[Community Pipelines] Long Prompt Weighting Stable Diffusion Pipelines (#907)
* [Community Pipelines] Long Prompt Weighting * Update README.md * fix * style * fix style * Update examples/community/README.md Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
This commit is contained in:
SkyTNT
GitHub
parent
ad9d7ce476
commit
2a0c823527
|
|
@ -6,12 +6,13 @@
|
|||
Please have a look at the following table to get an overview of all community examples. Click on the **Code Example** to get a copy-and-paste ready code example that you can try out.
|
||||
If a community doesn't work as expected, please open an issue and ping the author on it.
|
||||
|
||||
| Example | Description | Code Example | Colab | Author |
|
||||
|:----------|:----------------------|:-----------------|:-------------|----------:|
|
||||
| CLIP Guided Stable Diffusion | Doing CLIP guidance for text to image generation with Stable Diffusion| [CLIP Guided Stable Diffusion](#clip-guided-stable-diffusion) | [](https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/CLIP_Guided_Stable_diffusion_with_diffusers.ipynb) | [Suraj Patil](https://github.com/patil-suraj/) |
|
||||
| One Step U-Net (Dummy) | Example showcasing of how to use Community Pipelines (see https://github.com/huggingface/diffusers/issues/841) | [One Step U-Net](#one-step-unet) | - | [Patrick von Platen](https://github.com/patrickvonplaten/) |
|
||||
| Stable Diffusion Interpolation | Interpolate the latent space of Stable Diffusion between different prompts/seeds | [Stable Diffusion Interpolation](#stable-diffusion-interpolation) | - | [Nate Raw](https://github.com/nateraw/) |
|
||||
| Stable Diffusion Mega | **One** Stable Diffusion Pipeline with all functionalities of [Text2Image](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py), [Image2Image](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py) and [Inpainting](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py) | [Stable Diffusion Mega](#stable-diffusion-mega) | - | [Patrick von Platen](https://github.com/patrickvonplaten/) |
|
||||
| Example | Description | Code Example | Colab | Author |
|
||||
|:---------------------------------------|:---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|:------------------------------------------------------------------|:-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-----------------------------------------------------------:|
|
||||
| CLIP Guided Stable Diffusion | Doing CLIP guidance for text to image generation with Stable Diffusion | [CLIP Guided Stable Diffusion](#clip-guided-stable-diffusion) | [](https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/CLIP_Guided_Stable_diffusion_with_diffusers.ipynb) | [Suraj Patil](https://github.com/patil-suraj/) |
|
||||
| One Step U-Net (Dummy) | Example showcasing of how to use Community Pipelines (see https://github.com/huggingface/diffusers/issues/841) | [One Step U-Net](#one-step-unet) | - | [Patrick von Platen](https://github.com/patrickvonplaten/) |
|
||||
| Stable Diffusion Interpolation | Interpolate the latent space of Stable Diffusion between different prompts/seeds | [Stable Diffusion Interpolation](#stable-diffusion-interpolation) | - | [Nate Raw](https://github.com/nateraw/) |
|
||||
| Stable Diffusion Mega | **One** Stable Diffusion Pipeline with all functionalities of [Text2Image](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py), [Image2Image](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py) and [Inpainting](https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_inpaint.py) | [Stable Diffusion Mega](#stable-diffusion-mega) | - | [Patrick von Platen](https://github.com/patrickvonplaten/) |
|
||||
| Long Prompt Weighting Stable Diffusion | **One** Stable Diffusion Pipeline without tokens length limit, and support parsing weighting in prompt. | [Long Prompt Weighting Stable Diffusion](#long-prompt-weighting-stable-diffusion) | - | [SkyTNT](https://github.com/SkyTNT) |
|
||||
|
||||
To load a custom pipeline you just need to pass the `custom_pipeline` argument to `DiffusionPipeline`, as one of the files in `diffusers/examples/community`. Feel free to send a PR with your own pipelines, we will merge them quickly.
|
||||
```py
|
||||
|
|
@ -168,3 +169,50 @@ images = pipe.inpaint(prompt=prompt, init_image=init_image, mask_image=mask_imag
|
|||
|
||||
As shown above this one pipeline can run all both "text-to-image", "image-to-image", and "inpainting" in one pipeline.
|
||||
|
||||
### Long Prompt Weighting Stable Diffusion
|
||||
|
||||
The Pipeline lets you input prompt without 77 token length limit. And you can increase words weighting by using "()" or decrease words weighting by using "[]"
|
||||
The Pipeline also lets you use the main use cases of the stable diffusion pipeline in a single class.
|
||||
|
||||
#### pytorch
|
||||
|
||||
```python
|
||||
from diffusers import DiffusionPipeline
|
||||
import torch
|
||||
|
||||
pipe = DiffusionPipeline.from_pretrained(
|
||||
'hakurei/waifu-diffusion',
|
||||
custom_pipeline="lpw_stable_diffusion",
|
||||
revision="fp16",
|
||||
torch_dtype=torch.float16
|
||||
)
|
||||
pipe=pipe.to("cuda")
|
||||
|
||||
prompt = "best_quality (1girl:1.3) bow bride brown_hair closed_mouth frilled_bow frilled_hair_tubes frills (full_body:1.3) fox_ear hair_bow hair_tubes happy hood japanese_clothes kimono long_sleeves red_bow smile solo tabi uchikake white_kimono wide_sleeves cherry_blossoms"
|
||||
neg_prompt = "lowres, bad_anatomy, error_body, error_hair, error_arm, error_hands, bad_hands, error_fingers, bad_fingers, missing_fingers, error_legs, bad_legs, multiple_legs, missing_legs, error_lighting, error_shadow, error_reflection, text, error, extra_digit, fewer_digits, cropped, worst_quality, low_quality, normal_quality, jpeg_artifacts, signature, watermark, username, blurry"
|
||||
|
||||
pipe.text2img(prompt, negative_prompt=neg_prompt, width=512,height=512,max_embeddings_multiples=3).images[0]
|
||||
|
||||
```
|
||||
|
||||
#### onnxruntime
|
||||
|
||||
```python
|
||||
from diffusers import DiffusionPipeline
|
||||
import torch
|
||||
|
||||
pipe = DiffusionPipeline.from_pretrained(
|
||||
'CompVis/stable-diffusion-v1-4',
|
||||
custom_pipeline="lpw_stable_diffusion_onnx",
|
||||
revision="onnx",
|
||||
provider="CUDAExecutionProvider"
|
||||
)
|
||||
|
||||
prompt = "a photo of an astronaut riding a horse on mars, best quality"
|
||||
neg_prompt = "lowres, bad anatomy, error body, error hair, error arm, error hands, bad hands, error fingers, bad fingers, missing fingers, error legs, bad legs, multiple legs, missing legs, error lighting, error shadow, error reflection, text, error, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry"
|
||||
|
||||
pipe.text2img(prompt,negative_prompt=neg_prompt, width=512, height=512, max_embeddings_multiples=3).images[0]
|
||||
|
||||
```
|
||||
|
||||
if you see `Token indices sequence length is longer than the specified maximum sequence length for this model ( *** > 77 ) . Running this sequence through the model will result in indexing errors`. Do not worry, it is normal.
