Spaces:
Paused
Paused
Update custom_pipeline.py
Browse files- custom_pipeline.py +10 -0
custom_pipeline.py
CHANGED
|
@@ -64,14 +64,17 @@ EXAMPLE_DOC_STRING = """
|
|
| 64 |
>>> import torch
|
| 65 |
>>> from diffusers import StableDiffusionXLInstructPix2PixPipeline
|
| 66 |
>>> from diffusers.utils import load_image
|
|
|
|
| 67 |
>>> resolution = 768
|
| 68 |
>>> image = load_image(
|
| 69 |
... "https://hf.co/datasets/diffusers/diffusers-images-docs/resolve/main/mountain.png"
|
| 70 |
... ).resize((resolution, resolution))
|
| 71 |
>>> edit_instruction = "Turn sky into a cloudy one"
|
|
|
|
| 72 |
>>> pipe = StableDiffusionXLInstructPix2PixPipeline.from_pretrained(
|
| 73 |
... "diffusers/sdxl-instructpix2pix-768", torch_dtype=torch.float16
|
| 74 |
... ).to("cuda")
|
|
|
|
| 75 |
>>> edited_image = pipe(
|
| 76 |
... prompt=edit_instruction,
|
| 77 |
... image=image,
|
|
@@ -123,13 +126,16 @@ class CosStableDiffusionXLInstructPix2PixPipeline(
|
|
| 123 |
):
|
| 124 |
r"""
|
| 125 |
Pipeline for pixel-level image editing by following text instructions. Based on Stable Diffusion XL.
|
|
|
|
| 126 |
This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods the
|
| 127 |
library implements for all the pipelines (such as downloading or saving, running on a particular device, etc.)
|
|
|
|
| 128 |
The pipeline also inherits the following loading methods:
|
| 129 |
- [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
|
| 130 |
- [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files
|
| 131 |
- [`~loaders.StableDiffusionXLLoraLoaderMixin.load_lora_weights`] for loading LoRA weights
|
| 132 |
- [`~loaders.StableDiffusionXLLoraLoaderMixin.save_lora_weights`] for saving LoRA weights
|
|
|
|
| 133 |
Args:
|
| 134 |
vae ([`AutoencoderKL`]):
|
| 135 |
Variational Auto-Encoder (VAE) Model to encode and decode images to and from latent representations.
|
|
@@ -220,6 +226,7 @@ class CosStableDiffusionXLInstructPix2PixPipeline(
|
|
| 220 |
):
|
| 221 |
r"""
|
| 222 |
Encodes the prompt into text encoder hidden states.
|
|
|
|
| 223 |
Args:
|
| 224 |
prompt (`str` or `List[str]`, *optional*):
|
| 225 |
prompt to be encoded
|
|
@@ -621,6 +628,7 @@ class CosStableDiffusionXLInstructPix2PixPipeline(
|
|
| 621 |
):
|
| 622 |
r"""
|
| 623 |
Function invoked when calling the pipeline for generation.
|
|
|
|
| 624 |
Args:
|
| 625 |
prompt (`str` or `List[str]`, *optional*):
|
| 626 |
The prompt or prompts to guide the image generation. If not defined, one has to pass `prompt_embeds`.
|
|
@@ -731,7 +739,9 @@ class CosStableDiffusionXLInstructPix2PixPipeline(
|
|
| 731 |
Part of SDXL's micro-conditioning as explained in section 2.2 of
|
| 732 |
[https://huggingface.co/papers/2307.01952](https://huggingface.co/papers/2307.01952). Can be used to
|
| 733 |
simulate an aesthetic score of the generated image by influencing the negative text condition.
|
|
|
|
| 734 |
Examples:
|
|
|
|
| 735 |
Returns:
|
| 736 |
[`~pipelines.stable_diffusion_xl.StableDiffusionXLPipelineOutput`] or `tuple`:
|
| 737 |
[`~pipelines.stable_diffusion_xl.StableDiffusionXLPipelineOutput`] if `return_dict` is True, otherwise a
|
|
|
|
| 64 |
>>> import torch
|
| 65 |
>>> from diffusers import StableDiffusionXLInstructPix2PixPipeline
|
| 66 |
>>> from diffusers.utils import load_image
|
| 67 |
+
|
| 68 |
>>> resolution = 768
|
| 69 |
>>> image = load_image(
|
| 70 |
... "https://hf.co/datasets/diffusers/diffusers-images-docs/resolve/main/mountain.png"
|
| 71 |
... ).resize((resolution, resolution))
|
| 72 |
>>> edit_instruction = "Turn sky into a cloudy one"
|
| 73 |
+
|
| 74 |
>>> pipe = StableDiffusionXLInstructPix2PixPipeline.from_pretrained(
|
| 75 |
... "diffusers/sdxl-instructpix2pix-768", torch_dtype=torch.float16
|
| 76 |
... ).to("cuda")
|
| 77 |
+
|
| 78 |
>>> edited_image = pipe(
|
| 79 |
... prompt=edit_instruction,
|
| 80 |
... image=image,
|
|
|
|
| 126 |
):
|
| 127 |
r"""
|
| 128 |
Pipeline for pixel-level image editing by following text instructions. Based on Stable Diffusion XL.
|
| 129 |
+
|
| 130 |
This model inherits from [`DiffusionPipeline`]. Check the superclass documentation for the generic methods the
|
| 131 |
library implements for all the pipelines (such as downloading or saving, running on a particular device, etc.)
|
| 132 |
+
|
| 133 |
The pipeline also inherits the following loading methods:
|
| 134 |
- [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] for loading textual inversion embeddings
|
| 135 |
- [`~loaders.FromSingleFileMixin.from_single_file`] for loading `.ckpt` files
|
| 136 |
- [`~loaders.StableDiffusionXLLoraLoaderMixin.load_lora_weights`] for loading LoRA weights
|
| 137 |
- [`~loaders.StableDiffusionXLLoraLoaderMixin.save_lora_weights`] for saving LoRA weights
|
| 138 |
+
|
| 139 |
Args:
|
| 140 |
vae ([`AutoencoderKL`]):
|
| 141 |
Variational Auto-Encoder (VAE) Model to encode and decode images to and from latent representations.
|
|
|
|
| 226 |
):
|
| 227 |
r"""
|
| 228 |
Encodes the prompt into text encoder hidden states.
|
| 229 |
+
|
| 230 |
Args:
|
| 231 |
prompt (`str` or `List[str]`, *optional*):
|
| 232 |
prompt to be encoded
|
|
|
|
| 628 |
):
|
| 629 |
r"""
|
| 630 |
Function invoked when calling the pipeline for generation.
|
| 631 |
+
|
| 632 |
Args:
|
| 633 |
prompt (`str` or `List[str]`, *optional*):
|
| 634 |
The prompt or prompts to guide the image generation. If not defined, one has to pass `prompt_embeds`.
|
|
|
|
| 739 |
Part of SDXL's micro-conditioning as explained in section 2.2 of
|
| 740 |
[https://huggingface.co/papers/2307.01952](https://huggingface.co/papers/2307.01952). Can be used to
|
| 741 |
simulate an aesthetic score of the generated image by influencing the negative text condition.
|
| 742 |
+
|
| 743 |
Examples:
|
| 744 |
+
|
| 745 |
Returns:
|
| 746 |
[`~pipelines.stable_diffusion_xl.StableDiffusionXLPipelineOutput`] or `tuple`:
|
| 747 |
[`~pipelines.stable_diffusion_xl.StableDiffusionXLPipelineOutput`] if `return_dict` is True, otherwise a
|