콘텐츠로 이동

Stable Diffusion

Stable Diffusion은 텍스트 프롬프트로부터 이미지를 생성할 수 있는 텍스트-이미지 잠재 확산 모델입니다. RBLN NPU는 Optimum RBLN을 사용하여 Stable Diffusion 파이프라인을 가속화할 수 있습니다.

지원하는 파이프라인

Optimum RBLN은 여러 Stable Diffusion 파이프라인을 지원합니다:

  • 텍스트-이미지 변환(Text-to-Image): 텍스트 프롬프트에서 이미지 생성
  • 이미지-이미지 변환(Image-to-Image): 텍스트 프롬프트를 기반으로 기존 이미지 수정
  • 인페인팅(Inpainting): 텍스트 프롬프트에 따라 이미지의 마스킹된 영역 채우기

주요 클래스

중요: Guidance Scale에 따른 배치 크기 설정

배치 크기와 Guidance Scale

Stable Diffusion을 guidance scale > 1.0으로 사용할 때(기본값은 7.5), classifier-free guidance 기법으로 인해 UNet의 실제 배치 크기가 실행 시 2배가 됩니다.

RBLN NPU는 정적 그래프 컴파일을 사용하므로, 컴파일 시 UNet의 배치 크기가 실행 시 배치 크기와 일치해야 합니다. 그렇지 않으면 추론 중에 오류가 발생합니다.

기본 동작

UNet의 배치 크기를 명시적으로 지정하지 않는 경우, Optimum RBLN은 다음과 같이 동작합니다:

  • 기본 guidance scale(7.5)을 사용한다고 가정합니다
  • 자동으로 UNet의 배치 크기를 파이프라인 배치 크기의 2배로 설정합니다

기본 guidance scale(1.0보다 큰 값)을 사용할 계획이라면, 이 자동 구성이 올바르게 작동합니다. 그러나 다른 guidance scale을 사용하거나 더 많은 제어가 필요한 경우에는 UNet의 배치 크기를 명시적으로 구성해야 합니다.

예시: UNet 배치 크기 명시적 설정

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
from optimum.rbln import RBLNStableDiffusionPipelineConfig, RBLNStableDiffusionPipeline

# guidance_scale > 1.0인 경우 (기본값은 7.5)
# UNet의 배치 크기를 원하는 추론 배치 크기의 2배로 설정
config = RBLNStableDiffusionPipelineConfig(
    batch_size=2,  # 추론 배치 크기
    img_height=512,
    img_width=512,
    # UNet의 배치 크기를 2배로 구성
    unet=dict(batch_size=4)  # UNet 배치 크기를 2배로 설정
)

pipe = RBLNStableDiffusionPipeline.from_pretrained(
    "runwayml/stable-diffusion-v1-5",
    export=True,
    rbln_config=config
)

# 기본 guidance_scale=7.5로 일반 추론
prompts = ["A photo of an astronaut riding a horse on Mars", 
          "A portrait of a cat wearing a space suit"]
images = pipe(prompts).images  # 배치 크기 2

# 생성된 이미지 저장
for i, image in enumerate(images):
    image.save(f"생성된_이미지_{i}.png")
    print(f"이미지 {i}가 생성된_이미지_{i}.png로 저장되었습니다")

예시: Guidance Scale 1.0 사용

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
from optimum.rbln import RBLNStableDiffusionPipelineConfig, RBLNStableDiffusionPipeline

config = RBLNStableDiffusionPipelineConfig(
    img_height=512,
    img_width=512,
    # UNet 배치 크기가 추론 배치 크기와 일치
    unet=dict(batch_size=1) 
)

pipe = RBLNStableDiffusionPipeline.from_pretrained(
    "runwayml/stable-diffusion-v1-5",
    export=True,
    rbln_config=config
)

# 추론 시 반드시 guidance_scale=1.0 사용
prompt = "A photo of an astronaut riding a horse on Mars"
image = pipe(prompt, guidance_scale=1.0).images[0]

# 생성된 이미지 저장
image.save("가이던스_없는_이미지.png")
print("이미지가 가이던스_없는_이미지.png로 저장되었습니다")

사용 예제

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
from optimum.rbln import RBLNStableDiffusionPipeline, RBLNStableDiffusionPipelineConfig

# Create a configuration object with specific settings
config = RBLNStableDiffusionPipelineConfig(
    img_height=512,
    img_width=512
)

# Load and compile the model for RBLN NPU
pipe = RBLNStableDiffusionPipeline.from_pretrained(
    "runwayml/stable-diffusion-v1-5",
    export=True,
    rbln_config=config
)

# Generate an image
prompt = "A photo of an astronaut riding a horse on Mars"
image = pipe(prompt).images[0]

# 생성된 이미지 저장
image.save("우주비행사_화성.png")
print("이미지가 우주비행사_화성.png로 저장되었습니다")

API 참조

Classes

RBLNStableDiffusionPipeline

Bases: RBLNDiffusionMixin, StableDiffusionPipeline

Functions

from_pretrained(model_id, *, export=False, model_save_dir=None, rbln_config={}, lora_ids=None, lora_weights_names=None, lora_scales=None, **kwargs) classmethod

Load a pretrained diffusion pipeline from a model checkpoint, with optional compilation for RBLN NPUs.