|
||||
|
|
|
|||
File diff suppressed because it is too large
Load Diff
|
|
@ -0,0 +1,961 @@
|
|||
import inspect
|
||||
import re
|
||||
from typing import Callable, List, Optional, Union
|
||||
|
||||
import numpy as np
|
||||
import torch
|
||||
|
||||
import PIL
|
||||
from diffusers.onnx_utils import OnnxRuntimeModel
|
||||
from diffusers.pipeline_utils import DiffusionPipeline
|
||||
from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput
|
||||
from diffusers.schedulers import DDIMScheduler, LMSDiscreteScheduler, PNDMScheduler
|
||||
from diffusers.utils import logging
|
||||
from transformers import CLIPFeatureExtractor, CLIPTokenizer
|
||||
|
||||
|
||||
logger = logging.get_logger(__name__) # pylint: disable=invalid-name
|
||||
|
||||
re_attention = re.compile(
|
||||
r"""
|
||||
\\\(|
|
||||
\\\)|
|
||||
\\\[|
|
||||
\\]|
|
||||
\\\\|
|
||||
\\|
|
||||
\(|
|
||||
\[|
|
||||
:([+-]?[.\d]+)\)|
|
||||
\)|
|
||||
]|
|
||||
[^\\()\[\]:]+|
|
||||
:
|
||||
""",
|
||||
re.X,
|
||||
)
|
||||
|
||||
|
||||
def parse_prompt_attention(text):
|
||||
"""
|
||||
Parses a string with attention tokens and returns a list of pairs: text and its assoicated weight.
|
||||
Accepted tokens are:
|
||||
(abc) - increases attention to abc by a multiplier of 1.1
|
||||
(abc:3.12) - increases attention to abc by a multiplier of 3.12
|
||||
[abc] - decreases attention to abc by a multiplier of 1.1
|
||||
\( - literal character '('
|
||||
\[ - literal character '['
|
||||
\) - literal character ')'
|
||||
\] - literal character ']'
|
||||
\\ - literal character '\'
|
||||
anything else - just text
|
||||
>>> parse_prompt_attention('normal text')
|
||||
[['normal text', 1.0]]
|
||||
>>> parse_prompt_attention('an (important) word')
|
||||
[['an ', 1.0], ['important', 1.1], [' word', 1.0]]
|
||||
>>> parse_prompt_attention('(unbalanced')
|
||||
[['unbalanced', 1.1]]
|
||||
>>> parse_prompt_attention('\(literal\]')
|
||||
[['(literal]', 1.0]]
|
||||
>>> parse_prompt_attention('(unnecessary)(parens)')
|
||||
[['unnecessaryparens', 1.1]]
|
||||
>>> parse_prompt_attention('a (((house:1.3)) [on] a (hill:0.5), sun, (((sky))).')
|
||||
[['a ', 1.0],
|
||||
['house', 1.5730000000000004],
|
||||
[' ', 1.1],
|
||||
['on', 1.0],
|
||||
[' a ', 1.1],
|
||||
['hill', 0.55],
|
||||
[', sun, ', 1.1],
|
||||
['sky', 1.4641000000000006],
|
||||
['.', 1.1]]
|
||||
"""
|
||||
|
||||
res = []
|
||||
round_brackets = []
|
||||
square_brackets = []
|
||||
|
||||
round_bracket_multiplier = 1.1
|
||||
square_bracket_multiplier = 1 / 1.1
|
||||
|
||||
def multiply_range(start_position, multiplier):
|
||||
for p in range(start_position, len(res)):
|
||||
res[p][1] *= multiplier
|
||||
|
||||
for m in re_attention.finditer(text):
|
||||
text = m.group(0)
|
||||
weight = m.group(1)
|
||||
|
||||
if text.startswith("\\"):
|
||||
res.append([text[1:], 1.0])
|
||||
elif text == "(":
|
||||
round_brackets.append(len(res))
|
||||
elif text == "[":
|
||||
square_brackets.append(len(res))
|
||||
elif weight is not None and len(round_brackets) > 0:
|
||||
multiply_range(round_brackets.pop(), float(weight))
|
||||
elif text == ")" and len(round_brackets) > 0:
|
||||
multiply_range(round_brackets.pop(), round_bracket_multiplier)
|
||||
elif text == "]" and len(square_brackets) > 0:
|
||||
multiply_range(square_brackets.pop(), square_bracket_multiplier)
|
||||
else:
|
||||
res.append([text, 1.0])
|
||||
|
||||
for pos in round_brackets:
|
||||
multiply_range(pos, round_bracket_multiplier)
|
||||
|
||||
for pos in square_brackets:
|
||||
multiply_range(pos, square_bracket_multiplier)
|
||||
|
||||
if len(res) == 0:
|
||||
res = [["", 1.0]]
|
||||
|
||||
# merge runs of identical weights
|
||||
i = 0
|
||||
while i + 1 < len(res):
|
||||
if res[i][1] == res[i + 1][1]:
|
||||
res[i][0] += res[i + 1][0]
|
||||
res.pop(i + 1)
|
||||
else:
|
||||
i += 1
|
||||
|
||||
return res
|
||||
|
||||
|
||||
def get_prompts_with_weights(pipe, prompt: List[str], max_length: int):
|
||||
r"""
|
||||
Tokenize a list of prompts and return its tokens with weights of each token.
|
||||
|
||||
No padding, starting or ending token is included.
|
||||
"""
|
||||
tokens = []
|
||||
weights = []
|
||||
for text in prompt:
|
||||
texts_and_weights = parse_prompt_attention(text)
|
||||
text_token = []
|
||||
text_weight = []
|
||||
for word, weight in texts_and_weights:
|
||||
# tokenize and discard the starting and the ending token
|
||||
token = pipe.tokenizer(word, return_tensors="np").input_ids[0, 1:-1]
|
||||
text_token += list(token)
|
||||
|
||||
# copy the weight by length of token
|
||||
text_weight += [weight] * len(token)
|
||||
|
||||
# stop if the text is too long (longer than truncation limit)
|
||||
if len(text_token) > max_length:
|
||||
break
|
||||
|
||||
# truncate
|
||||
if len(text_token) > max_length:
|
||||
text_token = text_token[:max_length]
|
||||
text_weight = text_weight[:max_length]
|
||||
|
||||
tokens.append(text_token)
|
||||
weights.append(text_weight)
|
||||
return tokens, weights
|
||||
|
||||
|
||||
def pad_tokens_and_weights(tokens, weights, max_length, bos, eos, no_boseos_middle=True, chunk_length=77):
|
||||
r"""
|
||||
Pad the tokens (with starting and ending tokens) and weights (with 1.0) to max_length.
