StableMaterials / pipeline.py
gvecchio's picture
Add model
0f95e92
import contextlib
import inspect
from typing import Any, Dict, List, Optional, Union, get_args
import numpy as np
import torch
import torch.nn.functional as F
import torchvision.transforms.functional as TF
from diffusers.image_processor import PipelineImageInput, VaeImageProcessor
from diffusers.loaders import FromSingleFileMixin
from diffusers.models.transformers import Transformer2DModel
from diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion import (
rescale_noise_cfg,
retrieve_timesteps,
)
from diffusers.schedulers import KarrasDiffusionSchedulers
from diffusers.utils import (
BaseOutput,
deprecate,
logging,
)
from diffusers.utils.torch_utils import randn_tensor
from PIL import (
Image,
Jpeg2KImagePlugin,
JpegImagePlugin,
PngImagePlugin,
TiffImagePlugin,
)
from transformers import (
CLIPImageProcessor,
CLIPTextModel,
CLIPTokenizer,
CLIPVisionModel,
)
from diffusers import AutoencoderKL, DiffusionPipeline, UNet2DConditionModel
logger = logging.get_logger(__name__) # pylint: disable=invalid-name
from dataclasses import dataclass
ImageInput = Union[
PipelineImageInput,
JpegImagePlugin.JpegImageFile,
Jpeg2KImagePlugin.Jpeg2KImageFile,
PngImagePlugin.PngImageFile,
TiffImagePlugin.TiffImageFile,
]
import math
def postprocess(
image: torch.FloatTensor,
output_type: str = "pil",
):
"""
Postprocess the image output from tensor to `output_type`.
Args:
image (`torch.FloatTensor`):
The image input, should be a pytorch tensor with shape `B x C x H x W`.
output_type (`str`, *optional*, defaults to `pil`):
The output type of the image, can be one of `pil`, `np`, `pt`, `latent`.
Returns:
`PIL.Image.Image`, `np.ndarray` or `torch.FloatTensor`:
The postprocessed image.
"""
if not isinstance(image, torch.Tensor):
raise ValueError(
f"Input for postprocessing is in incorrect format: {type(image)}. We only support pytorch tensor"
)
if output_type not in ["latent", "pt", "np", "pil"]:
deprecation_message = (
f"the output_type {output_type} is outdated and has been set to `np`. Please make sure to set it to one of these instead: "
"`pil`, `np`, `pt`, `latent`"
)
deprecate(
"Unsupported output_type", "1.0.0", deprecation_message, standard_warn=False
)
output_type = "np"
image = image.detach().cpu()
image = image.to(torch.float32)
if output_type == "latent":
return image
# denormalize the image
image = image * 0.5 + 0.5 # .clamp(0, 1)
materials = []
for i in range(image.shape[0]):
material = StableMaterialsMaterial()
material.init_from_tensor(image[i], mode=output_type)
materials.append(material)
return materials
@dataclass
class StableMaterialsMaterial:
basecolor: torch.FloatTensor
normal: torch.FloatTensor
height: torch.FloatTensor
roughness: torch.FloatTensor
metallic: torch.FloatTensor
_mode: str = "tensor" # Default mode is tensor
def __init__(
self,
basecolor: Optional[Union[Image.Image, np.ndarray, torch.FloatTensor]] = None,
normal: Optional[Union[Image.Image, np.ndarray, torch.FloatTensor]] = None,
height: Optional[Union[Image.Image, np.ndarray, torch.FloatTensor]] = None,
roughness: Optional[Union[Image.Image, np.ndarray, torch.FloatTensor]] = None,
metallic: Optional[Union[Image.Image, np.ndarray, torch.FloatTensor]] = None,
mode: str = "tensor",
):
self._basecolor = self._to_pt(basecolor)
self._normal = self._to_pt(normal)
self._height = self._to_pt(height)
self._roughness = self._to_pt(roughness)
