examples/research_projects/realfill/train_realfill.py

import argparse
import copy
import itertools
import logging
import math
import os
import random
import shutil
from pathlib import Path

import numpy as np
import torch
import torch.nn.functional as F
import torch.utils.checkpoint
import torchvision.transforms.v2 as transforms_v2
import transformers
from accelerate import Accelerator
from accelerate.logging import get_logger
from accelerate.utils import set_seed
from huggingface_hub import create_repo, upload_folder
from packaging import version
from peft import LoraConfig, PeftModel, get_peft_model
from PIL import Image
from PIL.ImageOps import exif_transpose
from torch.utils.data import Dataset
from tqdm.auto import tqdm
from transformers import AutoTokenizer, CLIPTextModel

import diffusers
from diffusers import (
    AutoencoderKL,
    DDPMScheduler,
    DPMSolverMultistepScheduler,
    StableDiffusionInpaintPipeline,
    UNet2DConditionModel,
)
from diffusers.optimization import get_scheduler
from diffusers.utils import check_min_version, is_wandb_available
from diffusers.utils.import_utils import is_xformers_available


if is_wandb_available():
    import wandb

# Will error if the minimal version of diffusers is not installed. Remove at your own risks.
check_min_version("0.20.1")

logger = get_logger(__name__)


def make_mask(images, resolution, times=30):
    mask, times = torch.ones_like(images[0:1, :, :]), np.random.randint(1, times)
    min_size, max_size, margin = np.array([0.03, 0.25, 0.01]) * resolution
    max_size = min(max_size, resolution - margin * 2)

    for _ in range(times):
        width = np.random.randint(int(min_size), int(max_size))
        height = np.random.randint(int(min_size), int(max_size))

        x_start = np.random.randint(int(margin), resolution - int(margin) - width + 1)
        y_start = np.random.randint(int(margin), resolution - int(margin) - height + 1)
        mask[:, y_start : y_start + height, x_start : x_start + width] = 0

    mask = 1 - mask if random.random() < 0.5 else mask
    return mask


def save_model_card(
    repo_id: str,
    images=None,
    base_model=str,
    repo_folder=None,
):
    img_str = ""
    for i, image in enumerate(images):
        image.save(os.path.join(repo_folder, f"image_{i}.png"))
        img_str += f"![img_{i}](./image_{i}.png)\n"

    yaml = f"""
---
license: creativeml-openrail-m
base_model: {base_model}
prompt: "a photo of sks"
tags:
- stable-diffusion-inpainting
- stable-diffusion-inpainting-diffusers
- text-to-image
- diffusers
- realfill
- diffusers-training
inference: true
---
    """
    model_card = f"""
# RealFill - {repo_id}

This is a realfill model derived from {base_model}. The weights were trained using [RealFill](https://realfill.github.io/).
You can find some example images in the following. \n
{img_str}
"""
    with open(os.path.join(repo_folder, "README.md"), "w") as f:
        f.write(yaml + model_card)


def log_validation(
    text_encoder,
    tokenizer,
    unet,
    args,
    accelerator,
    weight_dtype,
    epoch,
):
    logger.info(f"Running validation... \nGenerating {args.num_validation_images} images")

    # create pipeline (note: unet and vae are loaded again in float32)
    pipeline = StableDiffusionInpaintPipeline.from_pretrained(
        args.pretrained_model_name_or_path,
        tokenizer=tokenizer,
        revision=args.revision,
        torch_dtype=weight_dtype,
    )

    # set `keep_fp32_wrapper` to True because we do not want to remove
    # mixed precision hooks while we are still training
    pipeline.unet = accelerator.unwrap_model(unet, keep_fp32_wrapper=True)
    pipeline.text_encoder = accelerator.unwrap_model(text_encoder, keep_fp32_wrapper=True)
    pipeline.scheduler = DPMSolverMultistepScheduler.from_config(pipeline.scheduler.config)

    pipeline = pipeline.to(accelerator.device)
    pipeline.set_progress_bar_config(disable=True)