|
||||
"""
|
||||
max_embeddings_multiples = (max_length - 2) // (chunk_length - 2)
|
||||
weights_length = max_length if no_boseos_middle else max_embeddings_multiples * chunk_length
|
||||
for i in range(len(tokens)):
|
||||
tokens[i] = [bos] + tokens[i] + [eos] * (max_length - 1 - len(tokens[i]))
|
||||
if no_boseos_middle:
|
||||
weights[i] = [1.0] + weights[i] + [1.0] * (max_length - 1 - len(weights[i]))
|
||||
else:
|
||||
w = []
|
||||
if len(weights[i]) == 0:
|
||||
w = [1.0] * weights_length
|
||||
else:
|
||||
for j in range((len(weights[i]) - 1) // chunk_length + 1):
|
||||
w.append(1.0) # weight for starting token in this chunk
|
||||
w += weights[i][j * chunk_length : min(len(weights[i]), (j + 1) * chunk_length)]
|
||||
w.append(1.0) # weight for ending token in this chunk
|
||||
w += [1.0] * (weights_length - len(w))
|
||||
weights[i] = w[:]
|
||||
|
||||
return tokens, weights
|
||||
|
||||
|
||||
def get_unweighted_text_embeddings(
|
||||
pipe, text_input: np.array, chunk_length: int, no_boseos_middle: Optional[bool] = True
|
||||
):
|
||||
"""
|
||||
When the length of tokens is a multiple of the capacity of the text encoder,
|
||||
it should be split into chunks and sent to the text encoder individually.
|
||||
"""
|
||||
max_embeddings_multiples = (text_input.shape[1] - 2) // (chunk_length - 2)
|
||||
if max_embeddings_multiples > 1:
|
||||
text_embeddings = []
|
||||
for i in range(max_embeddings_multiples):
|
||||
# extract the i-th chunk
|
||||
text_input_chunk = text_input[:, i * (chunk_length - 2) : (i + 1) * (chunk_length - 2) + 2].copy()
|
||||
|
||||
# cover the head and the tail by the starting and the ending tokens
|
||||
text_input_chunk[:, 0] = text_input[0, 0]
|
||||
text_input_chunk[:, -1] = text_input[0, -1]
|
||||
|
||||
text_embedding = pipe.text_encoder(input_ids=text_input_chunk)[0]
|
||||
|
||||
if no_boseos_middle:
|
||||
if i == 0:
|
||||
# discard the ending token
|
||||
text_embedding = text_embedding[:, :-1]
|
||||
elif i == max_embeddings_multiples - 1:
|
||||
# discard the starting token
|
||||
text_embedding = text_embedding[:, 1:]
|
||||
else:
|
||||
# discard both starting and ending tokens
|
||||
text_embedding = text_embedding[:, 1:-1]
|
||||
|
||||
text_embeddings.append(text_embedding)
|
||||
text_embeddings = np.concatenate(text_embeddings, axis=1)
|
||||
else:
|
||||
text_embeddings = pipe.text_encoder(input_ids=text_input)[0]
|
||||
return text_embeddings
|
||||
|
||||
|
||||
def get_weighted_text_embeddings(
|
||||
pipe,
|
||||
prompt: Union[str, List[str]],
|
||||
uncond_prompt: Optional[Union[str, List[str]]] = None,
|
||||
max_embeddings_multiples: Optional[int] = 4,
|
||||
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,
|
||||
prompt 'A (very beautiful) masterpiece' highlights the words 'very beautiful',
|
||||
and the embedding tokens corresponding to the words get multipled by a constant, 1.1.
|
||||
|
||||
Also, to regularize of the embedding, the weighted embedding would be scaled to preserve the origional mean.
|
||||
|
||||
Args:
|
||||
pipe (`DiffusionPipeline`):
|
||||
Pipe to provide access to the tokenizer and the text encoder.
|
||||
prompt (`str` or `List[str]`):
|
||||
The prompt or prompts to guide the image generation.
|
||||
uncond_prompt (`str` or `List[str]`):
|
||||
The unconditional prompt or prompts for guide the image generation. If unconditional prompt
|
||||
is provided, the embeddings of prompt and uncond_prompt are concatenated.
|
||||
max_embeddings_multiples (`int`, *optional*, defaults to `1`):
|
||||
The max multiple length of prompt embeddings compared to the max output length of text encoder.
|
||||
no_boseos_middle (`bool`, *optional*, defaults to `False`):
|
||||
If the length of text token is multiples of the capacity of text encoder, whether reserve the starting and
|
||||
ending token in each of the chunk in the middle.
|
||||
skip_parsing (`bool`, *optional*, defaults to `False`):
|
||||
Skip the parsing of brackets.
|
||||
skip_weighting (`bool`, *optional*, defaults to `False`):
|
||||
Skip the weighting. When the parsing is skipped, it is forced True.
