concept-ablation-diffusers/train.py

# This code is modified from the Huggingface repository: https://github.com/huggingface/diffusers/blob/main/examples/dreambooth/train_dreambooth_lora.py, and
import argparse
import hashlib
import itertools
import json
import logging
import math
import os
import warnings
from pathlib import Path

import numpy as np
import torch
import torch.nn.functional as F
import torch.utils.checkpoint
import transformers
from accelerate import Accelerator
from accelerate.logging import get_logger
from accelerate.utils import ProjectConfiguration, set_seed
from huggingface_hub import HfApi, create_repo
from model_pipeline import (
    CustomDiffusionAttnProcessor,
    CustomDiffusionPipeline,
    set_use_memory_efficient_attention_xformers,
)
from packaging import version
from tqdm.auto import tqdm
from transformers import AutoTokenizer, PretrainedConfig
from utils import (
    CustomDiffusionDataset,
    PromptDataset,
    collate_fn,
    filter,
    getanchorprompts,
)

import diffusers
from diffusers import (
    AutoencoderKL,
    DDPMScheduler,
    DiffusionPipeline,
    DPMSolverMultistepScheduler,
    UNet2DConditionModel,
)
from diffusers.models.cross_attention import CrossAttention
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

# Will error if the minimal version of diffusers is not installed. Remove at your own risks.
check_min_version("0.14.0")

logger = get_logger(__name__)


def create_custom_diffusion(unet, parameter_group):
    for name, params in unet.named_parameters():
        if parameter_group == "cross-attn":
            if 'attn2.to_k' in name or 'attn2.to_v' in name:
                params.requires_grad = True
            else:
                params.requires_grad = False
        elif parameter_group == 'full-weight':
            params.requires_grad = True
        elif parameter_group == 'embedding':
            params.requires_grad = False
        else:
            raise ValueError(
                "parameter_group argument only cross-attn, full-weight, embedding"
            )

    # change attn class
    def change_attn(unet):
        for layer in unet.children():
            if type(layer) == CrossAttention:
                bound_method = set_use_memory_efficient_attention_xformers.__get__(
                    layer, layer.__class__)
                setattr(
                    layer, 'set_use_memory_efficient_attention_xformers', bound_method)
            else:
                change_attn(layer)

    change_attn(unet)
    unet.set_attn_processor(CustomDiffusionAttnProcessor())
    return unet


def save_model_card(repo_id: str, images=None, base_model=str, prompt=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"./image_{i}.png\n"

    yaml = f"""
        ---
        license: creativeml-openrail-m
        base_model: {base_model}
        instance_prompt: {prompt}
        tags:
        - stable-diffusion
        - stable-diffusion-diffusers
        - text-to-image
        - diffusers
        - custom diffusion
        inference: true
        ---
            """
    model_card = f"""
        # Custom Diffusion - {repo_id}

        These are Custom Diffusion adaption weights for {base_model}. The weights were trained on {prompt} using [Custom Diffusion](https://www.cs.cmu.edu/~custom-diffusion). You can find some example images in the following. \n
        {img_str[0]}
        """
    with open(os.path.join(repo_folder, "README.md"), "w") as f:
        f.write(yaml + model_card)


def import_model_class_from_model_name_or_path(pretrained_model_name_or_path: str, revision: str):
    text_encoder_config = PretrainedConfig.from_pretrained(
        pretrained_model_name_or_path,
        subfolder="text_encoder",
        revision=revision,
    )
    model_class = text_encoder_config.architectures[0]

    if model_class == "CLIPTextModel":
        from transformers import CLIPTextModel

        return CLIPTextModel
    elif model_class == "RobertaSeriesModelWithTransformation":
        from diffusers.pipelines.alt_diffusion.modeling_roberta_series import (
            RobertaSeriesModelWithTransformation,
        )

        return RobertaSeriesModelWithTransformation
    else:
        raise ValueError(f"{model_class} is not supported.")


def freeze_params(params):
    for param in params:
        param.requires_grad = False