    # run inference
    generator = None if args.seed is None else torch.Generator(device=accelerator.device).manual_seed(args.seed)

    target_dir = Path(args.train_data_dir) / "target"
    target_image, target_mask = target_dir / "target.png", target_dir / "mask.png"
    image, mask_image = Image.open(target_image), Image.open(target_mask)

    if image.mode != "RGB":
        image = image.convert("RGB")

    images = []
    for _ in range(args.num_validation_images):
        image = pipeline(
            prompt="a photo of sks",
            image=image,
            mask_image=mask_image,
            num_inference_steps=25,
            guidance_scale=5,
            generator=generator,
        ).images[0]
        images.append(image)

    for tracker in accelerator.trackers:
        if tracker.name == "tensorboard":
            np_images = np.stack([np.asarray(img) for img in images])
            tracker.writer.add_images("validation", np_images, epoch, dataformats="NHWC")
        if tracker.name == "wandb":
            tracker.log({"validation": [wandb.Image(image, caption=str(i)) for i, image in enumerate(images)]})

    del pipeline
    torch.cuda.empty_cache()

    return images


def parse_args(input_args=None):
    parser = argparse.ArgumentParser(description="Simple example of a training script.")
    parser.add_argument(
        "--pretrained_model_name_or_path",
        type=str,
        default=None,
        required=True,
        help="Path to pretrained model or model identifier from huggingface.co/models.",
    )
    parser.add_argument(
        "--revision",
        type=str,
        default=None,
        required=False,
        help="Revision of pretrained model identifier from huggingface.co/models.",
    )
    parser.add_argument(
        "--tokenizer_name",
        type=str,
        default=None,
        help="Pretrained tokenizer name or path if not the same as model_name",
    )
    parser.add_argument(
        "--train_data_dir",
        type=str,
        default=None,
        required=True,
        help="A folder containing the training data of images.",
    )
    parser.add_argument(
        "--num_validation_images",
        type=int,
        default=4,
        help="Number of images that should be generated during validation with `validation_conditioning`.",
    )
    parser.add_argument(
        "--validation_steps",
        type=int,
        default=100,
        help=(
            "Run realfill validation every X steps. RealFill validation consists of running the conditioning"
            " `args.validation_conditioning` multiple times: `args.num_validation_images`."
        ),
    )
    parser.add_argument(
        "--output_dir",
        type=str,
        default="realfill-model",
        help="The output directory where the model predictions and checkpoints will be written.",
    )
    parser.add_argument("--seed", type=int, default=None, help="A seed for reproducible training.")
    parser.add_argument(
        "--resolution",
        type=int,
        default=512,
        help=(
            "The resolution for input images, all the images in the train/validation dataset will be resized to this"
            " resolution"
        ),
    )
    parser.add_argument(
        "--train_batch_size", type=int, default=4, help="Batch size (per device) for the training dataloader."
    )
    parser.add_argument("--num_train_epochs", type=int, default=1)
    parser.add_argument(
        "--max_train_steps",
        type=int,
        default=None,
        help="Total number of training steps to perform.  If provided, overrides num_train_epochs.",
    )
    parser.add_argument(
        "--checkpointing_steps",
        type=int,
        default=500,
        help=(
            "Save a checkpoint of the training state every X updates. These checkpoints can be used both as final"
            " checkpoints in case they are better than the last checkpoint, and are also suitable for resuming"
            " training using `--resume_from_checkpoint`."
        ),
    )
    parser.add_argument(
        "--checkpoints_total_limit",
        type=int,
        default=None,
        help=("Max number of checkpoints to store."),
    )
    parser.add_argument(
        "--resume_from_checkpoint",
        type=str,
        default=None,
        help=(
            "Whether training should be resumed from a previous checkpoint. Use a path saved by"
            ' `--checkpointing_steps`, or `"latest"` to automatically select the last available checkpoint.'
        ),
    )
    parser.add_argument(
        "--gradient_accumulation_steps",
        type=int,
        default=1,
        help="Number of updates steps to accumulate before performing a backward/update pass.",
    )
    parser.add_argument(
        "--gradient_checkpointing",
        action="store_true",
        help="Whether or not to use gradient checkpointing to save memory at the expense of slower backward pass.",
    )
    parser.add_argument(
        "--unet_learning_rate",
        type=float,
        default=2e-4,
        help="Learning rate to use for unet.",
    )
    parser.add_argument(
        "--text_encoder_learning_rate",
        type=float,
        default=4e-5,
