_mode():
# Equivalent to the above.
pipe.set_joint_mode()
sample = pipe(num_inference_steps=20, guidance_scale=8.0)When the mode is set manually, subsequent calls to the pipeline will use the set mode without attempting the infer the mode. You can reset the mode with UniDiffuserPipeline.reset_mode(), after which the pipeline will once again infer the mode.
You can also generate only an image or only text (which the UniDiffuser paper calls “marginal” generation since we sample from the marginal distribution of images and text, respectively):
# Unlike other generation tasks, image-only and text-only generation don't use classifier-free guidance
# Image-only generation
pipe.set_image_mode()
sample_image = pipe(num_inference_steps=20).images[0]
# Text-only generation
pipe.set_text_mode()
sample_text = pipe(num_inference_steps=20).text[0]Text-to-Image Generation
UniDiffuser is also capable of sampling from conditional distributions; that is, the distribution of images conditioned on a text prompt or the distribution of texts conditioned on an image. Here is an example of sampling from the conditional image distribution (text-to-image generation or text-conditioned image generation):
import torch
from diffusers import UniDiffuserPipeline
device = "cuda"
model_id_or_path = "thu-ml/unidiffuser-v1"
pipe = UniDiffuserPipeline.from_pretrained(model_id_or_path, torch_dtype=torch.float16)
pipe.to(device)
# Text-to-image generation
prompt = "an elephant under the sea"
sample = pipe(prompt=prompt, num_inference_steps=20, guidance_scale=8.0)
t2i_image = sample.images[0]
t2i_image.save("unidiffuser_text2img_sample_image.png")The text2img mode requires that either an input prompt or prompt_embeds be supplied. You can set the text2img mode manually with UniDiffuserPipeline.set_text_to_image_mode().
Image-to-Text Generation
Similarly, UniDiffuser can also produce text samples given an image (image-to-text or image-conditioned text generation):
import torch
from diffusers import UniDiffuserPipeline
from diffusers.utils import load_image
device = "cuda"
model_id_or_path = "thu-ml/unidiffuser-v1"
pipe = UniDiffuserPipeline.from_pretrained(model_id_or_path, torch_dtype=torch.float16)
pipe.to(device)
# Image-to-text generation
image_url = "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/unidiffuser/unidiffuser_example_image.jpg"
init_image = load_image(image_url).resize((512, 512))
sample = pipe(image=init_image, num_inference_steps=20, guidance_scale=8.0)
i2t_text = sample.text[0]
print(i2t_text)The img2text mode requires that an input image be supplied. You can set the img2text mode manually with UniDiffuserPipeline.set_image_to_text_mode().
Image Variation
The UniDiffuser authors suggest performing image variation through a “round-trip” generation method, where given an input image, we first perform an image-to-text generation, and the perform a text-to-image generation on the outputs of the first generation. This produces a new image which is semantically similar to the input image:
import torch
from diffusers import UniDiffuserPipeline
from diffusers.utils import load_image
device = "cuda"
model_id_or_path = "thu-ml/unidiffuser-v1"
pipe = UniDiffuserPipeline.from_pretrained(model_id_or_path, torch_dtype=torch.float16)
pipe.to(device)
# Image variation can be performed with a image-to-text generation followed by a text-to-image generation:
# 1. Image-to-text generation
image_url = "https://huggingface.co/datasets/hf-internal-testing/diffusers-images/resolve/main/unidiffuser/unidiffuser_example_image.jpg"
init_image = load_image(image_url).resize((512, 512))
sample = pipe(image=init_image, num_inference_steps=20, guidance_scale=8.0)
i2t_text = sample.text[0]
print(i2t_text)
# 2. Text-to-image generation
sample = pipe(prompt=i2t_text, num_inference_steps=20, guidance_scale=8.0)
final_image = sample.images[0]
final_image.save("unidiffuser_image_variation_sample.png")Text Variation
Similarly, text variation can be performed on an input prompt with a text-to-image generation followed by a image-to-text generation:
import torch
from diffusers import UniDiffuserPipeline
device = "cuda"
model_id_or_path = "thu-ml/unidiffuser-v1"
pipe = UniDiffuserPipeline.from_pretrained(model_id_or_path, torch_dtype=torch.float16)
pipe.to(device)
# Text variation can be performed with a text-to-image generation followed by a image-to-text generation:
# 1. Text-to-image generation
prompt = "an elephant under the sea"
sample = pipe(prompt=prompt, num_inference_steps=20, guidance_scale=8.0)
t2i_image = sample.images[0]
t2i_image.save("unidiffuser_text2img_sample_image.png")
# 2. Image-to-text generation
sample = pipe(image=t2i_image, num_inference_steps=20, guidance_scale=8.0)
final_prompt = sample.text[0]
print(final_prompt)UniDiffuserPipeline
class diffusers.UniDiffuserPipeline
< source >( vae: AutoencoderKL text_encoder: CLIPTextModel image_encoder: CLIPVisionModelWithProjection image_processor: CLIPImageProcessor clip_tokenizer: CLIPTokenizer text_decoder: UniDiffuserTextDecoder text_tokenizer: GPT2Tokenizer unet: UniDiffuserModel scheduler: KarrasDiffusionSchedulers )
Parameters
- vae (AutoencoderKL) — Variational Auto-Encoder (VAE) Model to encode and decode images to and from latent representations. This is part of the UniDiffuser image representation, along with the CLIP vision encoding.