|
||||
"""
|
||||
max_length = (pipe.tokenizer.model_max_length - 2) * max_embeddings_multiples + 2
|
||||
if isinstance(prompt, str):
|
||||
prompt = [prompt]
|
||||
|
||||
if not skip_parsing:
|
||||
prompt_tokens, prompt_weights = get_prompts_with_weights(pipe, prompt, max_length - 2)
|
||||
if uncond_prompt is not None:
|
||||
if isinstance(uncond_prompt, str):
|
||||
uncond_prompt = [uncond_prompt]
|
||||
uncond_tokens, uncond_weights = get_prompts_with_weights(pipe, uncond_prompt, max_length - 2)
|
||||
else:
|
||||
prompt_tokens = [
|
||||
token[1:-1]
|
||||
for token in pipe.tokenizer(prompt, max_length=max_length, truncation=True, return_tensors="np").input_ids
|
||||
]
|
||||
prompt_weights = [[1.0] * len(token) for token in prompt_tokens]
|
||||
if uncond_prompt is not None:
|
||||
if isinstance(uncond_prompt, str):
|
||||
uncond_prompt = [uncond_prompt]
|
||||
uncond_tokens = [
|
||||
token[1:-1]
|
||||
for token in pipe.tokenizer(
|
||||
uncond_prompt, max_length=max_length, truncation=True, return_tensors="np"
|
||||
).input_ids
|
||||
]
|
||||
uncond_weights = [[1.0] * len(token) for token in uncond_tokens]
|
||||
|
||||
# round up the longest length of tokens to a multiple of (model_max_length - 2)
|
||||
max_length = max([len(token) for token in prompt_tokens])
|
||||
if uncond_prompt is not None:
|
||||
max_length = max(max_length, max([len(token) for token in uncond_tokens]))
|
||||
|
||||
max_embeddings_multiples = min(
|
||||
max_embeddings_multiples, (max_length - 1) // (pipe.tokenizer.model_max_length - 2) + 1
|
||||
)
|
||||
max_embeddings_multiples = max(1, max_embeddings_multiples)
|
||||
max_length = (pipe.tokenizer.model_max_length - 2) * max_embeddings_multiples + 2
|
||||
|
||||
# pad the length of tokens and weights
|
||||
bos = pipe.tokenizer.bos_token_id
|
||||
eos = pipe.tokenizer.eos_token_id
|
||||
prompt_tokens, prompt_weights = pad_tokens_and_weights(
|
||||
prompt_tokens,
|
||||
prompt_weights,
|
||||
max_length,
|
||||
bos,
|
||||
eos,
|
||||
no_boseos_middle=no_boseos_middle,
|
||||
chunk_length=pipe.tokenizer.model_max_length,
|
||||
)
|
||||
prompt_tokens = np.array(prompt_tokens, dtype=np.int32)
|
||||
if uncond_prompt is not None:
|
||||
uncond_tokens, uncond_weights = pad_tokens_and_weights(
|
||||
uncond_tokens,
|
||||
uncond_weights,
|
||||
max_length,
|
||||
bos,
|
||||
eos,
|
||||
no_boseos_middle=no_boseos_middle,
|
||||
chunk_length=pipe.tokenizer.model_max_length,
|
||||
)
|
||||
uncond_tokens = np.array(uncond_tokens, dtype=np.int32)
|
||||
|
||||
# get the embeddings
|
||||
text_embeddings = get_unweighted_text_embeddings(
|
||||
pipe, prompt_tokens, pipe.tokenizer.model_max_length, no_boseos_middle=no_boseos_middle
|
||||
)
|
||||
prompt_weights = np.array(prompt_weights, dtype=text_embeddings.dtype)
|
||||
if uncond_prompt is not None:
|
||||
uncond_embeddings = get_unweighted_text_embeddings(
|
||||
pipe, uncond_tokens, pipe.tokenizer.model_max_length, no_boseos_middle=no_boseos_middle
|
||||
)
|
||||
uncond_weights = np.array(uncond_weights, dtype=uncond_embeddings.dtype)
|
||||
|
||||
# assign weights to the prompts and normalize in the sense of mean
|
||||
# TODO: should we normalize by chunk or in a whole (current implementation)?
|
||||
if (not skip_parsing) and (not skip_weighting):
|
||||
previous_mean = text_embeddings.mean(axis=(-2, -1))
|
||||
text_embeddings *= prompt_weights[:, :, None]
|
||||
text_embeddings *= (previous_mean / text_embeddings.mean(axis=(-2, -1)))[:, None, None]
|
||||
if uncond_prompt is not None:
|
||||
previous_mean = uncond_embeddings.mean(axis=(-2, -1))
|
||||
uncond_embeddings *= uncond_weights[:, :, None]
|
||||
uncond_embeddings *= (previous_mean / uncond_embeddings.mean(axis=(-2, -1)))[:, None, None]
|
||||
|
||||
# For classifier free guidance, we need to do two forward passes.
|
||||
# Here we concatenate the unconditional and text embeddings into a single batch
|
||||
# to avoid doing two forward passes
|
||||
if uncond_prompt is not None:
|
||||
return text_embeddings, uncond_embeddings
|
||||
|
||||
return text_embeddings
|
||||
|
||||
|
||||
def preprocess_image(image):
|
||||
w, h = image.size
|
||||
w, h = map(lambda x: x - x % 32, (w, h)) # resize to integer multiple of 32
|
||||
image = image.resize((w, h), resample=PIL.Image.LANCZOS)
|
||||
image = np.array(image).astype(np.float32) / 255.0
|
||||
image = image[None].transpose(0, 3, 1, 2)
|
||||
return 2.0 * image - 1.0
|
||||
|
||||
|
||||
def preprocess_mask(mask):
|
||||
mask = mask.convert("L")
|
||||
w, h = mask.size
|
||||
w, h = map(lambda x: x - x % 32, (w, h)) # resize to integer multiple of 32
|
||||
mask = mask.resize((w // 8, h // 8), resample=PIL.Image.NEAREST)
|
||||
mask = np.array(mask).astype(np.float32) / 255.0
|
||||
mask = np.tile(mask, (4, 1, 1))
|
||||
mask = mask[None].transpose(0, 1, 2, 3) # what does this step do?
|
||||
mask = 1 - mask # repaint white, keep black
|
||||
return mask
|
||||
|
||||
|
||||
class OnnxStableDiffusionLongPromptWeightingPipeline(DiffusionPipeline):
|
||||
r"""
|
||||
Pipeline for text-to-image generation using Stable Diffusion without tokens length limit, and support parsing
|
||||
weighting in prompt.
|
||||
|
||||
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.)
|
||||
"""
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
vae_encoder: OnnxRuntimeModel,
|
||||
vae_decoder: OnnxRuntimeModel,
|
||||
text_encoder: OnnxRuntimeModel,
|
||||
tokenizer: CLIPTokenizer,
|
||||
unet: OnnxRuntimeModel,
|
||||
scheduler: Union[DDIMScheduler, PNDMScheduler, LMSDiscreteScheduler],
|
||||
safety_checker: OnnxRuntimeModel,
|
||||
feature_extractor: CLIPFeatureExtractor,
|
||||
):
|
||||
super().__init__()
|
||||
self.register_modules(
|
||||
vae_encoder=vae_encoder,
|
||||
vae_decoder=vae_decoder,
|
||||
text_encoder=text_encoder,
|
||||
tokenizer=tokenizer,
|
||||
unet=unet,
|
||||
scheduler=scheduler,
|
||||
safety_checker=safety_checker,
|
||||
feature_extractor=feature_extractor,
|
||||
)
|
||||
|
||||
@torch.no_grad()
|
||||
def __call__(
|
||||
self,
|
||||
prompt: Union[str, List[str]],
|
||||
negative_prompt: Optional[Union[str, List[str]]] = None,
|
||||
init_image: Union[np.ndarray, PIL.Image.Image] = None,
|
||||
mask_image: Union[np.ndarray, PIL.Image.Image] = None,
|
||||
height: int = 512,
|
||||
width: int = 512,
|
||||
num_inference_steps: int = 50,
|
||||
guidance_scale: float = 7.5,
|
||||
strength: float = 0.8,
|
||||
num_images_per_prompt: Optional[int] = 1,
|
||||
eta: float = 0.0,
|
||||
generator: Optional[np.random.RandomState] = None,
|
||||
latents: Optional[np.ndarray] = None,
|
||||
max_embeddings_multiples: Optional[int] = 3,
|
||||
output_type: Optional[str] = "pil",
|
||||
return_dict: bool = True,
|
||||
callback: Optional[Callable[[int, int, np.ndarray], None]] = None,
|
||||
callback_steps: Optional[int] = 1,
|
||||
**kwargs,
|
||||
):
|
||||
r"""
|
||||
Function invoked when calling the pipeline for generation.