This method has two distinct operating modes:

  • When export=True: Takes a PyTorch-based diffusion model, compiles it for RBLN NPUs, and loads the compiled model
  • When export=False: Loads an already compiled RBLN model from model_id without recompilation

It supports various diffusion pipelines including Stable Diffusion, Kandinsky, ControlNet, and other diffusers-based models.

Parameters:

Name Type Description Default
model_id str

The model ID or path to the pretrained model to load. Can be either:

  • A model ID from the Hugging Face Hub
  • A local path to a saved model directory
 required
export bool

If True, takes a PyTorch model from model_id and compiles it for RBLN NPU execution. If False, loads an already compiled RBLN model from model_id without recompilation.

 False
model_save_dir Optional[PathLike]

Directory to save the compiled model artifacts. Only used when export=True. If not provided and export=True, a temporary directory is used.

 None
rbln_config Dict[str, Any]

Configuration options for RBLN compilation. Can include settings for specific submodules such as text_encoder, unet, and vae. Configuration can be tailored to the specific pipeline being compiled.

 {}
lora_ids Optional[Union[str, List[str]]]

LoRA adapter ID(s) to load and apply before compilation. LoRA weights are fused into the model weights during compilation. Only used when export=True.

 None
lora_weights_names Optional[Union[str, List[str]]]

Names of specific LoRA weight files to load, corresponding to lora_ids. Only used when export=True.

 None
lora_scales Optional[Union[float, List[float]]]

Scaling factor(s) to apply to the LoRA adapter(s). Only used when export=True.

 None
**kwargs Dict[str, Any]

Additional arguments to pass to the underlying diffusion pipeline constructor or the RBLN compilation process. These may include parameters specific to individual submodules or the particular diffusion pipeline being used.

 {}

Returns:

Type Description
Self

A compiled diffusion pipeline that can be used for inference on RBLN NPU. The returned object is an instance of the class that called this method, inheriting from RBLNDiffusionMixin.

RBLNStableDiffusionInpaintPipeline

Bases: RBLNDiffusionMixin, StableDiffusionInpaintPipeline

Functions

from_pretrained(model_id, *, export=False, model_save_dir=None, rbln_config={}, lora_ids=None, lora_weights_names=None, lora_scales=None, **kwargs) classmethod

Load a pretrained diffusion pipeline from a model checkpoint, with optional compilation for RBLN NPUs.

This method has two distinct operating modes:

  • When export=True: Takes a PyTorch-based diffusion model, compiles it for RBLN NPUs, and loads the compiled model
  • When export=False: Loads an already compiled RBLN model from model_id without recompilation

It supports various diffusion pipelines including Stable Diffusion, Kandinsky, ControlNet, and other diffusers-based models.

Parameters:

Name Type Description Default
model_id str

The model ID or path to the pretrained model to load. Can be either:

  • A model ID from the Hugging Face Hub
  • A local path to a saved model directory
 required
export bool

If True, takes a PyTorch model from model_id and compiles it for RBLN NPU execution. If False, loads an already compiled RBLN model from model_id without recompilation.

 False
model_save_dir Optional[PathLike]

Directory to save the compiled model artifacts. Only used when export=True. If not provided and export=True, a temporary directory is used.

 None
rbln_config Dict[str, Any]

Configuration options for RBLN compilation. Can include settings for specific submodules such as text_encoder, unet, and vae. Configuration can be tailored to the specific pipeline being compiled.

 {}
lora_ids Optional[Union[str, List[str]]]

LoRA adapter ID(s) to load and apply before compilation. LoRA weights are fused into the model weights during compilation. Only used when export=True.

 None
lora_weights_names Optional[Union[str, List[str]]]

Names of specific LoRA weight files to load, corresponding to lora_ids. Only used when export=True.

 None
lora_scales Optional[Union[float, List[float]]]

Scaling factor(s) to apply to the LoRA adapter(s). Only used when export=True.

 None
**kwargs Dict[str, Any]

Additional arguments to pass to the underlying diffusion pipeline constructor or the RBLN compilation process. These may include parameters specific to individual submodules or the particular diffusion pipeline being used.