self._metallic = self._to_pt(metallic)
self._mode = mode
def init_from_tensor(self, image: torch.FloatTensor, mode: str = "tensor"):
assert image.shape[0] >= 8, "Input tensor should have at least 8 channels"
self._basecolor = image[:3].clamp(0, 1)
self._normal = self.compute_normal_map_z_component(image[3:5])
self._height = image[5:6].clamp(0, 1)
self._roughness = image[6:7].clamp(0, 1)
self._metallic = image[7:8].clamp(0, 1)
self._mode = mode
def resize(self, size, antialias=True):
self._basecolor = TF.resize(self._basecolor, size, antialias=antialias)
self._normal = TF.resize(self._normal, size, antialias=antialias)
self._height = TF.resize(self._height, size, antialias=antialias)
self._roughness = TF.resize(self._roughness, size, antialias=antialias)
self._metallic = TF.resize(self._metallic, size, antialias=antialias)
return self
def tile(self, num_tiles):
self._basecolor = self._basecolor.repeat(1, num_tiles, num_tiles)
self._normal = self._normal.repeat(1, num_tiles, num_tiles)
self._height = self._height.repeat(1, num_tiles, num_tiles)
self._roughness = self._roughness.repeat(1, num_tiles, num_tiles)
self._metallic = self._metallic.repeat(1, num_tiles, num_tiles)
return self
def _to_numpy(self, image: torch.FloatTensor):
if image is None:
return None
return image.numpy()
def _to_pil(self, image: torch.FloatTensor, mode: str = "RGB"):
if image is None:
return None
return TF.to_pil_image(image).convert(mode)
def _to_pt(self, image):
if image is None:
return None
if isinstance(image, np.ndarray):
image = torch.from_numpy(image)
elif isinstance(image, Image.Image):
image = TF.to_tensor(image)
return image.cpu()
def compute_normal_map_z_component(self, normal: torch.FloatTensor):
normal = normal * 2 - 1
sum_sq = (normal**2).sum(dim=0, keepdim=True)[0]
z = torch.zeros_like(sum_sq)
mask = sum_sq <= 1
z[mask] = torch.sqrt(1 - sum_sq[mask])
mask_outlier = sum_sq > 1
scale_factor = torch.sqrt(sum_sq[mask_outlier])
normal[:, mask_outlier] = normal[:, mask_outlier] / scale_factor
normal = torch.cat([normal, z.unsqueeze(0)], dim=0)
normal = normal * 0.5 + 0.5
return normal.clamp(0, 1)
def _convert(self, image, mode="RGB"):
if self._mode == "numpy":
return self._to_numpy(image)
elif self._mode == "pil":
return self._to_pil(image, mode)
return image
@property
def size(self):
return list(self._basecolor.shape[-2:])
@property
def basecolor(self):
return self._convert(self._basecolor, mode="RGB")
@property
def normal(self):
return self._convert(self._normal, mode="RGB")
@property
def height(self):
return self._convert(self._height, mode="L")
@property
def roughness(self):
return self._convert(self._roughness, mode="L")
@property
def metallic(self):
return self._convert(self._metallic, mode="L")
def as_dict(self):
return {
"basecolor": self.basecolor,
"normal": self.normal,
"height": self.height,
"roughness": self.roughness,
"metallic": self.metallic,
}
def as_list(self):
return [
self.basecolor,
self.normal,
self.height,
self.roughness,
self.metallic,
]
def as_tensor(self):
return torch.cat(
[
self._basecolor,
self._normal[:2],
self._height,
self._roughness,
self._metallic,
],
dim=0,
)
@dataclass
class StableMaterialsPipelineOutput(BaseOutput):
"""
Output class for Stable Diffusion pipelines.
Args:
images (`List[PIL.Image.Image]` or `np.ndarray`)
List of denoised PIL images of length `batch_size` or NumPy array of shape `(batch_size, height, width,
num_channels)`.
