import inspect
from copy import deepcopy
from typing import Optional, Union
import numpy as np
import torch
from diffusers import AutoPipelineForText2Image
from mmengine import print_log
from mmengine.model import BaseModel
from peft import get_peft_model
from torch import nn
from diffengine.models.archs import create_peft_config
from diffengine.registry import MODELS
@MODELS.register_module()
[docs]class KandinskyV22Prior(BaseModel):
"""KandinskyV22 Prior.
Args:
----
tokenizer (dict): Config of tokenizer.
scheduler (dict): Config of scheduler.
text_encoder (dict): Config of text encoder.
image_encoder (dict): Config of image encoder.
prior (dict): Config of prior.
decoder_model (str): pretrained model name of decoder.
Defaults to "kandinsky-community/kandinsky-2-2-decoder".
prior_model (str): pretrained model name of prior.
Defaults to "kandinsky-community/kandinsky-2-2-prior".
loss (dict): Config of loss. Defaults to
``dict(type='L2Loss', loss_weight=1.0)``.
prior_lora_config (dict, optional): The LoRA config dict for Prior.
example. dict(type="LoRA", r=4). `type` is chosen from `LoRA`,
`LoHa`, `LoKr`. Other config are same as the config of PEFT.
https://github.com/huggingface/peft
Defaults to None.
prior_loss_weight (float): The weight of prior preservation loss.
It works when training dreambooth with class images.
data_preprocessor (dict, optional): The pre-process config of
:class:`SDDataPreprocessor`.
noise_generator (dict, optional): The noise generator config.
Defaults to ``dict(type='WhiteNoise')``.
timesteps_generator (dict, optional): The timesteps generator config.
Defaults to ``dict(type='TimeSteps')``.
input_perturbation_gamma (float): The gamma of input perturbation.
The recommended value is 0.1 for Input Perturbation.
Defaults to 0.0.
gradient_checkpointing (bool): Whether or not to use gradient
checkpointing to save memory at the expense of slower backward
pass. Defaults to False.
enable_xformers (bool): Whether or not to enable memory efficient
attention. Defaults to False.
"""
def __init__(
self,
tokenizer: dict,
scheduler: dict,
text_encoder: dict,
image_encoder: dict,
prior: dict,
decoder_model: str = "kandinsky-community/kandinsky-2-2-decoder",
prior_model: str = "kandinsky-community/kandinsky-2-2-prior",
loss: dict | None = None,
prior_lora_config: dict | None = None,
prior_loss_weight: float = 1.,
data_preprocessor: dict | nn.Module | None = None,
noise_generator: dict | None = None,
timesteps_generator: dict | None = None,
input_perturbation_gamma: float = 0.0,
*,
gradient_checkpointing: bool = False,
enable_xformers: bool = False,
) -> None:
if data_preprocessor is None:
data_preprocessor = {"type": "SDDataPreprocessor"}
if noise_generator is None:
noise_generator = {}
if timesteps_generator is None:
timesteps_generator = {}
if loss is None:
loss = {}
super().__init__(data_preprocessor=data_preprocessor)
assert gradient_checkpointing is False, (
"KandinskyV22Prior does not support gradient checkpointing.")
self.decoder_model = decoder_model
self.prior_lora_config = deepcopy(prior_lora_config)
self.prior_loss_weight = prior_loss_weight
self.gradient_checkpointing = gradient_checkpointing
self.input_perturbation_gamma = input_perturbation_gamma
self.enable_xformers = enable_xformers
if not isinstance(loss, nn.Module):
loss = MODELS.build(
loss,
default_args={"type": "L2Loss", "loss_weight": 1.0})
self.loss_module: nn.Module = loss
self.tokenizer = MODELS.build(tokenizer,
default_args={"pretrained_model_name_or_path": prior_model,
} if not inspect.isclass(tokenizer.get("type")) else None)
self.scheduler = MODELS.build(
scheduler,
default_args={"pretrained_model_name_or_path": prior_model,
} if not inspect.isclass(scheduler.get("type")) else None)
self.text_encoder = MODELS.build(
text_encoder,
default_args={"pretrained_model_name_or_path": prior_model,
} if not inspect.isclass(text_encoder.get("type")) else None)
self.image_encoder = MODELS.build(
image_encoder,
default_args={"pretrained_model_name_or_path": prior_model,
} if not inspect.isclass(image_encoder.get("type")) else None)
self.prior = MODELS.build(
prior,
default_args={"pretrained_model_name_or_path": prior_model,
} if not inspect.isclass(prior.get("type")) else None)
self.noise_generator = MODELS.build(
noise_generator,
default_args={"type": "WhiteNoise"})
self.timesteps_generator = MODELS.build(
timesteps_generator,
default_args={"type": "TimeSteps"})
self.register_buffer("clip_mean", self.prior.clip_mean.clone())
self.register_buffer("clip_std", self.prior.clip_std.clone())
self.prepare_model()
self.set_lora()
self.set_xformers()
[docs] def set_lora(self) -> None:
"""Set LORA for model."""
if self.prior_lora_config is not None:
prior_lora_config = create_peft_config(self.prior_lora_config)
self.prior = get_peft_model(self.prior, prior_lora_config)
self.prior.print_trainable_parameters()
[docs] def prepare_model(self) -> None:
"""Prepare model for training.