        help="Learning rate to use for text encoder.",
    )
    parser.add_argument(
        "--scale_lr",
        action="store_true",
        default=False,
        help="Scale the learning rate by the number of GPUs, gradient accumulation steps, and batch size.",
    )
    parser.add_argument(
        "--lr_scheduler",
        type=str,
        default="constant",
        help=(
            'The scheduler type to use. Choose between ["linear", "cosine", "cosine_with_restarts", "polynomial",'
            ' "constant", "constant_with_warmup"]'
        ),
    )
    parser.add_argument(
        "--lr_warmup_steps", type=int, default=500, help="Number of steps for the warmup in the lr scheduler."
    )
    parser.add_argument(
        "--lr_num_cycles",
        type=int,
        default=1,
        help="Number of hard resets of the lr in cosine_with_restarts scheduler.",
    )
    parser.add_argument("--lr_power", type=float, default=1.0, help="Power factor of the polynomial scheduler.")
    parser.add_argument(
        "--use_8bit_adam", action="store_true", help="Whether or not to use 8-bit Adam from bitsandbytes."
    )
    parser.add_argument("--adam_beta1", type=float, default=0.9, help="The beta1 parameter for the Adam optimizer.")
    parser.add_argument("--adam_beta2", type=float, default=0.999, help="The beta2 parameter for the Adam optimizer.")
    parser.add_argument("--adam_weight_decay", type=float, default=1e-2, help="Weight decay to use.")
    parser.add_argument("--adam_epsilon", type=float, default=1e-08, help="Epsilon value for the Adam optimizer")
    parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.")
    parser.add_argument("--push_to_hub", action="store_true", help="Whether or not to push the model to the Hub.")
    parser.add_argument("--hub_token", type=str, default=None, help="The token to use to push to the Model Hub.")
    parser.add_argument(
        "--hub_model_id",
        type=str,
        default=None,
        help="The name of the repository to keep in sync with the local `output_dir`.",
    )
    parser.add_argument(
        "--logging_dir",
        type=str,
        default="logs",
        help=(
            "[TensorBoard](https://www.tensorflow.org/tensorboard) log directory. Will default to"
            " *output_dir/runs/**CURRENT_DATETIME_HOSTNAME***."
        ),
    )
    parser.add_argument(
        "--allow_tf32",
        action="store_true",
        help=(
            "Whether or not to allow TF32 on Ampere GPUs. Can be used to speed up training. For more information, see"
            " https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices"
        ),
    )
    parser.add_argument(
        "--report_to",
        type=str,
        default="tensorboard",
        help=(
            'The integration to report the results and logs to. Supported platforms are `"tensorboard"`'
            ' (default), `"wandb"` and `"comet_ml"`. Use `"all"` to report to all integrations.'
        ),
    )
    parser.add_argument(
        "--wandb_key",
        type=str,
        default=None,
        help=("If report to option is set to wandb, api-key for wandb used for login to wandb "),
    )
    parser.add_argument(
        "--wandb_project_name",
        type=str,
        default=None,
        help=("If report to option is set to wandb, project name in wandb for log tracking  "),
    )
    parser.add_argument(
        "--mixed_precision",
        type=str,
        default=None,
        choices=["no", "fp16", "bf16"],
        help=(
            "Whether to use mixed precision. Choose between fp16 and bf16 (bfloat16). Bf16 requires PyTorch >="
            " 1.10.and an Nvidia Ampere GPU.  Default to the value of accelerate config of the current system or the"
            " flag passed with the `accelerate.launch` command. Use this argument to override the accelerate config."
        ),
    )
    parser.add_argument("--local_rank", type=int, default=-1, help="For distributed training: local_rank")
    parser.add_argument(
        "--enable_xformers_memory_efficient_attention", action="store_true", help="Whether or not to use xformers."
    )
    parser.add_argument(
        "--set_grads_to_none",
        action="store_true",
        help=(
            "Save more memory by using setting grads to None instead of zero. Be aware, that this changes certain"
            " behaviors, so disable this argument if it causes any problems. More info:"
            " https://pytorch.org/docs/stable/generated/torch.optim.Optimizer.zero_grad.html"
        ),
    )
    parser.add_argument(
        "--lora_rank",
        type=int,
        default=16,
        help=("The dimension of the LoRA update matrices."),
    )
    parser.add_argument(
        "--lora_alpha",
        type=int,
        default=27,
        help=("The alpha constant of the LoRA update matrices."),
    )
    parser.add_argument(
        "--lora_dropout",
        type=float,
        default=0.0,
        help="The dropout rate of the LoRA update matrices.",
    )
    parser.add_argument(
        "--lora_bias",
        type=str,
        default="none",
        help="The bias type of the Lora update matrices. Must be 'none', 'all' or 'lora_only'.",
    )

