DesignEdit / utils /sdxl.py
jiayueru
update code
37ee4a4
raw
history blame
50.7 kB
from typing import Any, Callable, Dict, List, Optional, Tuple, Union
# import seaborn as sns
import matplotlib.pyplot as plt
import torch
from diffusers import StableDiffusionXLPipeline
from typing import Optional, Union, Tuple, List, Callable, Dict
import numpy as np
import copy
import torch.nn.functional as F
from diffusers.loaders import LoraLoaderMixin, TextualInversionLoaderMixin
from diffusers.models.attention_processor import ( AttnProcessor2_0, LoRAAttnProcessor2_0, LoRAXFormersAttnProcessor, XFormersAttnProcessor, )
from diffusers.utils import ( logging, randn_tensor, replace_example_docstring, )
from diffusers.pipelines.stable_diffusion_xl import StableDiffusionXLPipelineOutput
from diffusers.pipelines.stable_diffusion_xl.pipeline_stable_diffusion_xl import rescale_noise_cfg
import os
logger = logging.get_logger(__name__)
EXAMPLE_DOC_STRING = """
Examples:
```py
>>> import torch
>>> from diffusers import StableDiffusionXLPipeline
>>> pipe = StableDiffusionXLPipeline.from_pretrained(
... "stabilityai/stable-diffusion-xl-base-0.9", torch_dtype=torch.float16
... )
>>> pipe = pipe.to("cuda")
>>> prompt = "a photo of an astronaut riding a horse on mars"
>>> image = pipe(prompt).images[0]
```
"""
class sdxl(StableDiffusionXLPipeline):
@replace_example_docstring(EXAMPLE_DOC_STRING)
@torch.no_grad()
def __call__(
self,
controller=None,
prompt: Union[str, List[str]] = None,
height: Optional[int] = None,
width: Optional[int] = None,
num_inference_steps: int = 50,
guidance_scale: float = 7.5,
negative_prompt: Optional[Union[str, List[str]]] = None,
num_images_per_prompt: Optional[int] = 1,
eta: float = 0.0,
generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
latents: Optional[torch.FloatTensor] = None,
prompt_embeds: Optional[torch.FloatTensor] = None,
negative_prompt_embeds: Optional[torch.FloatTensor] = None,
pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
negative_pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
output_type: Optional[str] = "pil",
return_dict: bool = True,
callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
callback_steps: int = 1,
cross_attention_kwargs: Optional[Dict[str, Any]] = None,
guidance_rescale: float = 0.0,
original_size: Optional[Tuple[int, int]] = None,
crops_coords_top_left: Tuple[int, int] = (0, 0),
target_size: Optional[Tuple[int, int]] = None,
same_init=False,
x_stars=None,
prox_guidance=True,
masa_control=False,
masa_mask=False,
masa_start_step=40,
masa_start_layer=55,
mask_file=None,
query_mask_time=[0, 10],
**kwargs
):
r"""
Function invoked when calling the pipeline for generation.
Args:
prompt (`str` or `List[str]`, *optional*):
The prompt or prompts to guide the image generation. If not defined, one has to pass `prompt_embeds`.
instead.
height (`int`, *optional*, defaults to self.unet.config.sample_size * self.vae_scale_factor):
The height in pixels of the generated image.
width (`int`, *optional*, defaults to self.unet.config.sample_size * self.vae_scale_factor):
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.
negative_prompt (`str` or `List[str]`, *optional*):
The prompt or prompts not to guide the image generation. If not defined, one has to pass
`negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
less than `1`).
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 (`torch.Generator` or `List[torch.Generator]`, *optional*):
One or a list of [torch generator(s)](https://pytorch.org/docs/stable/generated/torch.Generator.html)
to make generation deterministic.
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 ge generated by sampling using the supplied random `generator`.
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 from `prompt` input argument.
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 from `negative_prompt` input
argument.
pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
If not provided, pooled text embeddings will be generated from `prompt` input argument.
negative_pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
Pre-generated negative pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
weighting. If not provided, pooled negative_prompt_embeds will be generated from `negative_prompt`
input argument.
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.StableDiffusionXLPipelineOutput`] 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: torch.FloatTensor)`.