Disable gradient for some models.
"""
self.text_encoder.requires_grad_(requires_grad=False)
print_log("Set Text Encoder untrainable.", "current")
self.image_encoder.requires_grad_(requires_grad=False)
print_log("Set Image Encoder untrainable.", "current")
@property
[docs] def device(self) -> torch.device:
"""Get device information.
Returns
-------
torch.device: device.
"""
return next(self.parameters()).device
@torch.no_grad()
[docs] def infer(self,
prompt: list[str],
negative_prompt: str | None = None,
height: int | None = None,
width: int | None = None,
num_inference_steps: int = 50,
output_type: str = "pil",
**kwargs) -> list[np.ndarray]:
"""Inference function.
Args:
----
prompt (`List[str]`):
The prompt or prompts to guide the image generation.
negative_prompt (`Optional[str]`):
The prompt or prompts to guide the image generation.
Defaults to None.
height (int, optional):
The height in pixels of the generated image. Defaults to None.
width (int, optional):
The width in pixels of the generated image. Defaults to None.
num_inference_steps (int): Number of inference steps.
Defaults to 50.
output_type (str): The output format of the generate image.
Choose between 'pil' and 'latent'. Defaults to 'pil'.
**kwargs: Other arguments.
"""
if height is None:
height = 512
if width is None:
width = 512
pipeline = AutoPipelineForText2Image.from_pretrained(
self.decoder_model,
prior_image_encoder=self.image_encoder,
prior_text_encoder=self.text_encoder,
prior_tokenizer=self.tokenizer,
prior_prior=self.prior,
torch_dtype=torch.float32,
)
pipeline.set_progress_bar_config(disable=True)
images = []
for p in prompt:
image = pipeline(
p,
negative_prompt=negative_prompt,
num_inference_steps=num_inference_steps,
height=height,
width=width,
output_type=output_type,
**kwargs).images[0]
if output_type == "latent":
images.append(image)
else:
images.append(np.array(image))
del pipeline
torch.cuda.empty_cache()
return images
[docs] def val_step(
self,
data: Union[tuple, dict, list] # noqa
) -> list:
"""Val step."""
msg = "val_step is not implemented now, please use infer."
raise NotImplementedError(msg)
[docs] def test_step(
self,
data: Union[tuple, dict, list] # noqa
) -> list:
"""Test step."""
msg = "test_step is not implemented now, please use infer."
raise NotImplementedError(msg)
[docs] def loss(self,
model_pred: torch.Tensor,
noise: torch.Tensor, # noqa
latents: torch.Tensor,
timesteps: torch.Tensor,
weight: torch.Tensor | None = None) -> dict[str, torch.Tensor]:
"""Calculate loss."""
gt = latents
loss_dict = {}
# calculate loss in FP32
if self.loss_module.use_snr:
loss = self.loss_module(
model_pred.float(),
gt.float(),
timesteps,
self.scheduler.alphas_cumprod,
self.scheduler.config.prediction_type,
weight=weight)
else:
loss = self.loss_module(
model_pred.float(), gt.float(), weight=weight)
loss_dict["loss"] = loss
return loss_dict
[docs] def forward(
self,
inputs: dict,
data_samples: Optional[list] = None, # noqa
mode: str = "loss") -> dict:
"""Forward function.
Args:
----
inputs (dict): The input dict.
data_samples (Optional[list], optional): The data samples.
Defaults to None.
mode (str, optional): The mode. Defaults to "loss".
Returns:
-------
dict: The loss dict.
"""
assert mode == "loss"
inputs_text = self.tokenizer(
inputs["text"],
max_length=self.tokenizer.model_max_length,
padding="max_length",
truncation=True,
return_tensors="pt")
num_batches = len(inputs["img"])
if "result_class_image" in inputs:
# use prior_loss_weight
weight = torch.cat([
torch.ones((num_batches // 2, )),
torch.ones((num_batches // 2, )) * self.prior_loss_weight,
]).to(self.device).float().reshape(-1, 1, 1, 1)
else:
weight = None
with torch.no_grad():
text_encoder_output = self.text_encoder(
inputs_text.input_ids.to(self.device))
prompt_embeds = text_encoder_output.text_embeds
text_encoder_hidden_states = text_encoder_output.last_hidden_state
image_embeds = self.image_encoder(inputs["img"]).image_embeds
noise = self.noise_generator(image_embeds)
timesteps = self.timesteps_generator(self.scheduler, num_batches,
self.device)
image_embeds = (image_embeds - self.clip_mean) / self.clip_std
noisy_latents = self._preprocess_model_input(image_embeds, noise, timesteps)
model_pred = self.prior(
noisy_latents,
timesteps,
proj_embedding=prompt_embeds,
encoder_hidden_states=text_encoder_hidden_states,
attention_mask=inputs_text.attention_mask.to(self.device)).predicted_image_embedding
return self.loss(model_pred, noise, image_embeds, timesteps, weight)