"""
images: List[StableMaterialsMaterial]
def patch(x, patch_factor=2):
if isinstance(x, (list, tuple)):
pass
b, c, h, w = x.shape
patch_size = h // patch_factor
x = x.unfold(2, patch_size, patch_size).unfold(3, patch_size, patch_size)
x = x.permute(0, 2, 3, 1, 4, 5).contiguous().view(-1, c, patch_size, patch_size)
n_patches = x.shape[0] // b
return x, (b, h), n_patches, patch_size
def unpatch(x, b, h, n_patches, patch_size=32):
if isinstance(x, (list, tuple)):
if len(x) == 1:
x = x[0]
else:
pass
factor = patch_size / x.shape[-1]
h, w = int(h / factor), int(h / factor)
c, patch_size = x.shape[1], x.shape[2]
n_patches = x.shape[0] // b
x = x.reshape(b, n_patches, c, patch_size, patch_size)
x = x.permute(0, 2, 3, 4, 1).contiguous().view(b, c * patch_size * patch_size, -1)
x = F.fold(
x,
output_size=(h, w),
kernel_size=patch_size,
stride=patch_size,
)
return x
def roll(x):
roll_h = torch.randint(0, 256, (1,)).item() // 2 * 2
roll_w = torch.randint(0, 256, (1,)).item() // 2 * 2
x = torch.roll(x, shifts=(roll_h, roll_w), dims=(2, 3))
return x, (roll_h, roll_w)
def unroll(x, roll_h, roll_w, factor=1.0):
roll_h = int(roll_h * factor)
roll_w = int(roll_w * factor)
x = torch.roll(x, shifts=(-roll_h, -roll_w), dims=(2, 3))
return x
@contextlib.contextmanager
def rolled_conv(enabled=True):
conv = torch.nn.Conv2d
if enabled:
# Save the original conv's constructor
orig_forward = conv.forward
def forward(self, x, *args, **kwargs):
x, (roll_h, roll_w) = roll(x)
pad = 4
x = F.pad(x, (pad, pad, pad, pad), mode="circular")
h = x.shape[-2]
x = orig_forward(self, x, *args, **kwargs)
h1 = x.shape[-2]
factor = h1 / h
pad = int(pad * factor)
x = x[..., pad:-pad, pad:-pad]
x = unroll(x, roll_h, roll_w, factor)
return x
# Patch conv's constructor
conv.forward = forward
# conv.__init__ = __init__
yield conv
# Restore the original conv's constructor
conv.forward = orig_forward
else:
# Use the original conv
yield conv
@contextlib.contextmanager
def tiled_attn(enabled=True, scale_multiplier=4):
conv = Transformer2DModel
if enabled:
# Save the original conv's constructor
orig_forward = conv.forward
# mult = scale_multiplier
def forward(self, hidden_states, encoder_hidden_states, *args, **kwargs):
hidden_states, (roll_h, roll_w) = roll(hidden_states)
hidden_states, (b, h), n_patches, patch_size = patch(
hidden_states, self.scale_multiplier
)
encoder_hidden_states = encoder_hidden_states.repeat_interleave(
n_patches, dim=0
)
chunks = math.ceil(len(hidden_states) / 8)
hidden_states = hidden_states.chunk(chunks, dim=0)
encoder_hidden_states = encoder_hidden_states.chunk(chunks, dim=0)
result = []
for i in range(chunks):
result.append(
orig_forward(
self,
hidden_states[i],
encoder_hidden_states[i],
*args,
**kwargs,
)[0]
)
hidden_states = torch.cat(result, dim=0)
hidden_states = unpatch(hidden_states, b, h, n_patches, patch_size)
hidden_states = unroll(hidden_states, roll_h, roll_w)
return (hidden_states,)
# Patch conv's constructor
conv.scale_multiplier = scale_multiplier
conv.forward = forward
yield conv
# Restore the original conv's constructor
conv.forward = orig_forward
else:
# Use the original conv
yield conv
class StableMaterialsPipeline(DiffusionPipeline, FromSingleFileMixin):
model_cpu_offload_seq = "prompt_encoder->unet->vae"
def __init__(
self,
vae: AutoencoderKL,
unet: UNet2DConditionModel,
# prompt_encoder: nn.Module,
scheduler: KarrasDiffusionSchedulers,
text_encoder: CLIPTextModel,
tokenizer: CLIPTokenizer,
vision_encoder: CLIPVisionModel,
processor: CLIPImageProcessor,
):
super().__init__()
self.register_modules(
vae=vae,
unet=unet,
# prompt_encoder=prompt_encoder,
scheduler=scheduler,
# Conditioning modules
tokenizer=tokenizer,
processor=processor,
text_encoder=text_encoder,
vision_encoder=vision_encoder,
)
self.vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1)
self.image_processor = VaeImageProcessor(vae_scale_factor=self.vae_scale_factor)
def enable_vae_slicing(self):
r"""
Enable sliced VAE decoding. When this option is enabled, the VAE will split the input tensor in slices to
compute decoding in several steps. This is useful to save some memory and allow larger batch sizes.
"""
self.vae.enable_slicing()
def disable_vae_slicing(self):
r"""
Disable sliced VAE decoding. If `enable_vae_slicing` was previously enabled, this method will go back to
computing decoding in one step.