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.
cross_attention_kwargs (`dict`, *optional*):
A kwargs dictionary that if specified is passed along to the `AttentionProcessor` as defined under
`self.processor` in
[diffusers.cross_attention](https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/cross_attention.py).
guidance_rescale (`float`, *optional*, defaults to 0.7):
Guidance rescale factor proposed by [Common Diffusion Noise Schedules and Sample Steps are
Flawed](https://arxiv.org/pdf/2305.08891.pdf) `guidance_scale` is defined as `φ` in equation 16. of
[Common Diffusion Noise Schedules and Sample Steps are Flawed](https://arxiv.org/pdf/2305.08891.pdf).
Guidance rescale factor should fix overexposure when using zero terminal SNR.
Examples:
Returns:
[`~pipelines.stable_diffusion.StableDiffusionXLPipelineOutput`] or `tuple`:
[`~pipelines.stable_diffusion.StableDiffusionXLPipelineOutput`] 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`.
"""
# 0. Default height and width to unet
height = height or self.default_sample_size * self.vae_scale_factor
width = width or self.default_sample_size * self.vae_scale_factor
original_size = original_size or (height, width)
target_size = target_size or (height, width)
inv_batch_size = len(latents) if latents is not None else 1
# 1. Check inputs. Raise error if not correct
self.check_inputs(
prompt,
height,
width,
callback_steps,
negative_prompt,
prompt_embeds,
negative_prompt_embeds,
pooled_prompt_embeds,
negative_pooled_prompt_embeds,
)
# 2. Define call parameters
if prompt is not None and isinstance(prompt, str):
batch_size = 1
elif prompt is not None and isinstance(prompt, list):
batch_size = len(prompt)
else:
batch_size = prompt_embeds.shape[0]
device = self._execution_device
# 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
# 3. Encode input prompt
text_encoder_lora_scale = (
cross_attention_kwargs.get("scale", None) if cross_attention_kwargs is not None else None
)
(
prompt_embeds,
negative_prompt_embeds,
pooled_prompt_embeds,
negative_pooled_prompt_embeds,
) = self.encode_prompt(
prompt,
device,
num_images_per_prompt,
do_classifier_free_guidance,
negative_prompt,
prompt_embeds=prompt_embeds,
negative_prompt_embeds=negative_prompt_embeds,
pooled_prompt_embeds=pooled_prompt_embeds,
negative_pooled_prompt_embeds=negative_pooled_prompt_embeds,
lora_scale=text_encoder_lora_scale,
sample_ref_match=kwargs['sample_ref_match'] if 'sample_ref_match' in kwargs else None,
)
# 4. Prepare timesteps
self.scheduler.set_timesteps(num_inference_steps, device=device)
timesteps = self.scheduler.timesteps
# 5. Prepare latent variables
num_channels_latents = self.unet.config.in_channels
latents = self.prepare_latents(
batch_size * num_images_per_prompt,
num_channels_latents,
height,
width,
prompt_embeds.dtype,
device,
generator,
latents,
same_init=same_init, #ADD
sample_ref_match=kwargs['sample_ref_match'] if 'sample_ref_match' in kwargs else None,
)
# 6. Prepare extra step kwargs.
extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)
# 7. Prepare added time ids & embeddings
add_text_embeds = pooled_prompt_embeds
add_time_ids = self._get_add_time_ids(
original_size, crops_coords_top_left, target_size, dtype=prompt_embeds.dtype
)
if do_classifier_free_guidance:
prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds], dim=0)
add_text_embeds = torch.cat([negative_pooled_prompt_embeds, add_text_embeds], dim=0)
add_time_ids = torch.cat([add_time_ids, add_time_ids], dim=0)
prompt_embeds = prompt_embeds.to(device)
add_text_embeds = add_text_embeds.to(device)
add_time_ids = add_time_ids.to(device).repeat(batch_size * num_images_per_prompt, 1)
# 8. Denoising loop
num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order
with self.progress_bar(total=num_inference_steps) as progress_bar:
for i, t in enumerate(timesteps):
# expand the latents if we are doing classifier free guidance
latent_model_input = torch.cat([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
added_cond_kwargs = {"text_embeds": add_text_embeds, "time_ids": add_time_ids}
noise_pred = self.unet(
latent_model_input,
t,
encoder_hidden_states=prompt_embeds,
cross_attention_kwargs=cross_attention_kwargs,
added_cond_kwargs=added_cond_kwargs,
return_dict=False,
)[0]
# perform guidance
if do_classifier_free_guidance:
noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
# CHANGE START
score_delta,mask_edit=self.prox_regularization(
noise_pred_uncond,
noise_pred_text,
i,
t,
prox_guidance=prox_guidance,
)
if mask_edit is not None:
a = 1