 {}

Returns:

Type Description
Self

A compiled diffusion pipeline that can be used for inference on RBLN NPU. The returned object is an instance of the class that called this method, inheriting from RBLNDiffusionMixin.

RBLNStableDiffusionImg2ImgPipeline

Bases: RBLNDiffusionMixin, StableDiffusionImg2ImgPipeline

Functions

from_pretrained(model_id, *, export=False, model_save_dir=None, rbln_config={}, lora_ids=None, lora_weights_names=None, lora_scales=None, **kwargs) classmethod

Load a pretrained diffusion pipeline from a model checkpoint, with optional compilation for RBLN NPUs.

This method has two distinct operating modes:

  • When export=True: Takes a PyTorch-based diffusion model, compiles it for RBLN NPUs, and loads the compiled model
  • When export=False: Loads an already compiled RBLN model from model_id without recompilation

It supports various diffusion pipelines including Stable Diffusion, Kandinsky, ControlNet, and other diffusers-based models.

Parameters:

Name Type Description Default
model_id str

The model ID or path to the pretrained model to load. Can be either:

  • A model ID from the Hugging Face Hub
  • A local path to a saved model directory
 required
export bool

If True, takes a PyTorch model from model_id and compiles it for RBLN NPU execution. If False, loads an already compiled RBLN model from model_id without recompilation.

 False
model_save_dir Optional[PathLike]

Directory to save the compiled model artifacts. Only used when export=True. If not provided and export=True, a temporary directory is used.

 None
rbln_config Dict[str, Any]

Configuration options for RBLN compilation. Can include settings for specific submodules such as text_encoder, unet, and vae. Configuration can be tailored to the specific pipeline being compiled.

 {}
lora_ids Optional[Union[str, List[str]]]

LoRA adapter ID(s) to load and apply before compilation. LoRA weights are fused into the model weights during compilation. Only used when export=True.

 None
lora_weights_names Optional[Union[str, List[str]]]

Names of specific LoRA weight files to load, corresponding to lora_ids. Only used when export=True.

 None
lora_scales Optional[Union[float, List[float]]]

Scaling factor(s) to apply to the LoRA adapter(s). Only used when export=True.

 None
**kwargs Dict[str, Any]

Additional arguments to pass to the underlying diffusion pipeline constructor or the RBLN compilation process. These may include parameters specific to individual submodules or the particular diffusion pipeline being used.

 {}

Returns:

Type Description
Self

A compiled diffusion pipeline that can be used for inference on RBLN NPU. The returned object is an instance of the class that called this method, inheriting from RBLNDiffusionMixin.

Classes

RBLNStableDiffusionPipelineBaseConfig

Bases: RBLNModelConfig

Functions

__init__(text_encoder=None, unet=None, vae=None, *, batch_size=None, img_height=None, img_width=None, sample_size=None, image_size=None, guidance_scale=None, **kwargs)

Parameters:

Name Type Description Default
text_encoder Optional[RBLNCLIPTextModelConfig]

Configuration for the text encoder component. Initialized as RBLNCLIPTextModelConfig if not provided.

 None
unet Optional[RBLNUNet2DConditionModelConfig]

Configuration for the UNet model component. Initialized as RBLNUNet2DConditionModelConfig if not provided.

 None
vae Optional[RBLNAutoencoderKLConfig]

Configuration for the VAE model component. Initialized as RBLNAutoencoderKLConfig if not provided.

 None
batch_size Optional[int]

Batch size for inference, applied to all submodules.

 None
img_height Optional[int]

Height of the generated images.

 None
img_width Optional[int]

Width of the generated images.

 None
sample_size Optional[Tuple[int, int]]

Spatial dimensions for the UNet model.

 None
image_size Optional[Tuple[int, int]]

Alternative way to specify image dimensions. Cannot be used together with img_height/img_width.

 None
guidance_scale Optional[float]

Scale for classifier-free guidance. Deprecated parameter.

 None
**kwargs Dict[str, Any]

Additional arguments passed to the parent RBLNModelConfig.

 {}

Raises:

Type Description
ValueError

If both image_size and img_height/img_width are provided.
Note

When guidance_scale > 1.0, the UNet batch size is automatically doubled to accommodate classifier-free guidance.

RBLNStableDiffusionPipelineConfig

Bases: RBLNStableDiffusionPipelineBaseConfig

Functions

__init__(text_encoder=None, unet=None, vae=None, *, batch_size=None, img_height=None, img_width=None, sample_size=None, image_size=None, guidance_scale=None, **kwargs)

Parameters:

Name Type Description Default
text_encoder Optional[RBLNCLIPTextModelConfig]

Configuration for the text encoder component. Initialized as RBLNCLIPTextModelConfig if not provided.