"""
self.vae.disable_slicing()
def enable_vae_tiling(self):
r"""
Enable tiled VAE decoding. When this option is enabled, the VAE will split the input tensor into tiles to
compute decoding and encoding in several steps. This is useful for saving a large amount of memory and to allow
processing larger images.
"""
self.vae.enable_tiling()
def disable_vae_tiling(self):
r"""
Disable tiled VAE decoding. If `enable_vae_tiling` was previously enabled, this method will go back to
computing decoding in one step.
"""
self.vae.disable_tiling()
def __encode_text(self, text):
inputs = self.tokenizer(text, padding=True, return_tensors="pt")
inputs["input_ids"] = inputs["input_ids"].to(self.device)
inputs["attention_mask"] = inputs["attention_mask"].to(self.device)
outputs = self.text_encoder(**inputs)
return outputs.text_embeds.unsqueeze(1)
def __encode_image(self, image):
inputs = self.processor(images=image, return_tensors="pt")
inputs["pixel_values"] = inputs["pixel_values"].to(self.device)
outputs = self.vision_encoder(**inputs)
return outputs.image_embeds.unsqueeze(1)
def __encode_prompt(
self,
prompt,
):
if type(prompt) != list:
prompt = [prompt]
embs = []
for prompt in prompt:
if isinstance(prompt, str):
embs.append(self.__encode_text(prompt))
elif type(prompt) in get_args(ImageInput):
embs.append(self.__encode_image(prompt))
else:
raise NotImplementedError
return torch.cat(embs, dim=0)
def encode_prompt(
self,
prompt,
device,
num_images_per_prompt,
do_classifier_free_guidance,
negative_prompt=None,
prompt_embeds: Optional[torch.FloatTensor] = None,
negative_prompt_embeds: Optional[torch.FloatTensor] = 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.
"""
if (
prompt is not None
and isinstance(prompt, str)
or isinstance(prompt, Image.Image)
):
batch_size = 1
elif prompt is not None and isinstance(prompt, list):
batch_size = len(prompt)
else:
batch_size = prompt_embeds.shape[0]
if prompt_embeds is None:
prompt_embeds = self.__encode_prompt(prompt)
if self.unet is not None:
prompt_embeds_dtype = self.unet.dtype
else:
prompt_embeds_dtype = prompt_embeds.dtype
prompt_embeds = prompt_embeds.to(dtype=prompt_embeds_dtype, device=device)
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
)
if do_classifier_free_guidance and negative_prompt_embeds is None:
uncond_tokens: List[str]
if negative_prompt is None:
# uncond_tokens = [""] * batch_size
uncond_tokens = [Image.new("RGB", (512, 512), (0, 0, 0))] * batch_size
elif isinstance(negative_prompt, str):
uncond_tokens = [negative_prompt] * batch_size
elif len(negative_prompt) != batch_size:
raise ValueError(
"The `negative_prompt` must be a string, a list of strings of length `batch_size`, or `None`."
)
else:
uncond_tokens = negative_prompt
negative_prompt_embeds = self.__encode_prompt(uncond_tokens)
# get unconditional embeddings for classifier free guidance
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=prompt_embeds_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
)
return prompt_embeds, negative_prompt_embeds
def decode_latents(self, latents):
deprecation_message = "The decode_latents method is deprecated and will be removed in 1.0.0. Please use VaeImageProcessor.postprocess(...) instead"
deprecate("decode_latents", "1.0.0", deprecation_message, standard_warn=False)
latents = 1 / self.vae.config.scaling_factor * latents
image = self.vae.decode(latents, return_dict=False)[0]
image = (image / 2 + 0.5).clamp(0, 1)
# we always cast to float32 as this does not cause significant overhead and is compatible with bfloat16
image = image.cpu().permute(0, 2, 3, 1).float().numpy()
return image
def prepare_extra_step_kwargs(self, generator, eta):
# 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
# check if the scheduler accepts generator
accepts_generator = "generator" in set(
inspect.signature(self.scheduler.step).parameters.keys()
)
if accepts_generator:
extra_step_kwargs["generator"] = generator
return extra_step_kwargs
def check_inputs(
self,
prompt,
height,
width,
negative_prompt=None,
prompt_embeds=None,
negative_prompt_embeds=None,
):
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 prompt is not None and prompt_embeds is not None:
raise ValueError(
f"Cannot forward both `prompt`: {prompt} and `prompt_embeds`: {prompt_embeds}. Please make sure to"
" only forward one of the two."