noise_pred = noise_pred_uncond + guidance_scale * score_delta
# CHANGE END
if do_classifier_free_guidance and guidance_rescale > 0.0:
# Based on 3.4. in https://arxiv.org/pdf/2305.08891.pdf
noise_pred = rescale_noise_cfg(noise_pred, noise_pred_text, guidance_rescale=guidance_rescale)
# compute the previous noisy sample x_t -> x_t-1
latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs, return_dict=False)[0]
# ADD START
latents = self.proximal_guidance(
i,
t,
latents,
mask_edit,
prox_guidance=prox_guidance,
dtype=self.unet.dtype,
x_stars=x_stars,
controller=controller,
sample_ref_match=kwargs['sample_ref_match'] if 'sample_ref_match' in kwargs else None,
inv_batch_size=inv_batch_size,
only_inversion_align=kwargs['only_inversion_align'] if 'only_inversion_align' in kwargs else False,
)
# ADD END
if controller is not None:
latents = controller.step_callback(latents)
# call the callback, if provided
if i == len(timesteps) - 1 or ((i + 1) > num_warmup_steps and (i + 1) % self.scheduler.order == 0):
progress_bar.update()
if callback is not None and i % callback_steps == 0:
callback(i, t, latents)
# make sure the VAE is in float32 mode, as it overflows in float16
self.vae.to(dtype=torch.float32)
use_torch_2_0_or_xformers = isinstance(
self.vae.decoder.mid_block.attentions[0].processor,
(
AttnProcessor2_0,
XFormersAttnProcessor,
LoRAXFormersAttnProcessor,
LoRAAttnProcessor2_0,
),
)
# if xformers or torch_2_0 is used attention block does not need
# to be in float32 which can save lots of memory
if use_torch_2_0_or_xformers:
self.vae.post_quant_conv.to(latents.dtype)
self.vae.decoder.conv_in.to(latents.dtype)
self.vae.decoder.mid_block.to(latents.dtype)
else:
latents = latents.float()
if not output_type == "latent":
image = self.vae.decode(latents / self.vae.config.scaling_factor, return_dict=False)[0]
else:
image = latents
return StableDiffusionXLPipelineOutput(images=image)
image = self.watermark.apply_watermark(image)
image = self.image_processor.postprocess(image, output_type=output_type)
# Offload last model to CPU
if hasattr(self, "final_offload_hook") and self.final_offload_hook is not None:
self.final_offload_hook.offload()
return image
# Copied from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline.prepare_latents
def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None,same_init=False,sample_ref_match=None):
shape = (batch_size, num_channels_latents, height // self.vae_scale_factor, width // self.vae_scale_factor)
if isinstance(generator, list) and len(generator) != batch_size:
raise ValueError(
f"You have passed a list of generators of length {len(generator)}, but requested an effective batch"
f" size of {batch_size}. Make sure the batch size matches the length of the generators."
)
if sample_ref_match is not None:
new_latents=randn_tensor((batch_size,*shape[1:]), generator=generator, device=device, dtype=dtype)
for key,value in sample_ref_match.items():
new_latents[key]=latents[value].clone()
latents=new_latents
else:
if same_init is True:
if latents is None:
latents = randn_tensor((1,*shape[1:]), generator=generator, device=device, dtype=dtype).expand(shape).to(device)
else:
if batch_size>1 and latents.shape[0]==1:
latents=latents.expand(shape).to(device)
else:
latents = latents.to(device)
else:
if latents is None:
latents = randn_tensor(shape, generator=generator, device=device, dtype=dtype)
else:
latents = latents.to(device)
# scale the initial noise by the standard deviation required by the scheduler
latents = latents * self.scheduler.init_noise_sigma
return latents
def encode_prompt(
self,
prompt,
device: Optional[torch.device] = None,
num_images_per_prompt: int = 1,
do_classifier_free_guidance: bool = True,
negative_prompt=None,
prompt_embeds: Optional[torch.FloatTensor] = None,
negative_prompt_embeds: Optional[torch.FloatTensor] = None,
pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
negative_pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
lora_scale: Optional[float] = None,
sample_ref_match=None,
):
r"""
Encodes the prompt into text encoder hidden states.
Args:
prompt (`str` or `List[str]`, *optional*):
prompt to be encoded
device: (`torch.device`):
torch device
num_images_per_prompt (`int`):
number of images that should be generated per prompt
do_classifier_free_guidance (`bool`):
whether to use classifier free guidance or not
negative_prompt (`str` or `List[str]`, *optional*):
The prompt or prompts not to guide the image generation. If not defined, one has to pass
`negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
less than `1`).