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(
        "--concept_type",
        type=str,
        required=True,
        choices=['style', 'object', 'memorization'],
        help='the type of removed concepts'
    )
    parser.add_argument(
        "--caption_target",
        type=str,
        required=True,
        help="target style to remove, used when kldiv loss",
    )
    parser.add_argument(
        "--instance_data_dir",
        type=str,
        default=None,
        help="A folder containing the training data of instance images.",
    )
    parser.add_argument(
        "--class_data_dir",
        type=str,
        default=None,
        help="A folder containing the training data of class images.",
    )
    parser.add_argument(
        "--instance_prompt",
        type=str,
        help="The prompt with identifier specifying the instance",
    )
    parser.add_argument(
        "--class_prompt",
        type=str,
        default=None,
        help="The prompt to specify images in the same class as provided instance images.",
    )
    parser.add_argument(
        "--mem_impath",
        type=str,
        default="",
        help='the path to saved memorized image. Required when concept_type is memorization'
    )
    parser.add_argument(
        "--validation_prompt",
        type=str,
        default=None,
        help="A prompt that is used during validation to verify that the model is learning.",
    )
    parser.add_argument(
        "--num_validation_images",
        type=int,
        default=2,
        help="Number of images that should be generated during validation with `validation_prompt`.",
    )
    parser.add_argument(
        "--validation_steps",
        type=int,
        default=500,
        help=(
            "Run dreambooth validation every X epochs. Dreambooth validation consists of running the prompt"
            " `args.validation_prompt` multiple times: `args.num_validation_images`."
        ),
    )
    parser.add_argument(
        "--with_prior_preservation",
        default=False,
        action="store_true",
        help="Flag to add prior preservation loss.",
    )
    parser.add_argument("--prior_loss_weight", type=float,
                        default=1.0, help="The weight of prior preservation loss.")
    parser.add_argument(
        "--train_size",
        type=int,
        default=1000,
        help='the number of generated images used for ablating the concept'
    )
    parser.add_argument(
        "--output_dir",
        type=str,
        default="custom-diffusion-model",
        help="The output directory where the model predictions and checkpoints will be written.",
    )
    parser.add_argument(
        "--num_class_images",
        type=int,
        default=1000,
        help=(
            "Minimal anchor class images. If there are not enough images already present in"
            " class_data_dir, additional images will be sampled with class_prompt."
        ),
    )
    parser.add_argument(
        "--num_class_prompts",
        type=int,
        default=200,
        help=(
            "Minimal prompts used to generate anchor class images"
        ),
    )
    parser.add_argument("--seed", type=int, default=42,
                        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(
        "--center_crop",
        default=False,
        action="store_true",
        help=(
            "Whether to center crop the input images to the resolution. If not set, the images will be randomly"
            " cropped. The images will be resized to the resolution first before cropping."
        ),
    )
    parser.add_argument(
        "--train_batch_size", type=int, default=4, help="Batch size (per device) for the training dataloader."
    )
    parser.add_argument(
        "--sample_batch_size", type=int, default=4, help="Batch size (per device) for sampling images."
    )
    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=250,
        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. Passed as `total_limit` to the `Accelerator` `ProjectConfiguration`."
            " See Accelerator::save_state https://huggingface.co/docs/accelerate/package_reference/accelerator#accelerate.Accelerator.save_state"
            " for more docs"
        ),
    )
    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(
        "--learning_rate",
        type=float,
        default=1e-5,
        help="Initial learning rate (after the potential warmup period) to use.",
    )
    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(
        "--dataloader_num_workers",
        type=int,
        default=2,
        help=(
            "Number of subprocesses to use for data loading. 0 means that the data will be loaded in the main process."
        ),
    )
    parser.add_argument(
        "--parameter_group",
        type=str,
        default='cross-attn',
        choices=['full-weight', 'cross-attn', 'embedding'],
        help='parameter groups to finetune. Default: full-weight for memorization and cross-attn for others'
    )
    parser.add_argument(
        "--loss_type_reverse",
        type=str,
        default='model-based',
        help="loss type for reverse fine-tuning",
    )
    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(
        "--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(
        "--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(
        "--prior_generation_precision",
        type=str,
        default=None,
        choices=["no", "fp32", "fp16", "bf16"],
        help=(
            "Choose prior generation precision between fp32, fp16 and bf16 (bfloat16). Bf16 requires PyTorch >="
            " 1.10.and an Nvidia Ampere GPU.  Default to  fp16 if a GPU is available else fp32."
        ),
    )
    parser.add_argument(
        "--concepts_list",
        type=str,
        default=None,
        help="Path to json containing multiple concepts, will overwrite parameters like instance_prompt, class_prompt, etc.",
    )
    parser.add_argument(
        "--openai_key",
        type=str,
        default="",
        help=(
            "OPENAI API key. required for ablating objects and memorized images."
        ),
    )
    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("--hflip", action="store_true",
                        help="Apply horizontal flip data augmentation.")
    parser.add_argument("--noaug", action="store_true",
                        help="Dont apply augmentation during data augmentation when this flag is enabled.")