    if input_args is not None:
        args = parser.parse_args(input_args)
    else:
        args = parser.parse_args()

    env_local_rank = int(os.environ.get("LOCAL_RANK", -1))
    if env_local_rank != -1 and env_local_rank != args.local_rank:
        args.local_rank = env_local_rank

    return args


class RealFillDataset(Dataset):
    """
    A dataset to prepare the training and conditioning images and
    the masks with the dummy prompt for fine-tuning the model.
    It pre-processes the images, masks and tokenizes the prompts.
    """

    def __init__(
        self,
        train_data_root,
        tokenizer,
        size=512,
    ):
        self.size = size
        self.tokenizer = tokenizer

        self.ref_data_root = Path(train_data_root) / "ref"
        self.target_image = Path(train_data_root) / "target" / "target.png"
        self.target_mask = Path(train_data_root) / "target" / "mask.png"
        if not (self.ref_data_root.exists() and self.target_image.exists() and self.target_mask.exists()):
            raise ValueError("Train images root doesn't exists.")

        self.train_images_path = list(self.ref_data_root.iterdir()) + [self.target_image]
        self.num_train_images = len(self.train_images_path)
        self.train_prompt = "a photo of sks"

        self.transform = transforms_v2.Compose(
            [
                transforms_v2.ToImage(),
                transforms_v2.RandomResize(size, int(1.125 * size)),
                transforms_v2.RandomCrop(size),
                transforms_v2.ToDtype(torch.float32, scale=True),
                transforms_v2.Normalize([0.5], [0.5]),
            ]
        )

    def __len__(self):
        return self.num_train_images

    def __getitem__(self, index):
        example = {}

        image = Image.open(self.train_images_path[index])
        image = exif_transpose(image)

        if not image.mode == "RGB":
            image = image.convert("RGB")

        if index < len(self) - 1:
            weighting = Image.new("L", image.size)
        else:
            weighting = Image.open(self.target_mask)
            weighting = exif_transpose(weighting)

        image, weighting = self.transform(image, weighting)
        example["images"], example["weightings"] = image, weighting < 0

        if random.random() < 0.1:
            example["masks"] = torch.ones_like(example["images"][0:1, :, :])
        else:
            example["masks"] = make_mask(example["images"], self.size)

        example["conditioning_images"] = example["images"] * (example["masks"] < 0.5)

        train_prompt = "" if random.random() < 0.1 else self.train_prompt
        example["prompt_ids"] = self.tokenizer(
            train_prompt,
            truncation=True,
            padding="max_length",
            max_length=self.tokenizer.model_max_length,
            return_tensors="pt",
        ).input_ids

        return example


def collate_fn(examples):
    input_ids = [example["prompt_ids"] for example in examples]
    images = [example["images"] for example in examples]

    masks = [example["masks"] for example in examples]
    weightings = [example["weightings"] for example in examples]
    conditioning_images = [example["conditioning_images"] for example in examples]

    images = torch.stack(images)
    images = images.to(memory_format=torch.contiguous_format).float()

    masks = torch.stack(masks)
    masks = masks.to(memory_format=torch.contiguous_format).float()

    weightings = torch.stack(weightings)
    weightings = weightings.to(memory_format=torch.contiguous_format).float()

    conditioning_images = torch.stack(conditioning_images)
    conditioning_images = conditioning_images.to(memory_format=torch.contiguous_format).float()