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 from `prompt` input argument.
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 from `negative_prompt` input
argument.
pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
If not provided, pooled text embeddings will be generated from `prompt` input argument.
negative_pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
Pre-generated negative pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
weighting. If not provided, pooled negative_prompt_embeds will be generated from `negative_prompt`
input argument.
lora_scale (`float`, *optional*):
A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
"""
device = device or self._execution_device
# set lora scale so that monkey patched LoRA
# function of text encoder can correctly access it
if lora_scale is not None and isinstance(self, LoraLoaderMixin):
self._lora_scale = lora_scale
if prompt is not None and isinstance(prompt, str):
batch_size = 1
elif prompt is not None and isinstance(prompt, list):
batch_size = len(prompt)
else:
batch_size = prompt_embeds.shape[0]
# Define tokenizers and text encoders
tokenizers = [self.tokenizer, self.tokenizer_2] if self.tokenizer is not None else [self.tokenizer_2]
text_encoders = (
[self.text_encoder, self.text_encoder_2] if self.text_encoder is not None else [self.text_encoder_2]
)
if prompt_embeds is None:
# textual inversion: procecss multi-vector tokens if necessary
prompt_embeds_list = []
for tokenizer, text_encoder in zip(tokenizers, text_encoders):
if isinstance(self, TextualInversionLoaderMixin):
prompt = self.maybe_convert_prompt(prompt, tokenizer)
text_inputs = tokenizer(
prompt,
padding="max_length",
max_length=tokenizer.model_max_length,
truncation=True,
return_tensors="pt",
)
text_input_ids = text_inputs.input_ids
untruncated_ids = tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
text_input_ids, untruncated_ids
):
removed_text = tokenizer.batch_decode(untruncated_ids[:, tokenizer.model_max_length - 1 : -1])
logger.warning(
"The following part of your input was truncated because CLIP can only handle sequences up to"
f" {tokenizer.model_max_length} tokens: {removed_text}"
)
prompt_embeds = text_encoder(
text_input_ids.to(device),
output_hidden_states=True,
)
# We are only ALWAYS interested in the pooled output of the final text encoder
pooled_prompt_embeds = prompt_embeds[0]
prompt_embeds = prompt_embeds.hidden_states[-2]
bs_embed, seq_len, _ = prompt_embeds.shape
# duplicate text embeddings for each generation per prompt, using mps friendly method
prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)
prompt_embeds_list.append(prompt_embeds)
prompt_embeds = torch.concat(prompt_embeds_list, dim=-1)
# get unconditional embeddings for classifier free guidance
zero_out_negative_prompt = negative_prompt is None and self.config.force_zeros_for_empty_prompt
if do_classifier_free_guidance and negative_prompt_embeds is None and zero_out_negative_prompt:
negative_prompt_embeds = torch.zeros_like(prompt_embeds)
negative_pooled_prompt_embeds = torch.zeros_like(pooled_prompt_embeds)
elif do_classifier_free_guidance and negative_prompt_embeds is None:
negative_prompt = negative_prompt or ""
uncond_tokens: List[str]
if prompt is not None and type(prompt) is not type(negative_prompt):
raise TypeError(
f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
f" {type(prompt)}."
)
elif isinstance(negative_prompt, str):
uncond_tokens = [negative_prompt]
elif 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`."