|
||||
|
||||
Args:
|
||||
prompt (`str` or `List[str]`):
|
||||
The prompt or prompts to guide the image generation.
|
||||
negative_prompt (`str` or `List[str]`, *optional*):
|
||||
The prompt or prompts not to guide the image generation. Ignored when not using guidance (i.e., ignored
|
||||
if `guidance_scale` is less than `1`).
|
||||
init_image (`np.ndarray` or `PIL.Image.Image`):
|
||||
`Image`, or tensor representing an image batch, that will be used as the starting point for the
|
||||
process.
|
||||
mask_image (`np.ndarray` or `PIL.Image.Image`):
|
||||
`Image`, or tensor representing an image batch, to mask `init_image`. White pixels in the mask will be
|
||||
replaced by noise and therefore repainted, while black pixels will be preserved. If `mask_image` is a
|
||||
PIL image, it will be converted to a single channel (luminance) before use. If it's a tensor, it should
|
||||
contain one color channel (L) instead of 3, so the expected shape would be `(B, H, W, 1)`.
|
||||
height (`int`, *optional*, defaults to 512):
|
||||
The height in pixels of the generated image.
|
||||
width (`int`, *optional*, defaults to 512):
|
||||
The width in pixels of the generated image.
|
||||
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 7.5):
|
||||
Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
|
||||
`guidance_scale` is defined as `w` of equation 2. of [Imagen
|
||||
Paper](https://arxiv.org/pdf/2205.11487.pdf). Guidance scale is enabled by setting `guidance_scale >
|
||||
1`. Higher guidance scale encourages to generate images that are closely linked to the text `prompt`,
|
||||
usually at the expense of lower image quality.
|
||||
strength (`float`, *optional*, defaults to 0.8):
|
||||
Conceptually, indicates how much to transform the reference `init_image`. Must be between 0 and 1.
|
||||
`init_image` will be used as a starting point, adding more noise to it the larger the `strength`. The
|
||||
number of denoising steps depends on the amount of noise initially added. When `strength` is 1, added
|
||||
noise will be maximum and the denoising process will run for the full number of iterations specified in
|
||||
`num_inference_steps`. A value of 1, therefore, essentially ignores `init_image`.
|
||||
num_images_per_prompt (`int`, *optional*, defaults to 1):
|
||||
The number of images to generate per prompt.
|
||||
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 (`np.random.RandomState`, *optional*):
|
||||
A np.random.RandomState to make generation deterministic.
|
||||
latents (`np.ndarray`, *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 ge generated by sampling using the supplied random `generator`.
|
||||
max_embeddings_multiples (`int`, *optional*, defaults to `3`):
|
||||
The max multiple length of prompt embeddings compared to the max output length of text encoder.
|
||||
output_type (`str`, *optional*, defaults to `"pil"`):
|
||||
The output format of the generate image. Choose between
|
||||
[PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`.
|
||||
return_dict (`bool`, *optional*, defaults to `True`):
|
||||
Whether or not to return a [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] instead of a
|
||||
plain tuple.
|
||||
callback (`Callable`, *optional*):
|
||||
A function that will be called every `callback_steps` steps during inference. The function will be
|
||||
called with the following arguments: `callback(step: int, timestep: int, latents: np.ndarray)`.
|
||||
callback_steps (`int`, *optional*, defaults to 1):
|
||||
The frequency at which the `callback` function will be called. If not specified, the callback will be
|
||||
called at every step.
|
||||
|
||||
Returns:
|
||||
[`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] or `tuple`:
|
||||
[`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] 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 `bool`s denoting whether the corresponding generated image likely represents "not-safe-for-work"
|
||||
(nsfw) content, according to the `safety_checker`.
|
||||
"""
|
||||
|
||||
if isinstance(prompt, str):
|
||||
batch_size = 1
|
||||
prompt = [prompt]
|
||||
elif isinstance(prompt, list):
|
||||
batch_size = len(prompt)
|
||||
else:
|
||||
raise ValueError(f"`prompt` has to be of type `str` or `list` but is {type(prompt)}")
|
||||
|
||||
if strength < 0 or strength > 1:
|
||||
raise ValueError(f"The value of strength should in [0.0, 1.0] but is {strength}")
|
||||
|
||||
if height % 8 != 0 or width % 8 != 0:
|
||||
raise ValueError(f"`height` and `width` have to be divisible by 8 but are {height} and {width}.")
|
||||
|
||||
if (callback_steps is None) or (
|
||||
callback_steps is not None and (not isinstance(callback_steps, int) or callback_steps <= 0)
|
||||
):
|
||||
raise ValueError(
|
||||
f"`callback_steps` has to be a positive integer but is {callback_steps} of type"
|
||||
f" {type(callback_steps)}."
|
||||
)
|
||||
|
||||
# get prompt text embeddings
|
||||
|
||||
# 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`
|
||||
# corresponds to doing no classifier free guidance.
|
||||
do_classifier_free_guidance = guidance_scale > 1.0
|
||||
# get unconditional embeddings for classifier free guidance
|
||||
if negative_prompt is None:
|
||||
negative_prompt = [""] * batch_size
|
||||
elif isinstance(negative_prompt, str):
|
||||
negative_prompt = [negative_prompt] * batch_size
|
||||
if batch_size != len(negative_prompt):
|
||||
raise ValueError(
|
||||
f"`negative_prompt`: {negative_prompt} has batch size {len(negative_prompt)}, but `prompt`:"
|
||||
f" {prompt} has batch size {batch_size}. Please make sure that passed `negative_prompt` matches"
|
||||
" the batch size of `prompt`."