-
text_encoder (
CLIPTextModel) — Frozen text-encoder. Similar to Stable Diffusion, UniDiffuser uses the text portion of CLIP to encode text prompts. -
image_encoder (
CLIPVisionModel) — UniDiffuser uses the vision portion of CLIP to encode images as part of its image representation, along with the VAE latent representation. -
image_processor (
CLIPImageProcessor) — CLIP image processor of class CLIPImageProcessor, used to preprocess the image before CLIP encoding it withimage_encoder. -
clip_tokenizer (
CLIPTokenizer) — Tokenizer of class CLIPTokenizer which is used to tokenizer a prompt before encoding it withtext_encoder. -
text_decoder (
UniDiffuserTextDecoder) — Frozen text decoder. This is a GPT-style model which is used to generate text from the UniDiffuser embedding. -
text_tokenizer (
GPT2Tokenizer) — Tokenizer of class GPT2Tokenizer which is used along with thetext_decoderto decode text for text generation. -
unet (
UniDiffuserModel) — UniDiffuser uses a U-ViT model architecture, which is similar to a Transformer2DModel with U-Net-style skip connections between transformer layers. -
scheduler (SchedulerMixin) —
A scheduler to be used in combination with
unetto denoise the encoded image and/or text latents. The original UniDiffuser paper uses the DPMSolverMultistepScheduler scheduler.
Pipeline for a bimodal image-text UniDiffuser model, which supports unconditional text and image generation, text-conditioned image generation, image-conditioned text generation, and joint image-text generation.
This model inherits from DiffusionPipeline. Check the superclass documentation for the generic methods the library implements for all the pipelines (such as downloading or saving, running on a particular device, etc.)
__call__
< source >(
prompt: typing.Union[str, typing.List[str], NoneType] = None
image: typing.Union[torch.FloatTensor, PIL.Image.Image, NoneType] = None
height: typing.Optional[int] = None
width: typing.Optional[int] = None
data_type: typing.Optional[int] = 1
num_inference_steps: int = 50
guidance_scale: float = 8.0
negative_prompt: typing.Union[str, typing.List[str], NoneType] = None
num_images_per_prompt: typing.Optional[int] = 1
num_prompts_per_image: typing.Optional[int] = 1
eta: float = 0.0
generator: typing.Union[torch._C.Generator, typing.List[torch._C.Generator], NoneType] = None
latents: typing.Optional[torch.FloatTensor] = None
prompt_latents: typing.Optional[torch.FloatTensor] = None
vae_latents: typing.Optional[torch.FloatTensor] = None
clip_latents: typing.Optional[torch.FloatTensor] = None
prompt_embeds: typing.Optional[torch.FloatTensor] = None
negative_prompt_embeds: typing.Optional[torch.FloatTensor] = None
output_type: typing.Optional[str] = 'pil'
return_dict: bool = True
callback: typing.Union[typing.Callable[[int, int, torch.FloatTensor], NoneType], NoneType] = None
callback_steps: int = 1
)
→
ImageTextPipelineOutput or tuple
Parameters
-
prompt (
strorList[str], optional) — The prompt or prompts to guide the image generation. If not defined, one has to passprompt_embedsinstead. Required for text-conditioned image generation (text2img) mode. -
image (
torch.FloatTensororPIL.Image.Image, optional) —Image, or tensor representing an image batch. Required for image-conditioned text generation (img2text) mode. -
height (
int, optional, defaults toself.unet.config.sample_size * self.vae_scale_factor) — The height in pixels of the generated image. -
width (
int, optional, defaults toself.unet.config.sample_size * self.vae_scale_factor) — The width in pixels of the generated image. -
data_type (
int, optional, defaults to 1) — The data type (either 0 or 1). Only used if you are loading a checkpoint which supports a data type embedding; this is added for compatibility with the UniDiffuser-v1 checkpoint. -
num_inference_steps (
int, optional, defaults to 50) — The number of denoising steps. More denoising steps usually lead to a higher quality image at the expense of slower inference. -
guidance_scale (
float, optional, defaults to 8.0) — Guidance scale as defined in Classifier-Free Diffusion Guidance.guidance_scaleis defined aswof equation 2. of Imagen Paper. Guidance scale is enabled by settingguidance_scale > 1. Higher guidance scale encourages to generate images that are closely linked to the textprompt, usually at the expense of lower image quality. Note that the original UniDiffuser paper uses a different definition of the guidance scalew', which satisfiesw = w' + 1. -
negative_prompt (
strorList[str], optional) — The prompt or prompts not to guide the image generation. If not defined, one has to passnegative_prompt_embedsinstead. Ignored when not using guidance (i.e., ignored ifguidance_scaleis less than1). Used in text-conditioned image generation (text2img) mode. -
num_images_per_prompt (
int, optional, defaults to 1) — The number of images to generate per prompt. Used intext2img(text-conditioned image generation) andimgmode. If the mode is joint and bothnum_images_per_promptandnum_prompts_per_imageare supplied,min(num_images_per_prompt, num_prompts_per_image)samples will be generated. -