)
else:
uncond_tokens = negative_prompt
negative_prompt_embeds_list = []
for tokenizer, text_encoder in zip(tokenizers, text_encoders):
# textual inversion: procecss multi-vector tokens if necessary
if isinstance(self, TextualInversionLoaderMixin):
uncond_tokens = self.maybe_convert_prompt(uncond_tokens, tokenizer)
max_length = prompt_embeds.shape[1]
uncond_input = tokenizer(
uncond_tokens,
padding="max_length",
max_length=max_length,
truncation=True,
return_tensors="pt",
)
negative_prompt_embeds = text_encoder(
uncond_input.input_ids.to(device),
output_hidden_states=True,
)
# We are only ALWAYS interested in the pooled output of the final text encoder
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]
if do_classifier_free_guidance:
# duplicate unconditional embeddings for each generation per prompt, using mps friendly method
seq_len = negative_prompt_embeds.shape[1]
negative_prompt_embeds = negative_prompt_embeds.to(dtype=text_encoder.dtype, device=device)
negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
negative_prompt_embeds = negative_prompt_embeds.view(
batch_size * num_images_per_prompt, seq_len, -1
)
# 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
negative_prompt_embeds_list.append(negative_prompt_embeds)
negative_prompt_embeds = torch.concat(negative_prompt_embeds_list, dim=-1)
bs_embed = pooled_prompt_embeds.shape[0]
# ADD START
if sample_ref_match is not None:
new_negative_prompt_embeds=torch.zeros_like(prompt_embeds)
new_negative_pooled_prompt_embeds=torch.zeros_like(pooled_prompt_embeds)
for key,value in sample_ref_match.items():
new_negative_prompt_embeds[key]=negative_prompt_embeds[value].clone()
new_negative_pooled_prompt_embeds[key]=negative_pooled_prompt_embeds[value].clone()
negative_prompt_embeds=new_negative_prompt_embeds
negative_pooled_prompt_embeds=new_negative_pooled_prompt_embeds
else:
if negative_pooled_prompt_embeds.shape[0]==1 and bs_embed!=1:
negative_pooled_prompt_embeds=negative_pooled_prompt_embeds.repeat(bs_embed,1)
if negative_prompt_embeds.shape[0]==1 and bs_embed!=1:
negative_prompt_embeds=negative_prompt_embeds.repeat(bs_embed,1,1)
# ADD END
pooled_prompt_embeds = pooled_prompt_embeds.repeat(1, num_images_per_prompt).view(
bs_embed * num_images_per_prompt, -1
)
negative_pooled_prompt_embeds = negative_pooled_prompt_embeds.repeat(1, num_images_per_prompt).view(
bs_embed * num_images_per_prompt, -1
)
return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
def encode_prompt_not_zero_uncond(
self,
prompt,
device: Optional[torch.device] = None,
num_images_per_prompt: int = 1,
do_classifier_free_guidance: bool = True,
negative_prompt=None,
prompt_embeds: Optional[torch.FloatTensor] = None,
negative_prompt_embeds: Optional[torch.FloatTensor] = None,
pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
negative_pooled_prompt_embeds: Optional[torch.FloatTensor] = None,
lora_scale: Optional[float] = None,
):
r"""
Encodes the prompt into text encoder hidden states.
Args:
prompt (`str` or `List[str]`, *optional*):
prompt to be encoded
device: (`torch.device`):
torch device
num_images_per_prompt (`int`):
number of images that should be generated per prompt
do_classifier_free_guidance (`bool`):
whether to use classifier free guidance or not
negative_prompt (`str` or `List[str]`, *optional*):
The prompt or prompts not to guide the image generation. If not defined, one has to pass
`negative_prompt_embeds` instead. Ignored when not using guidance (i.e., ignored if `guidance_scale` is
less than `1`).
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 from `prompt` input argument.
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 from `negative_prompt` input
argument.
pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
Pre-generated pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting.
If not provided, pooled text embeddings will be generated from `prompt` input argument.
negative_pooled_prompt_embeds (`torch.FloatTensor`, *optional*):
Pre-generated negative pooled text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
weighting. If not provided, pooled negative_prompt_embeds will be generated from `negative_prompt`
input argument.
lora_scale (`float`, *optional*):
A lora scale that will be applied to all LoRA layers of the text encoder if LoRA layers are loaded.