|
||||
)
|
||||
|
||||
if generator is None:
|
||||
generator = np.random
|
||||
|
||||
text_embeddings, uncond_embeddings = get_weighted_text_embeddings(
|
||||
pipe=self,
|
||||
prompt=prompt,
|
||||
uncond_prompt=negative_prompt if do_classifier_free_guidance else None,
|
||||
max_embeddings_multiples=max_embeddings_multiples,
|
||||
**kwargs,
|
||||
)
|
||||
|
||||
text_embeddings = text_embeddings.repeat(num_images_per_prompt, 0)
|
||||
if do_classifier_free_guidance:
|
||||
uncond_embeddings = uncond_embeddings.repeat(num_images_per_prompt, 0)
|
||||
text_embeddings = np.concatenate([uncond_embeddings, text_embeddings])
|
||||
|
||||
# set timesteps
|
||||
self.scheduler.set_timesteps(num_inference_steps)
|
||||
|
||||
latents_dtype = text_embeddings.dtype
|
||||
init_latents_orig = None
|
||||
mask = None
|
||||
noise = None
|
||||
|
||||
if init_image is None:
|
||||
latents_shape = (batch_size * num_images_per_prompt, 4, height // 8, width // 8)
|
||||
|
||||
if latents is None:
|
||||
latents = generator.randn(*latents_shape).astype(latents_dtype)
|
||||
elif latents.shape != latents_shape:
|
||||
raise ValueError(f"Unexpected latents shape, got {latents.shape}, expected {latents_shape}")
|
||||
|
||||
timesteps = self.scheduler.timesteps.to(self.device)
|
||||
|
||||
# scale the initial noise by the standard deviation required by the scheduler
|
||||
latents = latents * self.scheduler.init_noise_sigma
|
||||
else:
|
||||
if isinstance(init_image, PIL.Image.Image):
|
||||
init_image = preprocess_image(init_image)
|
||||
# encode the init image into latents and scale the latents
|
||||
init_image = init_image.astype(latents_dtype)
|
||||
init_latents = self.vae_encoder(sample=init_image)[0]
|
||||
init_latents = 0.18215 * init_latents
|
||||
init_latents = np.concatenate([init_latents] * batch_size * num_images_per_prompt)
|
||||
init_latents_orig = init_latents
|
||||
|
||||
# preprocess mask
|
||||
if mask_image is not None:
|
||||
if isinstance(mask_image, PIL.Image.Image):
|
||||
mask_image = preprocess_mask(mask_image)
|
||||
mask_image = mask_image.astype(latents_dtype)
|
||||
mask = np.concatenate([mask_image] * batch_size * num_images_per_prompt)
|
||||
|
||||
# check sizes
|
||||
if not mask.shape == init_latents.shape:
|
||||
print(mask.shape, init_latents.shape)
|
||||
raise ValueError("The mask and init_image should be the same size!")
|
||||
|
||||
# get the original timestep using init_timestep
|
||||
offset = self.scheduler.config.get("steps_offset", 0)
|
||||
init_timestep = int(num_inference_steps * strength) + offset
|
||||
init_timestep = min(init_timestep, num_inference_steps)
|
||||
|
||||
timesteps = self.scheduler.timesteps[-init_timestep]
|
||||
timesteps = torch.tensor([timesteps] * batch_size * num_images_per_prompt)
|
||||
|
||||
# add noise to latents using the timesteps
|
||||
noise = generator.randn(*init_latents.shape).astype(latents_dtype)
|
||||
latents = self.scheduler.add_noise(
|
||||
torch.from_numpy(init_latents), torch.from_numpy(noise), timesteps
|
||||
).numpy()
|
||||
|
||||
t_start = max(num_inference_steps - init_timestep + offset, 0)
|
||||
timesteps = self.scheduler.timesteps[t_start:]
|
||||
|
||||
# prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
|
||||
# eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
|
||||
# eta corresponds to η in DDIM paper: https://arxiv.org/abs/2010.02502
|
||||
# and should be between [0, 1]
|
||||
accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
|
||||
extra_step_kwargs = {}
|
||||
if accepts_eta:
|
||||
extra_step_kwargs["eta"] = eta
|
||||
|
||||
for i, t in enumerate(self.progress_bar(timesteps)):
|
||||
# expand the latents if we are doing classifier free guidance
|
||||
latent_model_input = np.concatenate([latents] * 2) if do_classifier_free_guidance else latents
|
||||
latent_model_input = self.scheduler.scale_model_input(latent_model_input, t)
|
||||
|
||||
# predict the noise residual
|
||||
noise_pred = self.unet(
|
||||
sample=latent_model_input, timestep=np.array([t]), encoder_hidden_states=text_embeddings
|
||||
)
|
||||
noise_pred = noise_pred[0]
|
||||
|
||||
# perform guidance
|
||||
if do_classifier_free_guidance:
|
||||
noise_pred_uncond, noise_pred_text = np.split(noise_pred, 2)
|
||||
noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond)
|
||||
|
||||
# compute the previous noisy sample x_t -> x_t-1
|
||||
latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs).prev_sample.numpy()
|
||||
|
||||
if mask is not None:
|
||||
# masking
|
||||
init_latents_proper = self.scheduler.add_noise(
|
||||
torch.from_numpy(init_latents_orig), torch.from_numpy(noise), torch.tensor([t])
|
||||
).numpy()
|
||||
latents = (init_latents_proper * mask) + (latents * (1 - mask))
|
||||
|
||||
# call the callback, if provided
|
||||
if callback is not None and i % callback_steps == 0:
|
||||
callback(i, t, latents)
|
||||
|
||||
latents = 1 / 0.18215 * latents
|
||||
# image = self.vae_decoder(latent_sample=latents)[0]
|
||||
# it seems likes there is a problem for using half-precision vae decoder if batchsize>1
|
||||
image = []
|
||||
for i in range(latents.shape[0]):
|
||||
image.append(self.vae_decoder(latent_sample=latents[i : i + 1])[0])
|
||||
image = np.concatenate(image)
|
||||
|
||||
image = np.clip(image / 2 + 0.5, 0, 1)
|
||||
image = image.transpose((0, 2, 3, 1))
|
||||
|
||||
if self.safety_checker is not None:
|
||||
safety_checker_input = self.feature_extractor(
|
||||
self.numpy_to_pil(image), return_tensors="np"
|
||||
).pixel_values.astype(image.dtype)
|
||||
# There will throw an error if use safety_checker directly and batchsize>1
|
||||
images, has_nsfw_concept = [], []
|
||||
for i in range(image.shape[0]):
|
||||
image_i, has_nsfw_concept_i = self.safety_checker(
|
||||
clip_input=safety_checker_input[i : i + 1], images=image[i : i + 1]
|
||||
)
|
||||
images.append(image_i)
|
||||
has_nsfw_concept.append(has_nsfw_concept_i)
|
||||
image = np.concatenate(images)
|
||||
else:
|
||||
has_nsfw_concept = None
|
||||
|
||||
if output_type == "pil":
|
||||
image = self.numpy_to_pil(image)
|
||||
|
||||
if not return_dict:
|
||||
return (image, has_nsfw_concept)
|
||||
|
||||
return StableDiffusionPipelineOutput(images=image, nsfw_content_detected=has_nsfw_concept)
|
||||
|
||||
def text2img(
|
||||
self,
|
||||
prompt: Union[str, List[str]],
|
||||
negative_prompt: Optional[Union[str, List[str]]] = None,
|
||||
height: int = 512,
|
||||
width: int = 512,
|
||||
num_inference_steps: int = 50,
|
||||
guidance_scale: float = 7.5,
|
||||
num_images_per_prompt: Optional[int] = 1,
|
||||
eta: float = 0.0,
|
||||
generator: Optional[np.random.RandomState] = None,
|
||||
latents: Optional[np.ndarray] = None,
|
||||
max_embeddings_multiples: Optional[int] = 3,
|
||||
output_type: Optional[str] = "pil",
|
||||
return_dict: bool = True,
|
||||
callback: Optional[Callable[[int, int, np.ndarray], None]] = None,
|
||||
callback_steps: Optional[int] = 1,
|
||||
**kwargs,
|
||||
):
|
||||
r"""
|
||||
Function for text-to-image generation.
