import inspect
from copy import deepcopy
from typing import Optional, Union
import numpy as np
import torch
from diffusers import PixArtAlphaPipeline
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 PixArtAlpha(BaseModel):
"""PixArt Alpha.
Args:
----
tokenizer (dict): Config of tokenizer.
scheduler (dict): Config of scheduler.
text_encoder (dict): Config of text encoder.
vae (dict): Config of vae.
transformer (dict): Config of transformer.
model (str): pretrained model name of stable diffusion.
Defaults to 'PixArt-alpha/PixArt-XL-2-1024-MS'.
loss (dict): Config of loss. Defaults to
``dict(type='L2Loss', loss_weight=1.0)``.
transformer_lora_config (dict, optional): The LoRA config dict for
Transformer. 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.
text_encoder_lora_config (dict, optional): The LoRA config dict for
Text Encoder. 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.
tokenizer_max_length (int): The max length of tokenizer.
Defaults to 120.
prediction_type (str): The prediction_type that shall be used for
training. Choose between 'epsilon' or 'v_prediction' or leave
`None`. If left to `None` the default prediction type of the
scheduler will be used. Defaults to None.
data_preprocessor (dict, optional): The pre-process config of
:class:`PixArtAlphaDataPreprocessor`.
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.
vae_batch_size (int): The batch size of vae. Defaults to 8.
finetune_text_encoder (bool, optional): Whether to fine-tune text
encoder. Defaults to False.
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,
vae: dict,
transformer: dict,
model: str = "PixArt-alpha/PixArt-XL-2-1024-MS",
loss: dict | None = None,
transformer_lora_config: dict | None = None,
text_encoder_lora_config: dict | None = None,
prior_loss_weight: float = 1.,
tokenizer_max_length: int = 120,
prediction_type: str | None = None,
data_preprocessor: dict | nn.Module | None = None,
noise_generator: dict | None = None,
timesteps_generator: dict | None = None,
input_perturbation_gamma: float = 0.0,
vae_batch_size: int = 8,
*,
finetune_text_encoder: bool = False,
gradient_checkpointing: bool = False,
enable_xformers: bool = False,
) -> None:
if data_preprocessor is None:
data_preprocessor = {"type": "PixArtAlphaDataPreprocessor"}
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)
if (
transformer_lora_config is not None) and (
text_encoder_lora_config is not None) and (
not finetune_text_encoder):
print_log(
"You are using LoRA for Transformer and text encoder. "
"But you are not set `finetune_text_encoder=True`. "
"We will set `finetune_text_encoder=True` for you.")
finetune_text_encoder = True
if text_encoder_lora_config is not None:
assert finetune_text_encoder, (
"If you want to use LoRA for text encoder, "
"you should set finetune_text_encoder=True."
)
if finetune_text_encoder and transformer_lora_config is not None:
assert text_encoder_lora_config is not None, (
"If you want to finetune text encoder with LoRA Transformer, "
"you should set text_encoder_lora_config."
)
self.model = model
self.transformer_lora_config = deepcopy(transformer_lora_config)
self.text_encoder_lora_config = deepcopy(text_encoder_lora_config)
self.finetune_text_encoder = finetune_text_encoder
self.prior_loss_weight = prior_loss_weight
self.tokenizer_max_length = tokenizer_max_length
self.gradient_checkpointing = gradient_checkpointing
self.input_perturbation_gamma = input_perturbation_gamma
self.enable_xformers = enable_xformers
self.vae_batch_size = vae_batch_size
if not isinstance(loss, nn.Module):
loss = MODELS.build(
loss,
default_args={"type": "L2Loss", "loss_weight": 1.0})
self.loss_module: nn.Module = loss
assert prediction_type in [None, "epsilon", "v_prediction"]
self.prediction_type = prediction_type
self.tokenizer = MODELS.build(
tokenizer,
default_args={
"pretrained_model_name_or_path": model,
} if not inspect.isclass(tokenizer.get("type")) else None)
self.scheduler = MODELS.build(
scheduler,
default_args={
"pretrained_model_name_or_path": model,
} if not inspect.isclass(scheduler.get("type")) else None)
self.text_encoder = MODELS.build(
text_encoder,
default_args={
"pretrained_model_name_or_path": model,
} if not inspect.isclass(text_encoder.get("type")) else None)
self.vae = MODELS.build(
vae,
default_args={
"pretrained_model_name_or_path": model,
} if not inspect.isclass(vae.get("type")) else None)
self.transformer = MODELS.build(
transformer,
default_args={
"pretrained_model_name_or_path": model,
} if not inspect.isclass(transformer.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.prepare_model()
self.set_lora()
self.set_xformers()
[docs] def set_lora(self) -> None:
"""Set LORA for model."""