    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

    if args.with_prior_preservation:
        if args.concepts_list is None:
            if args.class_data_dir is None:
                raise ValueError(
                    "You must specify a data directory for class images.")
            if args.class_prompt is None:
                raise ValueError("You must specify prompt for class images.")
    else:
        # logger is not available yet
        if args.class_data_dir is not None:
            warnings.warn(
                "You need not use --class_data_dir without --with_prior_preservation.")
        if args.class_prompt is not None:
            warnings.warn(
                "You need not use --class_prompt without --with_prior_preservation.")

    return args


def main(args):
    logging_dir = Path(args.output_dir, args.logging_dir)

    accelerator_project_config = ProjectConfiguration(
        total_limit=args.checkpoints_total_limit)

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

    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.")
        import wandb

    # Currently, it's not possible to do gradient accumulation when training two models with accelerate.accumulate
    # This will be enabled soon in accelerate. For now, we don't allow gradient accumulation when training two models.
    # TODO (patil-suraj): Remove this check when gradient accumulation with two models is enabled in accelerate.
    # 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()

    # 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:
        print(vars(args))
        accelerator.init_trackers("custom-diffusion", config=vars(args))

    # If passed along, set the training seed now.
    if args.seed is not None:
        set_seed(args.seed)
    if args.concepts_list is None:
        args.concepts_list = [
            {
                "instance_prompt": args.instance_prompt,
                "class_prompt": args.class_prompt,
                "instance_data_dir": args.instance_data_dir,
                "class_data_dir": args.class_data_dir,
                "caption_target": args.caption_target,
            }
        ]
    else:
        with open(args.concepts_list, "r") as f:
            args.concepts_list = json.load(f)

    # Generate class images if prior preservation is enabled.
    for i, concept in enumerate(args.concepts_list):
        # directly path to ablation images and its corresponding prompts is provided.
        if (concept['instance_prompt'] is not None and concept['instance_data_dir'] is not None):
            break

        class_images_dir = Path(concept['class_data_dir'])
        if not class_images_dir.exists():
            class_images_dir.mkdir(parents=True, exist_ok=True)
        os.makedirs(f'{class_images_dir}/images', exist_ok=True)

        # we need to generate training images
        if len(list(Path(os.path.join(class_images_dir, 'images')).iterdir())) < args.num_class_images:
            torch_dtype = torch.float16 if accelerator.device.type == "cuda" else torch.float32
            if args.prior_generation_precision == "fp32":
                torch_dtype = torch.float32
            elif args.prior_generation_precision == "fp16":
                torch_dtype = torch.float16
            elif args.prior_generation_precision == "bf16":
                torch_dtype = torch.bfloat16
            pipeline = DiffusionPipeline.from_pretrained(
                args.pretrained_model_name_or_path,
                torch_dtype=torch_dtype,
                safety_checker=None,
                revision=args.revision,
            )
            pipeline.scheduler = DPMSolverMultistepScheduler.from_config(
                                    pipeline.scheduler.config)

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

            # need to create prompts using class_prompt.
            if not os.path.isfile(concept['class_prompt']):
                # style based prompts are retrieved from laion dataset
                if args.concept_type == 'style':
                    with open(os.path.join(class_images_dir, 'painting.txt')) as f:
                        class_prompt_collection = [
                                x.strip() for x in f.readlines()]