    input_ids = torch.cat(input_ids, dim=0)

    batch = {
        "input_ids": input_ids,
        "images": images,
        "masks": masks,
        "weightings": weightings,
        "conditioning_images": conditioning_images,
    }
    return batch


def main(args):
    if args.report_to == "wandb" and args.hub_token is not None:
        raise ValueError(
            "You cannot use both --report_to=wandb and --hub_token due to a security risk of exposing your token."
            " Please use `huggingface-cli login` to authenticate with the Hub."
        )

    logging_dir = Path(args.output_dir, args.logging_dir)

    accelerator = Accelerator(
        gradient_accumulation_steps=args.gradient_accumulation_steps,
        mixed_precision=args.mixed_precision,
        log_with=args.report_to,
        project_dir=logging_dir,
    )

    if args.report_to == "wandb":
        if not is_wandb_available():
            raise ImportError("Make sure to install wandb if you want to use it for logging during training.")

        wandb.login(key=args.wandb_key)
        wandb.init(project=args.wandb_project_name)

    # Make one log on every process with the configuration for debugging.
    logging.basicConfig(
        format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
        datefmt="%m/%d/%Y %H:%M:%S",
        level=logging.INFO,
    )
    logger.info(accelerator.state, main_process_only=False)
    if accelerator.is_local_main_process:
        transformers.utils.logging.set_verbosity_warning()
        diffusers.utils.logging.set_verbosity_info()
    else:
        transformers.utils.logging.set_verbosity_error()
        diffusers.utils.logging.set_verbosity_error()

    # If passed along, set the training seed now.
    if args.seed is not None:
        set_seed(args.seed)

    # Handle the repository creation
    if accelerator.is_main_process:
        if args.output_dir is not None:
            os.makedirs(args.output_dir, exist_ok=True)

        if args.push_to_hub:
            repo_id = create_repo(
                repo_id=args.hub_model_id or Path(args.output_dir).name, exist_ok=True, token=args.hub_token
            ).repo_id

    # Load the tokenizer
    if args.tokenizer_name:
        tokenizer = AutoTokenizer.from_pretrained(args.tokenizer_name, revision=args.revision, use_fast=False)
    elif args.pretrained_model_name_or_path:
        tokenizer = AutoTokenizer.from_pretrained(
            args.pretrained_model_name_or_path,
            subfolder="tokenizer",
            revision=args.revision,
            use_fast=False,
        )

    # Load scheduler and models
    noise_scheduler = DDPMScheduler.from_pretrained(args.pretrained_model_name_or_path, subfolder="scheduler")
    text_encoder = CLIPTextModel.from_pretrained(
        args.pretrained_model_name_or_path, subfolder="text_encoder", revision=args.revision
    )
    vae = AutoencoderKL.from_pretrained(args.pretrained_model_name_or_path, subfolder="vae", revision=args.revision)
    unet = UNet2DConditionModel.from_pretrained(
        args.pretrained_model_name_or_path, subfolder="unet", revision=args.revision
    )

    config = LoraConfig(
        r=args.lora_rank,
        lora_alpha=args.lora_alpha,
        target_modules=["to_k", "to_q", "to_v", "key", "query", "value"],
        lora_dropout=args.lora_dropout,
        bias=args.lora_bias,
    )
    unet = get_peft_model(unet, config)

    config = LoraConfig(
        r=args.lora_rank,
        lora_alpha=args.lora_alpha,
        target_modules=["k_proj", "q_proj", "v_proj"],
        lora_dropout=args.lora_dropout,
        bias=args.lora_bias,
    )
    text_encoder = get_peft_model(text_encoder, config)

    vae.requires_grad_(False)

    if args.enable_xformers_memory_efficient_attention:
        if is_xformers_available():
            import xformers

            xformers_version = version.parse(xformers.__version__)
            if xformers_version == version.parse("0.0.16"):
                logger.warning(
                    "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                )
            unet.enable_xformers_memory_efficient_attention()
        else:
            raise ValueError("xformers is not available. Make sure it is installed correctly")

    if args.gradient_checkpointing:
        unet.enable_gradient_checkpointing()
        text_encoder.gradient_checkpointing_enable()