 None
unet Optional[RBLNUNet2DConditionModelConfig]

Configuration for the UNet model component. Initialized as RBLNUNet2DConditionModelConfig if not provided.

 None
vae Optional[RBLNAutoencoderKLConfig]

Configuration for the VAE model component. Initialized as RBLNAutoencoderKLConfig if not provided.

 None
batch_size Optional[int]

Batch size for inference, applied to all submodules.

 None
img_height Optional[int]

Height of the generated images.

 None
img_width Optional[int]

Width of the generated images.

 None
sample_size Optional[Tuple[int, int]]

Spatial dimensions for the UNet model.

 None
image_size Optional[Tuple[int, int]]

Alternative way to specify image dimensions. Cannot be used together with img_height/img_width.

 None
guidance_scale Optional[float]

Scale for classifier-free guidance. Deprecated parameter.

 None
**kwargs Dict[str, Any]

Additional arguments passed to the parent RBLNModelConfig.

 {}

Raises:

Type Description
ValueError

If both image_size and img_height/img_width are provided.
Note

When guidance_scale > 1.0, the UNet batch size is automatically doubled to accommodate classifier-free guidance.

RBLNStableDiffusionImg2ImgPipelineConfig

Bases: RBLNStableDiffusionPipelineBaseConfig

Functions

__init__(text_encoder=None, unet=None, vae=None, *, batch_size=None, img_height=None, img_width=None, sample_size=None, image_size=None, guidance_scale=None, **kwargs)

Parameters:

Name Type Description Default
text_encoder Optional[RBLNCLIPTextModelConfig]

Configuration for the text encoder component. Initialized as RBLNCLIPTextModelConfig if not provided.

 None
unet Optional[RBLNUNet2DConditionModelConfig]

Configuration for the UNet model component. Initialized as RBLNUNet2DConditionModelConfig if not provided.

 None
vae Optional[RBLNAutoencoderKLConfig]

Configuration for the VAE model component. Initialized as RBLNAutoencoderKLConfig if not provided.

 None
batch_size Optional[int]

Batch size for inference, applied to all submodules.

 None
img_height Optional[int]

Height of the generated images.

 None
img_width Optional[int]

Width of the generated images.

 None
sample_size Optional[Tuple[int, int]]

Spatial dimensions for the UNet model.

 None
image_size Optional[Tuple[int, int]]

Alternative way to specify image dimensions. Cannot be used together with img_height/img_width.

 None
guidance_scale Optional[float]

Scale for classifier-free guidance. Deprecated parameter.

 None
**kwargs Dict[str, Any]

Additional arguments passed to the parent RBLNModelConfig.

 {}

Raises:

Type Description
ValueError

If both image_size and img_height/img_width are provided.
Note

When guidance_scale > 1.0, the UNet batch size is automatically doubled to accommodate classifier-free guidance.

RBLNStableDiffusionInpaintPipelineConfig

Bases: RBLNStableDiffusionPipelineBaseConfig

Functions

__init__(text_encoder=None, unet=None, vae=None, *, batch_size=None, img_height=None, img_width=None, sample_size=None, image_size=None, guidance_scale=None, **kwargs)

Parameters:

Name Type Description Default
text_encoder Optional[RBLNCLIPTextModelConfig]

Configuration for the text encoder component. Initialized as RBLNCLIPTextModelConfig if not provided.

 None
unet Optional[RBLNUNet2DConditionModelConfig]

Configuration for the UNet model component. Initialized as RBLNUNet2DConditionModelConfig if not provided.

 None
vae Optional[RBLNAutoencoderKLConfig]

Configuration for the VAE model component. Initialized as RBLNAutoencoderKLConfig if not provided.

 None
batch_size Optional[int]

Batch size for inference, applied to all submodules.

 None
img_height Optional[int]

Height of the generated images.

 None
img_width Optional[int]

Width of the generated images.

 None
sample_size Optional[Tuple[int, int]]

Spatial dimensions for the UNet model.

 None
image_size Optional[Tuple[int, int]]

Alternative way to specify image dimensions. Cannot be used together with img_height/img_width.

 None
guidance_scale Optional[float]

Scale for classifier-free guidance. Deprecated parameter.

 None
**kwargs Dict[str, Any]

Additional arguments passed to the parent RBLNModelConfig.

 {}

Raises:

Type Description
ValueError

If both image_size and img_height/img_width are provided.
Note

When guidance_scale > 1.0, the UNet batch size is automatically doubled to accommodate classifier-free guidance.