|
||||
Args:
|
||||
prompt (`str` or `List[str]`):
|
||||
The prompt or prompts to guide the image generation.
|
||||
negative_prompt (`str` or `List[str]`, *optional*):
|
||||
The prompt or prompts not to guide the image generation. Ignored when not using guidance (i.e., ignored
|
||||
if `guidance_scale` is less than `1`).
|
||||
height (`int`, *optional*, defaults to 512):
|
||||
The height in pixels of the generated image.
|
||||
width (`int`, *optional*, defaults to 512):
|
||||
The width in pixels of the generated image.
|
||||
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 7.5):
|
||||
Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
|
||||
`guidance_scale` is defined as `w` of equation 2. of [Imagen
|
||||
Paper](https://arxiv.org/pdf/2205.11487.pdf). Guidance scale is enabled by setting `guidance_scale >
|
||||
1`. Higher guidance scale encourages to generate images that are closely linked to the text `prompt`,
|
||||
usually at the expense of lower image quality.
|
||||
num_images_per_prompt (`int`, *optional*, defaults to 1):
|
||||
The number of images to generate per prompt.
|
||||
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 (`np.random.RandomState`, *optional*):
|
||||
A np.random.RandomState to make generation deterministic.
|
||||
latents (`np.ndarray`, *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 ge generated by sampling using the supplied random `generator`.
|
||||
max_embeddings_multiples (`int`, *optional*, defaults to `3`):
|
||||
The max multiple length of prompt embeddings compared to the max output length of text encoder.
|
||||
output_type (`str`, *optional*, defaults to `"pil"`):
|
||||
The output format of the generate image. Choose between
|
||||
[PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`.
|
||||
return_dict (`bool`, *optional*, defaults to `True`):
|
||||
Whether or not to return a [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] instead of a
|
||||
plain tuple.
|
||||
callback (`Callable`, *optional*):
|
||||
A function that will be called every `callback_steps` steps during inference. The function will be
|
||||
called with the following arguments: `callback(step: int, timestep: int, latents: np.ndarray)`.
|
||||
callback_steps (`int`, *optional*, defaults to 1):
|
||||
The frequency at which the `callback` function will be called. If not specified, the callback will be
|
||||
called at every step.
|
||||
Returns:
|
||||
[`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] or `tuple`:
|
||||
[`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] 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 `bool`s denoting whether the corresponding generated image likely represents "not-safe-for-work"
|
||||
(nsfw) content, according to the `safety_checker`.
|
||||
"""
|
||||
return self.__call__(
|
||||
prompt=prompt,
|
||||
negative_prompt=negative_prompt,
|
||||
height=height,
|
||||
width=width,
|
||||
num_inference_steps=num_inference_steps,
|
||||
guidance_scale=guidance_scale,
|
||||
num_images_per_prompt=num_images_per_prompt,
|
||||
eta=eta,
|
||||
generator=generator,
|
||||
latents=latents,
|
||||
max_embeddings_multiples=max_embeddings_multiples,
|
||||
output_type=output_type,
|
||||
return_dict=return_dict,
|
||||
callback=callback,
|
||||
callback_steps=callback_steps,
|
||||
**kwargs,
|
||||
)
|
||||
|
||||
def img2img(
|
||||
self,
|
||||
init_image: Union[np.ndarray, PIL.Image.Image],
|
||||
prompt: Union[str, List[str]],
|
||||
negative_prompt: Optional[Union[str, List[str]]] = None,
|
||||
strength: float = 0.8,
|
||||
num_inference_steps: Optional[int] = 50,
|
||||
guidance_scale: Optional[float] = 7.5,
|
||||
num_images_per_prompt: Optional[int] = 1,
|
||||
eta: Optional[float] = 0.0,
|
||||
generator: Optional[np.random.RandomState] = None,
|
||||
max_embeddings_multiples: Optional[int] = 3,
|
||||
output_type: Optional[str] = "pil",
|
||||
return_dict: bool = True,
|
||||
callback: Optional[Callable[[int, int, np.ndarray], None]] = None,
|
||||
callback_steps: Optional[int] = 1,
|
||||
**kwargs,
|
||||
):
|
||||
r"""
|
||||
Function for image-to-image generation.
|
||||
Args:
|
||||
init_image (`np.ndarray` or `PIL.Image.Image`):
|
||||
`Image`, or ndarray representing an image batch, that will be used as the starting point for the
|
||||
process.
|
||||
prompt (`str` or `List[str]`):
|
||||
The prompt or prompts to guide the image generation.
|
||||
negative_prompt (`str` or `List[str]`, *optional*):
|
||||
The prompt or prompts not to guide the image generation. Ignored when not using guidance (i.e., ignored
|
||||
if `guidance_scale` is less than `1`).
|
||||
strength (`float`, *optional*, defaults to 0.8):
|
||||
Conceptually, indicates how much to transform the reference `init_image`. Must be between 0 and 1.
|
||||
`init_image` will be used as a starting point, adding more noise to it the larger the `strength`. The
|
||||
number of denoising steps depends on the amount of noise initially added. When `strength` is 1, added
|
||||
noise will be maximum and the denoising process will run for the full number of iterations specified in
|
||||
`num_inference_steps`. A value of 1, therefore, essentially ignores `init_image`.
|
||||
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. This parameter will be modulated by `strength`.
|
||||
guidance_scale (`float`, *optional*, defaults to 7.5):
|
||||
Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
|
||||
`guidance_scale` is defined as `w` of equation 2. of [Imagen
|
||||
Paper](https://arxiv.org/pdf/2205.11487.pdf). Guidance scale is enabled by setting `guidance_scale >
|
||||
1`. Higher guidance scale encourages to generate images that are closely linked to the text `prompt`,
|
||||
usually at the expense of lower image quality.
|
||||
num_images_per_prompt (`int`, *optional*, defaults to 1):
|
||||
The number of images to generate per prompt.
|
||||
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 (`np.random.RandomState`, *optional*):
|
||||
A np.random.RandomState to make generation deterministic.
|
||||
max_embeddings_multiples (`int`, *optional*, defaults to `3`):
|
||||
The max multiple length of prompt embeddings compared to the max output length of text encoder.
|
||||
output_type (`str`, *optional*, defaults to `"pil"`):
|
||||
The output format of the generate image. Choose between
|
||||
[PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`.
|
||||
return_dict (`bool`, *optional*, defaults to `True`):
|
||||
Whether or not to return a [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] instead of a
|
||||
plain tuple.