    # create custom saving & loading hooks so that `accelerator.save_state(...)` serializes in a nice format
    def save_model_hook(models, weights, output_dir):
        if accelerator.is_main_process:
            for model in models:
                sub_dir = (
                    "unet"
                    if isinstance(model.base_model.model, type(accelerator.unwrap_model(unet).base_model.model))
                    else "text_encoder"
                )
                model.save_pretrained(os.path.join(output_dir, sub_dir))

                # make sure to pop weight so that corresponding model is not saved again
                weights.pop()

    def load_model_hook(models, input_dir):
        while len(models) > 0:
            # pop models so that they are not loaded again
            model = models.pop()

            sub_dir = (
                "unet"
                if isinstance(model.base_model.model, type(accelerator.unwrap_model(unet).base_model.model))
                else "text_encoder"
            )
            model_cls = (
                UNet2DConditionModel
                if isinstance(model.base_model.model, type(accelerator.unwrap_model(unet).base_model.model))
                else CLIPTextModel
            )

            load_model = model_cls.from_pretrained(args.pretrained_model_name_or_path, subfolder=sub_dir)
            load_model = PeftModel.from_pretrained(load_model, input_dir, subfolder=sub_dir)

            model.load_state_dict(load_model.state_dict())
            del load_model

    accelerator.register_save_state_pre_hook(save_model_hook)
    accelerator.register_load_state_pre_hook(load_model_hook)

    # Enable TF32 for faster training on Ampere GPUs,
    # cf https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices
    if args.allow_tf32:
        torch.backends.cuda.matmul.allow_tf32 = True

    if args.scale_lr:
        args.unet_learning_rate = (
            args.unet_learning_rate
            * args.gradient_accumulation_steps
            * args.train_batch_size
            * accelerator.num_processes
        )

        args.text_encoder_learning_rate = (
            args.text_encoder_learning_rate
            * args.gradient_accumulation_steps
            * args.train_batch_size
            * accelerator.num_processes
        )

    # Use 8-bit Adam for lower memory usage or to fine-tune the model in 16GB GPUs
    if args.use_8bit_adam:
        try:
            import bitsandbytes as bnb
        except ImportError:
            raise ImportError(
                "To use 8-bit Adam, please install the bitsandbytes library: `pip install bitsandbytes`."
            )

        optimizer_class = bnb.optim.AdamW8bit
    else:
        optimizer_class = torch.optim.AdamW

    # Optimizer creation
    optimizer = optimizer_class(
        [
            {"params": unet.parameters(), "lr": args.unet_learning_rate},
            {"params": text_encoder.parameters(), "lr": args.text_encoder_learning_rate},
        ],
        betas=(args.adam_beta1, args.adam_beta2),
        weight_decay=args.adam_weight_decay,
        eps=args.adam_epsilon,
    )

    # Dataset and DataLoaders creation:
    train_dataset = RealFillDataset(
        train_data_root=args.train_data_dir,
        tokenizer=tokenizer,
        size=args.resolution,
    )

    train_dataloader = torch.utils.data.DataLoader(
        train_dataset,
        batch_size=args.train_batch_size,
        shuffle=True,
        collate_fn=collate_fn,
        num_workers=1,
    )

    # Scheduler and math around the number of training steps.
    overrode_max_train_steps = False
    num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
    if args.max_train_steps is None:
        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
        overrode_max_train_steps = True

    lr_scheduler = get_scheduler(
        args.lr_scheduler,
        optimizer=optimizer,
        num_warmup_steps=args.lr_warmup_steps * args.gradient_accumulation_steps,
        num_training_steps=args.max_train_steps * args.gradient_accumulation_steps,
        num_cycles=args.lr_num_cycles,
        power=args.lr_power,
    )

    # Prepare everything with our `accelerator`.
    unet, text_encoder, optimizer, train_dataloader = accelerator.prepare(
        unet, text_encoder, optimizer, train_dataloader
    )

    # For mixed precision training we cast all non-trainable weigths (vae, non-lora text_encoder and non-lora unet) to half-precision
    # as these weights are only used for inference, keeping weights in full precision is not required.
    weight_dtype = torch.float32
    if accelerator.mixed_precision == "fp16":
        weight_dtype = torch.float16
    elif accelerator.mixed_precision == "bf16":
        weight_dtype = torch.bfloat16