)
elif prompt is None and prompt_embeds is None:
raise ValueError(
"Provide either `prompt` or `prompt_embeds`. Cannot leave both `prompt` and `prompt_embeds` undefined."
)
elif prompt is not None and (not isinstance(prompt, (str, list, Image.Image))):
raise ValueError(
f"`prompt` has to be of type `str` or `list` but is {type(prompt)}"
)
if negative_prompt is not None and negative_prompt_embeds is not None:
raise ValueError(
f"Cannot forward both `negative_prompt`: {negative_prompt} and `negative_prompt_embeds`:"
f" {negative_prompt_embeds}. Please make sure to only forward one of the two."
)
if prompt_embeds is not None and negative_prompt_embeds is not None:
if prompt_embeds.shape != negative_prompt_embeds.shape:
raise ValueError(
"`prompt_embeds` and `negative_prompt_embeds` must have the same shape when passed directly, but"
f" got: `prompt_embeds` {prompt_embeds.shape} != `negative_prompt_embeds`"
f" {negative_prompt_embeds.shape}."
)
def prepare_latents(
self,
batch_size,
num_channels_latents,
height,
width,
dtype,
device,
generator,
latents=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 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 enable_freeu(self, s1: float, s2: float, b1: float, b2: float):
r"""Enables the FreeU mechanism as in https://arxiv.org/abs/2309.11497.
The suffixes after the scaling factors represent the stages where they are being applied.
Please refer to the [official repository](https://github.com/ChenyangSi/FreeU) for combinations of the values
that are known to work well for different pipelines such as Stable Diffusion v1, v2, and Stable Diffusion XL.
Args:
s1 (`float`):
Scaling factor for stage 1 to attenuate the contributions of the skip features. This is done to
mitigate "oversmoothing effect" in the enhanced denoising process.
s2 (`float`):
Scaling factor for stage 2 to attenuate the contributions of the skip features. This is done to
mitigate "oversmoothing effect" in the enhanced denoising process.
b1 (`float`): Scaling factor for stage 1 to amplify the contributions of backbone features.
b2 (`float`): Scaling factor for stage 2 to amplify the contributions of backbone features.
"""
if not hasattr(self, "unet"):
raise ValueError("The pipeline must have `unet` for using FreeU.")
self.unet.enable_freeu(s1=s1, s2=s2, b1=b1, b2=b2)
def disable_freeu(self):
"""Disables the FreeU mechanism if enabled."""
self.unet.disable_freeu()
# Copied from diffusers.pipelines.latent_consistency_models.pipeline_latent_consistency_text2img.LatentConsistencyModelPipeline.get_guidance_scale_embedding
def get_guidance_scale_embedding(self, w, embedding_dim=512, dtype=torch.float32):
"""
See https://github.com/google-research/vdm/blob/dc27b98a554f65cdc654b800da5aa1846545d41b/model_vdm.py#L298
Args:
timesteps (`torch.Tensor`):
generate embedding vectors at these timesteps
embedding_dim (`int`, *optional*, defaults to 512):
dimension of the embeddings to generate
dtype:
data type of the generated embeddings
Returns:
`torch.FloatTensor`: Embedding vectors with shape `(len(timesteps), embedding_dim)`
"""
assert len(w.shape) == 1
w = w * 1000.0
half_dim = embedding_dim // 2
emb = torch.log(torch.tensor(10000.0)) / (half_dim - 1)
emb = torch.exp(torch.arange(half_dim, dtype=dtype) * -emb)
emb = w.to(dtype)[:, None] * emb[None, :]
emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=1)
if embedding_dim % 2 == 1: # zero pad
emb = torch.nn.functional.pad(emb, (0, 1))
assert emb.shape == (w.shape[0], embedding_dim)
return emb
@property
def guidance_scale(self):
return self._guidance_scale
@property
def guidance_rescale(self):