if self.text_encoder_lora_config is not None:
text_encoder_lora_config = create_peft_config(
self.text_encoder_lora_config)
self.text_encoder = get_peft_model(
self.text_encoder, text_encoder_lora_config)
self.text_encoder.print_trainable_parameters()
if self.transformer_lora_config is not None:
transformer_lora_config = create_peft_config(self.transformer_lora_config)
self.transformer = get_peft_model(self.transformer, transformer_lora_config)
self.transformer.print_trainable_parameters()
[docs] def prepare_model(self) -> None:
"""Prepare model for training.
Disable gradient for some models.
"""
if self.gradient_checkpointing:
self.transformer.enable_gradient_checkpointing()
if self.finetune_text_encoder:
self.text_encoder.gradient_checkpointing_enable()
self.vae.requires_grad_(requires_grad=False)
print_log("Set VAE untrainable.", "current")
if not self.finetune_text_encoder:
self.text_encoder.requires_grad_(requires_grad=False)
print_log("Set Text 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.
"""
pipeline = PixArtAlphaPipeline.from_pretrained(
self.model,
vae=self.vae,
tokenizer=self.tokenizer,
transformer=self.transformer,
torch_dtype=torch.float32,
)
if self.finetune_text_encoder:
# TODO(takuoko): When parsing text_encoder directly, the # noqa
# results are different. So we need to parse here.
pipeline.text_encoder = self.text_encoder
pipeline.to(self.device)
if self.prediction_type is not None:
# set prediction_type of scheduler if defined
scheduler_args = {"prediction_type": self.prediction_type}
pipeline.scheduler = pipeline.scheduler.from_config(
pipeline.scheduler.config, **scheduler_args)
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,
latents: torch.Tensor,
timesteps: torch.Tensor,
weight: torch.Tensor | None = None) -> dict[str, torch.Tensor]:
"""Calculate loss."""
if self.prediction_type is not None:
# set prediction_type of scheduler if defined
self.scheduler.register_to_config(
prediction_type=self.prediction_type)
if self.scheduler.config.prediction_type == "epsilon":
gt = noise
elif self.scheduler.config.prediction_type == "v_prediction":
gt = self.scheduler.get_velocity(latents, noise, timesteps)
else:
msg = f"Unknown prediction type {self.scheduler.config.prediction_type}"
raise ValueError(msg)
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_vae(self, img: torch.Tensor, num_batches: int,
) -> torch.Tensor:
"""Forward vae."""
latents = [
self.vae.encode(
img[i : i + self.vae_batch_size],
).latent_dist.sample() for i in range(
0, num_batches, self.vae_batch_size)
]
latents = torch.cat(latents, dim=0)
return latents * self.vae.config.scaling_factor
[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"
text_inputs = self.tokenizer(
inputs["text"],
max_length=self.tokenizer_max_length,
padding="max_length",
truncation=True,
return_tensors="pt")
inputs["text"] = text_inputs.input_ids.to(self.device)
inputs["attention_mask"] = text_inputs.attention_mask.to(self.device)
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
latents = self._forward_vae(inputs["img"], num_batches)
noise = self.noise_generator(latents)
timesteps = self.timesteps_generator(self.scheduler, num_batches,
self.device)
noisy_latents = self._preprocess_model_input(latents, noise, timesteps)
encoder_hidden_states = self.text_encoder(
inputs["text"], attention_mask=inputs["attention_mask"])[0]
if self.transformer.config.sample_size == 128: # noqa
added_cond_kwargs = {"resolution": inputs["resolution"],
"aspect_ratio": inputs["aspect_ratio"]}
else:
added_cond_kwargs = {"resolution": None, "aspect_ratio": None}
model_pred = self.transformer(
noisy_latents,
encoder_hidden_states=encoder_hidden_states,
encoder_attention_mask=inputs["attention_mask"],
timestep=timesteps,
added_cond_kwargs=added_cond_kwargs).sample
latent_channels = self.transformer.config.in_channels
if self.transformer.config.out_channels // 2 == latent_channels:
model_pred = model_pred.chunk(2, dim=1)[0]
return self.loss(model_pred, noise, latents, timesteps, weight)