    # Move vae to device and cast to weight_dtype
    vae.to(accelerator.device, dtype=weight_dtype)

    # We need to recalculate our total training steps as the size of the training dataloader may have changed.
    num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
    if overrode_max_train_steps:
        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
    # Afterwards we recalculate our number of training epochs
    args.num_train_epochs = math.ceil(args.max_train_steps / num_update_steps_per_epoch)

    # We need to initialize the trackers we use, and also store our configuration.
    # The trackers initializes automatically on the main process.
    if accelerator.is_main_process:
        tracker_config = vars(copy.deepcopy(args))
        accelerator.init_trackers("realfill", config=tracker_config)

    # Train!
    total_batch_size = args.train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps

    logger.info("***** Running training *****")
    logger.info(f"  Num examples = {len(train_dataset)}")
    logger.info(f"  Num batches each epoch = {len(train_dataloader)}")
    logger.info(f"  Num Epochs = {args.num_train_epochs}")
    logger.info(f"  Instantaneous batch size per device = {args.train_batch_size}")
    logger.info(f"  Total train batch size (w. parallel, distributed & accumulation) = {total_batch_size}")
    logger.info(f"  Gradient Accumulation steps = {args.gradient_accumulation_steps}")
    logger.info(f"  Total optimization steps = {args.max_train_steps}")
    global_step = 0
    first_epoch = 0

    # Potentially load in the weights and states from a previous save
    if args.resume_from_checkpoint:
        if args.resume_from_checkpoint != "latest":
            path = os.path.basename(args.resume_from_checkpoint)
        else:
            # Get the mos recent checkpoint
            dirs = os.listdir(args.output_dir)
            dirs = [d for d in dirs if d.startswith("checkpoint")]
            dirs = sorted(dirs, key=lambda x: int(x.split("-")[1]))
            path = dirs[-1] if len(dirs) > 0 else None

        if path is None:
            accelerator.print(
                f"Checkpoint '{args.resume_from_checkpoint}' does not exist. Starting a new training run."
            )
            args.resume_from_checkpoint = None
            initial_global_step = 0
        else:
            accelerator.print(f"Resuming from checkpoint {path}")
            accelerator.load_state(os.path.join(args.output_dir, path))
            global_step = int(path.split("-")[1])

            initial_global_step = global_step
            first_epoch = global_step // num_update_steps_per_epoch
    else:
        initial_global_step = 0

    progress_bar = tqdm(
        range(0, args.max_train_steps),
        initial=initial_global_step,
        desc="Steps",
        # Only show the progress bar once on each machine.
        disable=not accelerator.is_local_main_process,
    )

    for epoch in range(first_epoch, args.num_train_epochs):
        unet.train()
        text_encoder.train()

        for step, batch in enumerate(train_dataloader):
            with accelerator.accumulate(unet, text_encoder):
                # Convert images to latent space
                latents = vae.encode(batch["images"].to(dtype=weight_dtype)).latent_dist.sample()
                latents = latents * 0.18215

                # Convert masked images to latent space
                conditionings = vae.encode(batch["conditioning_images"].to(dtype=weight_dtype)).latent_dist.sample()
                conditionings = conditionings * 0.18215

                # Downsample mask and weighting so that they match with the latents
                masks, size = batch["masks"].to(dtype=weight_dtype), latents.shape[2:]
                masks = F.interpolate(masks, size=size)

                weightings = batch["weightings"].to(dtype=weight_dtype)
                weightings = F.interpolate(weightings, size=size)

                # Sample noise that we'll add to the latents
                noise = torch.randn_like(latents)
                bsz = latents.shape[0]

                # Sample a random timestep for each image
                timesteps = torch.randint(0, noise_scheduler.config.num_train_timesteps, (bsz,), device=latents.device)
                timesteps = timesteps.long()

                # Add noise to the latents according to the noise magnitude at each timestep
                # (this is the forward diffusion process)
                noisy_latents = noise_scheduler.add_noise(latents, noise, timesteps)