                # LLM based prompt collection.
                else:
                    class_prompt = concept['class_prompt']
                    # in case of object query chatGPT to generate captions containing the anchor category
                    if args.concept_type == 'object':
                        class_prompt_collection, _ = getanchorprompts(
                            pipeline, accelerator, class_prompt, args.concept_type, class_images_dir, args.openai_key, args.num_class_prompts)
                        with open(class_images_dir / 'caption_anchor.txt', 'w') as f:
                            for prompt in class_prompt_collection:
                                f.write(prompt + '\n')
                    # in case of memorization query chatGPT to generate different captions that can be paraphrase of the origianl caption
                    elif args.concept_type == 'memorization':
                        class_prompt_collection, caption_target = getanchorprompts(
                            pipeline, accelerator, class_prompt, args.concept_type, class_images_dir, args.openai_key, args.num_class_prompts, mem_impath=args.mem_impath)
                        concept['caption_target'] += f';*+{caption_target}'
                        with open(class_images_dir / 'caption_target.txt', 'w') as f:
                            f.write(concept['caption_target'])
                        print(class_prompt_collection,
                              concept['caption_target'])
            # class_prompt is filepath to prompts.
            else:
                with open(concept['class_prompt']) as f:
                    class_prompt_collection = [
                        x.strip() for x in f.readlines()]

            num_new_images = args.num_class_images
            logger.info(
                f"Number of class images to sample: {num_new_images}.")

            sample_dataset = PromptDataset(
                class_prompt_collection, num_new_images)
            sample_dataloader = torch.utils.data.DataLoader(
                sample_dataset, batch_size=args.sample_batch_size)

            sample_dataloader = accelerator.prepare(sample_dataloader)

            if os.path.exists(f'{class_images_dir}/caption.txt'):
                os.remove(f'{class_images_dir}/caption.txt')
            if os.path.exists(f'{class_images_dir}/images.txt'):
                os.remove(f'{class_images_dir}/images.txt')

            for example in tqdm(
                sample_dataloader, desc="Generating class images", disable=not accelerator.is_local_main_process
            ):
                accelerator.wait_for_everyone()
                with open(f'{class_images_dir}/caption.txt', 'a') as f1, open(f'{class_images_dir}/images.txt', 'a') as f2:
                    images = pipeline(example["prompt"], num_inference_steps=25, guidance_scale=6., eta=1.).images

                    for i, image in enumerate(images):
                        hash_image = hashlib.sha1(
                            image.tobytes()).hexdigest()
                        image_filename = class_images_dir / \
                            f"images/{example['index'][i]}-{hash_image}.jpg"
                        image.save(image_filename)
                        f2.write(str(image_filename)+'\n')
                    f1.write('\n'.join(example["prompt"]) + '\n')
                    accelerator.wait_for_everyone()

            del pipeline

        if args.concept_type == 'memorization':
            filter(class_images_dir, args.mem_impath,
                    outpath=str(class_images_dir / 'filtered'))
            if os.path.exists(class_images_dir / 'caption_target.txt'):
                with open(class_images_dir / 'caption_target.txt', 'r') as f:
                    concept['caption_target'] = f.readlines()[0].strip()
            class_images_dir = class_images_dir / 'filtered'

        concept['class_prompt'] = os.path.join(
            class_images_dir, 'caption.txt')
        concept['class_data_dir'] = os.path.join(
            class_images_dir, 'images.txt')
        concept['instance_prompt'] = os.path.join(
            class_images_dir, 'caption.txt')
        concept['instance_data_dir'] = os.path.join(
            class_images_dir, 'images.txt')

        if torch.cuda.is_available():
            torch.cuda.empty_cache()

    # 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:
            print(args.hub_model_id or Path(args.output_dir).name)
            repo_id = create_repo(
                repo_id=args.hub_model_id or Path(args.output_dir).name, exist_ok=True, token=args.hub_token
            )
            print(repo_id)
            repo_id = args.hub_model_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,
        )

    # import correct text encoder class
    text_encoder_cls = import_model_class_from_model_name_or_path(
        args.pretrained_model_name_or_path, args.revision)

    # Load scheduler and models
    noise_scheduler = DDPMScheduler.from_pretrained(
        args.pretrained_model_name_or_path, subfolder="scheduler")
    text_encoder = text_encoder_cls.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
    )

    vae.requires_grad_(False)
    if args.parameter_group != 'embedding':
        text_encoder.requires_grad_(False)
    unet = create_custom_diffusion(unet, args.parameter_group)