"""
device = device or self._execution_device
# set lora scale so that monkey patched LoRA
# function of text encoder can correctly access it
if lora_scale is not None and isinstance(self, LoraLoaderMixin):
self._lora_scale = lora_scale
if prompt is not None and isinstance(prompt, str):
batch_size = 1
elif prompt is not None and isinstance(prompt, list):
batch_size = len(prompt)
else:
batch_size = prompt_embeds.shape[0]
# Define tokenizers and text encoders
tokenizers = [self.tokenizer, self.tokenizer_2] if self.tokenizer is not None else [self.tokenizer_2]
text_encoders = (
[self.text_encoder, self.text_encoder_2] if self.text_encoder is not None else [self.text_encoder_2]
)
if prompt_embeds is None:
# textual inversion: procecss multi-vector tokens if necessary
prompt_embeds_list = []
for tokenizer, text_encoder in zip(tokenizers, text_encoders):
if isinstance(self, TextualInversionLoaderMixin):
prompt = self.maybe_convert_prompt(prompt, tokenizer)
text_inputs = tokenizer(
prompt,
padding="max_length",
max_length=tokenizer.model_max_length,
truncation=True,
return_tensors="pt",
)
text_input_ids = text_inputs.input_ids
untruncated_ids = tokenizer(prompt, padding="longest", return_tensors="pt").input_ids
if untruncated_ids.shape[-1] >= text_input_ids.shape[-1] and not torch.equal(
text_input_ids, untruncated_ids
):
removed_text = tokenizer.batch_decode(untruncated_ids[:, tokenizer.model_max_length - 1 : -1])
logger.warning(
"The following part of your input was truncated because CLIP can only handle sequences up to"
f" {tokenizer.model_max_length} tokens: {removed_text}"
)
prompt_embeds = text_encoder(text_input_ids.to(device),output_hidden_states=True)
# We are only ALWAYS interested in the pooled output of the final text encoder
pooled_prompt_embeds = prompt_embeds[0]
prompt_embeds = prompt_embeds.hidden_states[-2]
bs_embed, seq_len, _ = prompt_embeds.shape
# duplicate text embeddings for each generation per prompt, using mps friendly method
prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
prompt_embeds = prompt_embeds.view(bs_embed * num_images_per_prompt, seq_len, -1)
prompt_embeds_list.append(prompt_embeds)
prompt_embeds = torch.concat(prompt_embeds_list, dim=-1)
# get unconditional embeddings for classifier free guidance
if do_classifier_free_guidance and negative_prompt_embeds is None:
negative_prompt = negative_prompt or ""
uncond_tokens: List[str]
if prompt is not None and isinstance(prompt,List) and negative_prompt == "":
negative_prompt = ["" for i in range(len(prompt))]
if prompt is not None and type(prompt) is not type(negative_prompt):
raise TypeError(
f"`negative_prompt` should be the same type to `prompt`, but got {type(negative_prompt)} !="
f" {type(prompt)}."
)
elif isinstance(negative_prompt, str):
uncond_tokens = [negative_prompt]
elif 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`."
)
else:
uncond_tokens = negative_prompt
negative_prompt_embeds_list = []
for tokenizer, text_encoder in zip(tokenizers, text_encoders):
# textual inversion: procecss multi-vector tokens if necessary
if isinstance(self, TextualInversionLoaderMixin):
uncond_tokens = self.maybe_convert_prompt(uncond_tokens, tokenizer)
max_length = prompt_embeds.shape[1]
uncond_input = tokenizer(
uncond_tokens,
padding="max_length",
max_length=max_length,
truncation=True,
return_tensors="pt",
)
negative_prompt_embeds = text_encoder(
uncond_input.input_ids.to(device),
output_hidden_states=True,
)
# We are only ALWAYS interested in the pooled output of the final text encoder
negative_pooled_prompt_embeds = negative_prompt_embeds[0]
negative_prompt_embeds = negative_prompt_embeds.hidden_states[-2]
if do_classifier_free_guidance:
# duplicate unconditional embeddings for each generation per prompt, using mps friendly method
seq_len = negative_prompt_embeds.shape[1]
negative_prompt_embeds = negative_prompt_embeds.to(dtype=text_encoder.dtype, device=device)
negative_prompt_embeds = negative_prompt_embeds.repeat(1, num_images_per_prompt, 1)
negative_prompt_embeds = negative_prompt_embeds.view(
batch_size * num_images_per_prompt, seq_len, -1
)