num_prompts_per_image (
int, optional, defaults to 1) — The number of prompts to generate per image. Used inimg2text(image-conditioned text generation) andtextmode. If the mode is joint and bothnum_images_per_promptandnum_prompts_per_imageare supplied,min(num_images_per_prompt, num_prompts_per_image)samples will be generated. -
eta (
float, optional, defaults to 0.0) — Corresponds to parameter eta (η) in the DDIM paper: https://arxiv.org/abs/2010.02502. Only applies to schedulers.DDIMScheduler, will be ignored for others. -
generator (
torch.GeneratororList[torch.Generator], optional) — One or a list of torch generator(s) to make generation deterministic. -
latents (
torch.FloatTensor, optional) — Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for joint image-text generation. Can be used to tweak the same generation with different prompts. If not provided, a latents tensor will be generated by sampling using the supplied randomgenerator. Note that this is assumed to be a full set of VAE, CLIP, and text latents, if supplied, this will override the value ofprompt_latents,vae_latents, andclip_latents. -
prompt_latents (
torch.FloatTensor, optional) — Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for text generation. Can be used to tweak the same generation with different prompts. If not provided, a latents tensor will be generated by sampling using the supplied randomgenerator. -
vae_latents (
torch.FloatTensor, optional) — Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image generation. Can be used to tweak the same generation with different prompts. If not provided, a latents tensor will be generated by sampling using the supplied randomgenerator. -
clip_latents (
torch.FloatTensor, optional) — Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image generation. Can be used to tweak the same generation with different prompts. If not provided, a latents tensor will be generated by sampling using the supplied randomgenerator. -
prompt_embeds (
torch.FloatTensor, optional) — Pre-generated text embeddings. Can be used to easily tweak text inputs, e.g. prompt weighting. If not provided, text embeddings will be generated frompromptinput argument. Used in text-conditioned image generation (text2img) mode. -
negative_prompt_embeds (
torch.FloatTensor, optional) — Pre-generated negative text embeddings. Can be used to easily tweak text inputs, e.g. prompt weighting. If not provided, negative_prompt_embeds will be generated fromnegative_promptinput argument. Used in text-conditioned image generation (text2img) mode. -
output_type (
str, optional, defaults to"pil") — The output format of the generate image. Choose between PIL:PIL.Image.Imageornp.array. -
return_dict (
bool, optional, defaults toTrue) — Whether or not to return a ImageTextPipelineOutput instead of a plain tuple. -
callback (
Callable, optional) — A function that will be called everycallback_stepssteps during inference. The function will be called with the following arguments:callback(step: int, timestep: int, latents: torch.FloatTensor). -
callback_steps (
int, optional, defaults to 1) — The frequency at which thecallbackfunction will be called. If not specified, the callback will be called at every step.
Returns
ImageTextPipelineOutput or tuple
pipelines.unidiffuser.ImageTextPipelineOutput if return_dict is True, otherwise a tuple. When
returning a tuple, the first element is a list with the generated images, and the second element is a list
of generated texts.
Function invoked when calling the pipeline for generation.
Offloads all models to CPU using accelerate, reducing memory usage with a low impact on performance. Compared
to enable_sequential_cpu_offload, this method moves one whole model at a time to the GPU when its forward
method is called, and the model remains in GPU until the next model runs. Memory savings are lower than with
enable_sequential_cpu_offload, but performance is much better due to the iterative execution of the unet.
Offloads all models to CPU using accelerate, significantly reducing memory usage. When called, unet,
text_encoder, vae and safety checker have their state dicts saved to CPU and then are moved to a
torch.device('meta') and loaded to GPU only when their specific submodule has its forward method called.
Note that offloading happens on a submodule basis. Memory savings are higher than with
enable_model_cpu_offload, but performance is lower.
Removes a manually set mode; after calling this, the pipeline will infer the mode from inputs.
Manually set the generation mode to unconditional (“marginal”) image generation.
Manually set the generation mode to image-conditioned text generation.
Manually set the generation mode to unconditional joint image-text generation.
Manually set the generation mode to unconditional (“marginal”) text generation.
Manually set the generation mode to text-conditioned image generation.