    # For mixed precision training we cast the text_encoder and vae weights to half-precision
    # as these models 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 unet, vae and text_encoder to device and cast to weight_dtype
    if accelerator.mixed_precision != "fp16":
        unet.to(accelerator.device, dtype=weight_dtype)
        text_encoder.to(accelerator.device, dtype=weight_dtype)
    vae.to(accelerator.device, dtype=weight_dtype)

    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.warn(
                    "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()
        if args.parameter_group == 'embedding':
            text_encoder.gradient_checkpointing_enable()
    # 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.learning_rate = (
            args.learning_rate * args.gradient_accumulation_steps *
            args.train_batch_size * accelerator.num_processes
        )
        if args.with_prior_preservation:
            args.learning_rate = args.learning_rate * 2.

    # 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

    # Adding a modifier token which is optimized ####
    # Code taken from https://github.com/huggingface/diffusers/blob/main/examples/textual_inversion/textual_inversion.py
    modifier_token_id = []
    if args.parameter_group == 'embedding':
        assert args.concept_type != 'memorization', "embedding finetuning is not supported for memorization"

        for concept in args.concept_list:
            # Convert the caption_target to ids
            token_ids = tokenizer.encode(
                [concept['caption_target']], add_special_tokens=False)
            print(token_ids)
        # Check if initializer_token is a single token or a sequence of tokens
        modifier_token_id += token_ids

        # Freeze all parameters except for the token embeddings in text encoder
        params_to_freeze = itertools.chain(
            text_encoder.text_model.encoder.parameters(),
            text_encoder.text_model.final_layer_norm.parameters(),
            text_encoder.text_model.embeddings.position_embedding.parameters(),
        )
        freeze_params(params_to_freeze)
        params_to_optimize = itertools.chain(
            text_encoder.get_input_embeddings().parameters())
    else:
        if args.parameter_group == 'cross-attn':
            params_to_optimize = itertools.chain([x[1] for x in unet.named_parameters() if (
                'attn2.to_k' in x[0] or 'attn2.to_v' in x[0])])
        if args.parameter_group == 'full-weight':
            params_to_optimize = itertools.chain(unet.parameters())

    # Optimizer creation
    optimizer = optimizer_class(
        params_to_optimize,
        lr=args.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 = CustomDiffusionDataset(
        concepts_list=args.concepts_list,
        concept_type=args.concept_type,
        tokenizer=tokenizer,
        with_prior_preservation=args.with_prior_preservation,
        size=args.resolution,
        center_crop=args.center_crop,
        num_class_images=args.num_class_images,
        hflip=args.hflip, aug=not args.noaug,
    )

    train_dataloader = torch.utils.data.DataLoader(
        train_dataset,
        batch_size=args.train_batch_size,
        shuffle=True,
        collate_fn=lambda examples: collate_fn(
            examples, args.with_prior_preservation),
        num_workers=args.dataloader_num_workers,
    )

    # 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,
    )

    # Prepare everything with our `accelerator`.
    if args.parameter_group == 'embedding':
        text_encoder, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
            text_encoder, optimizer, train_dataloader, lr_scheduler
        )
    else:
        unet, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
            unet, optimizer, train_dataloader, lr_scheduler
        )

    # 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)

    # 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
        else:
            accelerator.print(f"Resuming from checkpoint {path}")
            accelerator.load_state(os.path.join(args.output_dir, path))
            global_step = int(path.split("-")[1])

            resume_global_step = global_step * args.gradient_accumulation_steps
            first_epoch = global_step // num_update_steps_per_epoch
            resume_step = resume_global_step % (
                num_update_steps_per_epoch * args.gradient_accumulation_steps)

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

    for epoch in range(first_epoch, args.num_train_epochs):
        if args.parameter_group == 'embedding':
            text_encoder.train()
        else:
            unet.train()
        for step, batch in enumerate(train_dataloader):
            # Skip steps until we reach the resumed step
            if args.resume_from_checkpoint and epoch == first_epoch and step < resume_step:
                if step % args.gradient_accumulation_steps == 0:
                    progress_bar.update(1)
                continue

            with accelerator.accumulate(unet) if args.parameter_group != 'embedding' else accelerator.accumulate(text_encoder):
                # Convert images to latent space
                latents = vae.encode(batch["pixel_values"].to(
                    dtype=weight_dtype)).latent_dist.sample()
                latents = latents * vae.config.scaling_factor