                # Concatenate noisy latents, masks and conditionings to get inputs to unet
                inputs = torch.cat([noisy_latents, masks, conditionings], dim=1)

                # Get the text embedding for conditioning
                encoder_hidden_states = text_encoder(batch["input_ids"])[0]

                # Predict the noise residual
                model_pred = unet(inputs, timesteps, encoder_hidden_states).sample

                # Compute the diffusion loss
                assert noise_scheduler.config.prediction_type == "epsilon"
                loss = (weightings * F.mse_loss(model_pred.float(), noise.float(), reduction="none")).mean()

                # Backpropagate
                accelerator.backward(loss)
                if accelerator.sync_gradients:
                    params_to_clip = itertools.chain(unet.parameters(), text_encoder.parameters())
                    accelerator.clip_grad_norm_(params_to_clip, args.max_grad_norm)

                optimizer.step()
                lr_scheduler.step()
                optimizer.zero_grad(set_to_none=args.set_grads_to_none)

            # Checks if the accelerator has performed an optimization step behind the scenes
            if accelerator.sync_gradients:
                progress_bar.update(1)
                if args.report_to == "wandb":
                    accelerator.print(progress_bar)
                global_step += 1

                if accelerator.is_main_process:
                    if global_step % args.checkpointing_steps == 0:
                        # _before_ saving state, check if this save would set us over the `checkpoints_total_limit`
                        if args.checkpoints_total_limit is not None:
                            checkpoints = os.listdir(args.output_dir)
                            checkpoints = [d for d in checkpoints if d.startswith("checkpoint")]
                            checkpoints = sorted(checkpoints, key=lambda x: int(x.split("-")[1]))

                            # before we save the new checkpoint, we need to have at _most_ `checkpoints_total_limit - 1` checkpoints
                            if len(checkpoints) >= args.checkpoints_total_limit:
                                num_to_remove = len(checkpoints) - args.checkpoints_total_limit + 1
                                removing_checkpoints = checkpoints[0:num_to_remove]

                                logger.info(
                                    f"{len(checkpoints)} checkpoints already exist, removing {len(removing_checkpoints)} checkpoints"
                                )
                                logger.info(f"removing checkpoints: {', '.join(removing_checkpoints)}")

                                for removing_checkpoint in removing_checkpoints:
                                    removing_checkpoint = os.path.join(args.output_dir, removing_checkpoint)
                                    shutil.rmtree(removing_checkpoint)

                        save_path = os.path.join(args.output_dir, f"checkpoint-{global_step}")
                        accelerator.save_state(save_path)
                        logger.info(f"Saved state to {save_path}")

                    if global_step % args.validation_steps == 0:
                        log_validation(
                            text_encoder,
                            tokenizer,
                            unet,
                            args,
                            accelerator,
                            weight_dtype,
                            global_step,
                        )

            logs = {"loss": loss.detach().item()}
            progress_bar.set_postfix(**logs)
            accelerator.log(logs, step=global_step)

            if global_step >= args.max_train_steps:
                break

    # Save the lora layers
    accelerator.wait_for_everyone()
    if accelerator.is_main_process:
        pipeline = StableDiffusionInpaintPipeline.from_pretrained(
            args.pretrained_model_name_or_path,
            unet=accelerator.unwrap_model(unet, keep_fp32_wrapper=True).merge_and_unload(),
            text_encoder=accelerator.unwrap_model(text_encoder, keep_fp32_wrapper=True).merge_and_unload(),
            revision=args.revision,
        )

        pipeline.save_pretrained(args.output_dir)

        # Final inference
        images = log_validation(
            text_encoder,
            tokenizer,
            unet,
            args,
            accelerator,
            weight_dtype,
            global_step,
        )

        if args.push_to_hub:
            save_model_card(
                repo_id,
                images=images,
                base_model=args.pretrained_model_name_or_path,
                repo_folder=args.output_dir,
            )
            upload_folder(
                repo_id=repo_id,
                folder_path=args.output_dir,
                commit_message="End of training",
                ignore_patterns=["step_*", "epoch_*"],
            )

    accelerator.end_training()


if __name__ == "__main__":
    args = parse_args()
    main(args)