# 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
negative_prompt_embeds_list.append(negative_prompt_embeds)
negative_prompt_embeds = torch.concat(negative_prompt_embeds_list, dim=-1)
bs_embed = pooled_prompt_embeds.shape[0]
pooled_prompt_embeds = pooled_prompt_embeds.repeat(1, num_images_per_prompt).view(
bs_embed * num_images_per_prompt, -1
)
negative_pooled_prompt_embeds = negative_pooled_prompt_embeds.repeat(1, num_images_per_prompt).view(
bs_embed * num_images_per_prompt, -1
)
return prompt_embeds, negative_prompt_embeds, pooled_prompt_embeds, negative_pooled_prompt_embeds
def prox_regularization(
self,
noise_pred_uncond,
noise_pred_text,
i,
t,
prox_guidance=False,
prox=None,
quantile=0.75,
recon_t=400,
dilate_radius=2,
):
if prox_guidance is True:
mask_edit = None
if prox == 'l1':
score_delta = (noise_pred_text - noise_pred_uncond).float()
if quantile > 0:
threshold = score_delta.abs().quantile(quantile)
else:
threshold = -quantile # if quantile is negative, use it as a fixed threshold
score_delta -= score_delta.clamp(-threshold, threshold)
score_delta = torch.where(score_delta > 0, score_delta-threshold, score_delta)
score_delta = torch.where(score_delta < 0, score_delta+threshold, score_delta)
if (recon_t > 0 and t < recon_t) or (recon_t < 0 and t > -recon_t):
mask_edit = (score_delta.abs() > threshold).float()
if dilate_radius > 0:
radius = int(dilate_radius)
mask_edit = dilate(mask_edit.float(), kernel_size=2*radius+1, padding=radius)
elif prox == 'l0':
score_delta = (noise_pred_text - noise_pred_uncond).float()
if quantile > 0:
threshold = score_delta.abs().quantile(quantile)
else:
threshold = -quantile # if quantile is negative, use it as a fixed threshold
score_delta -= score_delta.clamp(-threshold, threshold)
if (recon_t > 0 and t < recon_t) or (recon_t < 0 and t > -recon_t):
mask_edit = (score_delta.abs() > threshold).float()
if dilate_radius > 0:
radius = int(dilate_radius)
mask_edit = dilate(mask_edit.float(), kernel_size=2*radius+1, padding=radius)
elif prox==None:
score_delta = (noise_pred_text - noise_pred_uncond).float()
if quantile > 0:
threshold = score_delta.abs().quantile(quantile)
else:
threshold = -quantile # if quantile is negative, use it as a fixed threshold
if (recon_t > 0 and t < recon_t) or (recon_t < 0 and t > -recon_t):
mask_edit = (score_delta.abs() > threshold).float()
if dilate_radius > 0:
radius = int(dilate_radius)
mask_edit = dilate(mask_edit.float(), kernel_size=2*radius+1, padding=radius)
else:
raise NotImplementedError
return score_delta,mask_edit
else:
return noise_pred_text - noise_pred_uncond,None
def proximal_guidance(
self,
i,
t,
latents,
mask_edit,
dtype,
prox_guidance=False,
recon_t=400,
recon_end=0,
recon_lr=0.1,
x_stars=None,
controller=None,
sample_ref_match=None,
inv_batch_size=1,
only_inversion_align=False,
):
if mask_edit is not None and prox_guidance and (recon_t > recon_end and t < recon_t) or (recon_t < -recon_end and t > -recon_t):
if controller.layer_fusion.remove_mask is not None:
fix_mask = copy.deepcopy(controller.layer_fusion.remove_mask)
mask_edit[1] = (mask_edit[1]+fix_mask).clamp(0,1)
if mask_edit.shape[0] > 2:
mask_edit[2].fill_(1)
recon_mask = 1 - mask_edit
target_latents=x_stars[len(x_stars)-i-2]
new_target_latents=torch.zeros_like(latents)
for key,value in sample_ref_match.items():
new_target_latents[key]=target_latents[value].clone()
latents = latents - recon_lr * (latents - new_target_latents) * recon_mask
return latents.to(dtype)
def slerp(val, low, high):
""" taken from https://discuss.pytorch.org/t/help-regarding-slerp-function-for-generative-model-sampling/32475/4
"""
low_norm = low/torch.norm(low, dim=1, keepdim=True)
high_norm = high/torch.norm(high, dim=1, keepdim=True)
omega = torch.acos((low_norm*high_norm).sum(1))
so = torch.sin(omega)
res = (torch.sin((1.0-val)*omega)/so).unsqueeze(1)*low + (torch.sin(val*omega)/so).unsqueeze(1) * high
return res
def slerp_tensor(val, low, high):
shape = low.shape
res = slerp(val, low.flatten(1), high.flatten(1))
return res.reshape(shape)
def dilate(image, kernel_size, stride=1, padding=0):
"""
Perform dilation on a binary image using a square kernel.