return self._guidance_rescale
# 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.
@property
def do_classifier_free_guidance(self):
return self._guidance_scale > 1 and self.unet.config.time_cond_proj_dim is None
@property
def cross_attention_kwargs(self):
return self._cross_attention_kwargs
@property
def num_timesteps(self):
return self._num_timesteps
@property
def interrupt(self):
return self._interrupt
@torch.no_grad()
# @replace_example_docstring(EXAMPLE_DOC_STRING)
def __call__(
self,
prompt: Union[
str, List[str], PipelineImageInput, List[PipelineImageInput]
] = None,
height: Optional[int] = None,
width: Optional[int] = None,
tileable: bool = False,
patched: bool = False,
num_inference_steps: int = 50,
timesteps: List[int] = None,
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,
output_type: Optional[str] = "pil",
return_dict: bool = True,
cross_attention_kwargs: Optional[Dict[str, Any]] = None,
guidance_rescale: float = 0.0,
**kwargs,
):
# 0. Default height and width to unet
height = height or self.unet.config.sample_size * self.vae_scale_factor
width = width or self.unet.config.sample_size * self.vae_scale_factor
# 1. Check inputs. Raise error if not correct
self.check_inputs(
prompt,
height,
width,
negative_prompt,
prompt_embeds,
negative_prompt_embeds,
)
self._guidance_scale = guidance_scale
self._guidance_rescale = guidance_rescale
self._cross_attention_kwargs = cross_attention_kwargs
self._interrupt = False
# 2. Define call parameters
if prompt is not None and (
isinstance(prompt, str) or isinstance(prompt, Image.Image)
):
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
# 3. Encode input prompt
prompt_embeds, negative_prompt_embeds = self.encode_prompt(
prompt,
device,
num_images_per_prompt,
self.do_classifier_free_guidance,
negative_prompt,
prompt_embeds=prompt_embeds,
negative_prompt_embeds=negative_prompt_embeds,
)
# 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 self.do_classifier_free_guidance:
prompt_embeds = torch.cat([negative_prompt_embeds, prompt_embeds])
# 4. Prepare timesteps
timesteps, num_inference_steps = retrieve_timesteps(
self.scheduler, num_inference_steps, device, 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,
)
# 6. Prepare extra step kwargs.
extra_step_kwargs = self.prepare_extra_step_kwargs(generator, eta)
# 6.2 Optionally get Guidance Scale Embedding
timestep_cond = None
if self.unet.config.time_cond_proj_dim is not None:
guidance_scale_tensor = torch.tensor(self.guidance_scale - 1).repeat(
batch_size * num_images_per_prompt
)
timestep_cond = self.get_guidance_scale_embedding(
guidance_scale_tensor, embedding_dim=self.unet.config.time_cond_proj_dim
).to(device=device, dtype=latents.dtype)
# 7. Denoising loop
self._num_timesteps = len(timesteps)
with self.progress_bar(total=num_inference_steps) as progress_bar:
for i, t in enumerate(timesteps):
if self.interrupt:
continue
# expand the latents if we are doing classifier free guidance
latent_model_input = (
torch.cat([latents] * 2)
if self.do_classifier_free_guidance
else latents
)
latent_model_input = self.scheduler.scale_model_input(
latent_model_input, t
)
scale_multiplier = (
latent_model_input.shape[-1]
) // self.unet.config.sample_size
past_mid = i >= len(timesteps) // 4
# predict the noise residual
with rolled_conv(enabled=(tileable & past_mid)):
with tiled_attn(enabled=patched, scale_multiplier=scale_multiplier):
noise_pred = self.unet(
latent_model_input,
t,
encoder_hidden_states=prompt_embeds,
timestep_cond=timestep_cond,
cross_attention_kwargs=self.cross_attention_kwargs,
return_dict=False,
)[0]
# perform guidance
if self.do_classifier_free_guidance:
noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
noise_pred = noise_pred_uncond + self.guidance_scale * (
noise_pred_text - noise_pred_uncond
)
if self.do_classifier_free_guidance and self.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=self.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]
# call the callback, if provided
if i == len(timesteps) - 1 or (i + 1) % self.scheduler.order == 0:
progress_bar.update()
if not output_type == "latent":
if tileable:
# decode padded latent to preserve tileability
l_height = height // self.vae_scale_factor
l_width = width // self.vae_scale_factor
pad = l_height // 4
latents = TF.center_crop(
latents.repeat(1, 1, 3, 3), (l_height + pad, l_width + pad)
)
# decode the latents
image = self.vae.decode(
latents / self.vae.config.scaling_factor,
return_dict=False,
generator=generator,
)[0]
# crop to original size
image = TF.center_crop(image, (height, width))
else:
image = latents
image = postprocess(image, output_type=output_type)
# Offload all models
self.maybe_free_model_hooks()
if not return_dict:
return image
return StableMaterialsPipelineOutput(images=image)