|
||||
callback (`Callable`, *optional*):
|
||||
A function that will be called every `callback_steps` steps during inference. The function will be
|
||||
called with the following arguments: `callback(step: int, timestep: int, latents: np.ndarray)`.
|
||||
callback_steps (`int`, *optional*, defaults to 1):
|
||||
The frequency at which the `callback` function will be called. If not specified, the callback will be
|
||||
called at every step.
|
||||
Returns:
|
||||
[`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] or `tuple`:
|
||||
[`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] 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 `bool`s denoting whether the corresponding generated image likely represents "not-safe-for-work"
|
||||
(nsfw) content, according to the `safety_checker`.
|
||||
"""
|
||||
return self.__call__(
|
||||
prompt=prompt,
|
||||
negative_prompt=negative_prompt,
|
||||
init_image=init_image,
|
||||
num_inference_steps=num_inference_steps,
|
||||
guidance_scale=guidance_scale,
|
||||
strength=strength,
|
||||
num_images_per_prompt=num_images_per_prompt,
|
||||
eta=eta,
|
||||
generator=generator,
|
||||
max_embeddings_multiples=max_embeddings_multiples,
|
||||
output_type=output_type,
|
||||
return_dict=return_dict,
|
||||
callback=callback,
|
||||
callback_steps=callback_steps,
|
||||
**kwargs,
|
||||
)
|
||||
|
||||
def inpaint(
|
||||
self,
|
||||
init_image: Union[np.ndarray, PIL.Image.Image],
|
||||
mask_image: Union[np.ndarray, PIL.Image.Image],
|
||||
prompt: Union[str, List[str]],
|
||||
negative_prompt: Optional[Union[str, List[str]]] = None,
|
||||
strength: float = 0.8,
|
||||
num_inference_steps: Optional[int] = 50,
|
||||
guidance_scale: Optional[float] = 7.5,
|
||||
num_images_per_prompt: Optional[int] = 1,
|
||||
eta: Optional[float] = 0.0,
|
||||
generator: Optional[np.random.RandomState] = None,
|
||||
max_embeddings_multiples: Optional[int] = 3,
|
||||
output_type: Optional[str] = "pil",
|
||||
return_dict: bool = True,
|
||||
callback: Optional[Callable[[int, int, np.ndarray], None]] = None,
|
||||
callback_steps: Optional[int] = 1,
|
||||
**kwargs,
|
||||
):
|
||||
r"""
|
||||
Function for inpaint.
|
||||
Args:
|
||||
init_image (`np.ndarray` or `PIL.Image.Image`):
|
||||
`Image`, or tensor representing an image batch, that will be used as the starting point for the
|
||||
process. This is the image whose masked region will be inpainted.
|
||||
mask_image (`np.ndarray` or `PIL.Image.Image`):
|
||||
`Image`, or tensor representing an image batch, to mask `init_image`. White pixels in the mask will be
|
||||
replaced by noise and therefore repainted, while black pixels will be preserved. If `mask_image` is a
|
||||
PIL image, it will be converted to a single channel (luminance) before use. If it's a tensor, it should
|
||||
contain one color channel (L) instead of 3, so the expected shape would be `(B, H, W, 1)`.
|
||||
prompt (`str` or `List[str]`):
|
||||
The prompt or prompts to guide the image generation.
|
||||
negative_prompt (`str` or `List[str]`, *optional*):
|
||||
The prompt or prompts not to guide the image generation. Ignored when not using guidance (i.e., ignored
|
||||
if `guidance_scale` is less than `1`).
|
||||
strength (`float`, *optional*, defaults to 0.8):
|
||||
Conceptually, indicates how much to inpaint the masked area. Must be between 0 and 1. When `strength`
|
||||
is 1, the denoising process will be run on the masked area for the full number of iterations specified
|
||||
in `num_inference_steps`. `init_image` will be used as a reference for the masked area, adding more
|
||||
noise to that region the larger the `strength`. If `strength` is 0, no inpainting will occur.
|
||||
num_inference_steps (`int`, *optional*, defaults to 50):
|
||||
The reference number of denoising steps. More denoising steps usually lead to a higher quality image at
|
||||
the expense of slower inference. This parameter will be modulated by `strength`, as explained above.
|
||||
guidance_scale (`float`, *optional*, defaults to 7.5):
|
||||
Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
|
||||
`guidance_scale` is defined as `w` of equation 2. of [Imagen
|
||||
Paper](https://arxiv.org/pdf/2205.11487.pdf). Guidance scale is enabled by setting `guidance_scale >
|
||||
1`. Higher guidance scale encourages to generate images that are closely linked to the text `prompt`,
|
||||
usually at the expense of lower image quality.
|
||||
num_images_per_prompt (`int`, *optional*, defaults to 1):
|
||||
The number of images to generate per prompt.
|
||||
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 (`np.random.RandomState`, *optional*):
|
||||
A np.random.RandomState to make generation deterministic.
|
||||
max_embeddings_multiples (`int`, *optional*, defaults to `3`):
|
||||
The max multiple length of prompt embeddings compared to the max output length of text encoder.
|
||||
output_type (`str`, *optional*, defaults to `"pil"`):
|
||||
The output format of the generate image. Choose between
|
||||
[PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`.
|
||||
return_dict (`bool`, *optional*, defaults to `True`):
|
||||
Whether or not to return a [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] instead of a
|
||||
plain tuple.
|
||||
callback (`Callable`, *optional*):
|
||||
A function that will be called every `callback_steps` steps during inference. The function will be
|
||||
called with the following arguments: `callback(step: int, timestep: int, latents: np.ndarray)`.
|
||||
callback_steps (`int`, *optional*, defaults to 1):
|
||||
The frequency at which the `callback` function will be called. If not specified, the callback will be
|
||||
called at every step.
|
||||
Returns:
|
||||
[`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] or `tuple`:
|
||||
[`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] 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 `bool`s denoting whether the corresponding generated image likely represents "not-safe-for-work"
|
||||
(nsfw) content, according to the `safety_checker`.
|
||||
"""
|
||||
return self.__call__(
|
||||
prompt=prompt,
|
||||
negative_prompt=negative_prompt,
|
||||
init_image=init_image,
|
||||
mask_image=mask_image,
|
||||
num_inference_steps=num_inference_steps,
|
||||
guidance_scale=guidance_scale,
|
||||
strength=strength,
|
||||
num_images_per_prompt=num_images_per_prompt,
|
||||
eta=eta,
|
||||
generator=generator,
|
||||
max_embeddings_multiples=max_embeddings_multiples,
|
||||
output_type=output_type,
|
||||
return_dict=return_dict,
|
||||
callback=callback,
|
||||
callback_steps=callback_steps,
|
||||
**kwargs,
|
||||
)
|
||||
Loading…
Reference in New Issue