"""
# Ensure the image is binary
assert image.max() <= 1 and image.min() >= 0
# Get the maximum value in each neighborhood
dilated_image = F.max_pool2d(image, kernel_size, stride, padding)
return dilated_image
def exec_classifier_free_guidance(model,latents,controller,t,guidance_scale,
do_classifier_free_guidance,noise_pred,guidance_rescale,
prox=None, quantile=0.75,image_enc=None, recon_lr=0.1, recon_t=400,recon_end_t=0,
inversion_guidance=False, reconstruction_guidance=False,x_stars=None, i=0,
use_localblend_mask=False,
save_heatmap=False,**kwargs):
# perform guidance
if do_classifier_free_guidance:
noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
#noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond)
if prox is None and inversion_guidance is True:
prox = 'l1'
step_kwargs = {
'ref_image': None,
'recon_lr': 0,
'recon_mask': None,
}
mask_edit = None
if prox is not None:
if prox == 'l1':
score_delta = (noise_pred_text - noise_pred_uncond).float()
if quantile > 0:
threshold = score_delta.abs().quantile(quantile)
else:
threshold = -quantile # if quantile is negative, use it as a fixed threshold
score_delta -= score_delta.clamp(-threshold, threshold)
score_delta = torch.where(score_delta > 0, score_delta-threshold, score_delta)
score_delta = torch.where(score_delta < 0, score_delta+threshold, score_delta)
if (recon_t > 0 and t < recon_t) or (recon_t < 0 and t > -recon_t):
step_kwargs['ref_image'] = image_enc
step_kwargs['recon_lr'] = recon_lr
score_delta_norm=score_delta.abs()
score_delta_norm=(score_delta_norm - score_delta_norm.min ()) / (score_delta_norm.max () - score_delta_norm.min ())
mask_edit = (score_delta.abs() > threshold).float()
if save_heatmap and i%10==0:
for kk in range(4):
sns.heatmap(mask_edit[1][kk].clone().cpu(), cmap='coolwarm')
plt.savefig(f'./vis/prox_inv/heatmap1_mask_{i}_{kk}.png')
plt.clf()
if kwargs.get('dilate_mask', 2) > 0:
radius = int(kwargs.get('dilate_mask', 2))
mask_edit = dilate(mask_edit.float(), kernel_size=2*radius+1, padding=radius)
if save_heatmap and i%10==0:
for kk in range(4):
sns.heatmap(mask_edit[1][kk].clone().cpu(), cmap='coolwarm')
plt.savefig(f'./vis/prox_inv/heatmap1_mask_dilate_{i}_{kk}.png')
plt.clf()
step_kwargs['recon_mask'] = 1 - mask_edit
elif prox == 'l0':
score_delta = (noise_pred_text - noise_pred_uncond).float()
if quantile > 0:
threshold = score_delta.abs().quantile(quantile)
else:
threshold = -quantile # if quantile is negative, use it as a fixed threshold
score_delta -= score_delta.clamp(-threshold, threshold)
if (recon_t > 0 and t < recon_t) or (recon_t < 0 and t > -recon_t):
step_kwargs['ref_image'] = image_enc
step_kwargs['recon_lr'] = recon_lr
mask_edit = (score_delta.abs() > threshold).float()
if kwargs.get('dilate_mask', 2) > 0:
radius = int(kwargs.get('dilate_mask', 2))
mask_edit = dilate(mask_edit.float(), kernel_size=2*radius+1, padding=radius)
step_kwargs['recon_mask'] = 1 - mask_edit
else:
raise NotImplementedError
noise_pred = (noise_pred_uncond + guidance_scale * score_delta).to(model.unet.dtype)
else:
noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond)
if do_classifier_free_guidance and guidance_rescale > 0.0:
# Based on 3.4. in https://arxiv.org/pdf/2305.08891.pdf
noise_pred = rescale_noise_cfg(noise_pred, noise_pred_text, guidance_rescale=guidance_rescale)
if reconstruction_guidance:
kwargs.update(step_kwargs)
latents = model.scheduler.step(noise_pred, t, latents, **kwargs, return_dict=False)[0]
if mask_edit is not None and inversion_guidance and (recon_t > recon_end_t and t < recon_t) or (recon_t < recon_end_t and t > -recon_t):
if use_localblend_mask:
assert hasattr(controller,"layer_fusion")
if save_heatmap and i%10==0:
sns.heatmap(controller.layer_fusion.mask[0][0].clone().cpu(), cmap='coolwarm')
plt.savefig(f'./vis/prox_inv/heatmap0_localblendmask_{i}.png')
plt.clf()
sns.heatmap(controller.layer_fusion.mask[1][0].clone().cpu(), cmap='coolwarm')
plt.savefig(f'./vis/prox_inv/heatmap1_localblendmask_{i}.png')
plt.clf()
layer_fusion_mask=controller.layer_fusion.mask.float()
layer_fusion_mask[0]=layer_fusion_mask[1]
recon_mask=1-layer_fusion_mask.expand_as(latents)
else:
recon_mask = 1 - mask_edit
target_latents=x_stars[len(x_stars)-i-2].expand_as(latents)
# if target_latents有四维
if len(target_latents.shape)==4:
target_latents=target_latents[0]
latents = latents - recon_lr * (latents - target_latents) * recon_mask
# controller
if controller is not None:
latents = controller.step_callback(latents)
return latents.to(model.unet.dtype)