                # 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)

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

                # Predict the noise residual
                model_pred = unet(noisy_latents, timesteps,
                                  encoder_hidden_states).sample
                with torch.no_grad():
                    model_pred_anchor = unet(noisy_latents[:encoder_anchor_hidden_states.size(
                        0)], timesteps[:encoder_anchor_hidden_states.size(0)], encoder_anchor_hidden_states).sample

                # Get the target for loss depending on the prediction type
                if args.loss_type_reverse == 'model-based':
                    if args.with_prior_preservation:
                        target_prior = torch.chunk(noise, 2, dim=0)[1]
                    target = model_pred_anchor
                else:
                    if noise_scheduler.config.prediction_type == "epsilon":
                        target = noise
                    elif noise_scheduler.config.prediction_type == "v_prediction":
                        target = noise_scheduler.get_velocity(
                            latents, noise, timesteps)
                    else:
                        raise ValueError(
                            f"Unknown prediction type {noise_scheduler.config.prediction_type}")
                    if args.with_prior_preservation:
                        target, target_prior = torch.chunk(target, 2, dim=0)

                if args.with_prior_preservation:
                    # Chunk the noise and model_pred into two parts and compute the loss on each part separately.
                    model_pred, model_pred_prior = torch.chunk(
                        model_pred, 2, dim=0)
                    mask = torch.chunk(batch["mask"], 2, dim=0)[0]
                    # Compute instance loss
                    loss = F.mse_loss(model_pred.float(),
                                      target.float(), reduction="none")
                    loss = (
                        (loss * mask).sum([1, 2, 3]) / mask.sum([1, 2, 3])).mean()

                    # Compute prior loss
                    prior_loss = F.mse_loss(
                        model_pred_prior.float(), target_prior.float(), reduction="mean")

                    # Add the prior loss to the instance loss.
                    loss = loss + args.prior_loss_weight * prior_loss
                else:
                    mask = batch["mask"]
                    loss = F.mse_loss(model_pred.float(),
                                      target.float(), reduction="none")
                    loss = (
                        (loss * mask).sum([1, 2, 3]) / mask.sum([1, 2, 3])).mean()

                accelerator.backward(loss)
                # Zero out the gradients for all token embeddings except the newly added
                # embeddings for the concept, as we only want to optimize the concept embeddings
                if args.parameter_group == 'embedding':
                    if accelerator.num_processes > 1:
                        grads_text_encoder = text_encoder.module.get_input_embeddings().weight.grad
                    else:
                        grads_text_encoder = text_encoder.get_input_embeddings().weight.grad
                    # Get the index for tokens that we want to zero the grads for
                    index_grads_to_zero = torch.arange(
                        len(tokenizer)) != modifier_token_id[0]
                    for i in range(len(modifier_token_id[1:])):
                        index_grads_to_zero = index_grads_to_zero & (
                            torch.arange(len(tokenizer)) != modifier_token_id[i])
                    grads_text_encoder.data[index_grads_to_zero,
                                            :] = grads_text_encoder.data[index_grads_to_zero, :].fill_(0)

                if accelerator.sync_gradients:
                    params_to_clip = (
                        itertools.chain(text_encoder.parameters())
                        if args.parameter_group == 'embedding'
                        else itertools.chain([x[1] for x in unet.named_parameters() if ('attn2' in x[0])])
                        if args.parameter_group == 'cross-attn'
                        else itertools.chain(unet.parameters())
                    )
                    accelerator.clip_grad_norm_(
                        params_to_clip, args.max_grad_norm)
                optimizer.step()
                lr_scheduler.step()
                optimizer.zero_grad()

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

                if global_step % args.checkpointing_steps == 0:
                    if accelerator.is_main_process:
                        pipeline = CustomDiffusionPipeline.from_pretrained(
                            args.pretrained_model_name_or_path,
                            unet=accelerator.unwrap_model(unet),
                            text_encoder=accelerator.unwrap_model(
                                text_encoder),
                            tokenizer=tokenizer,
                            revision=args.revision,
                            modifier_token_id=modifier_token_id,
                        )
                        save_path = os.path.join(
                            args.output_dir, f"delta-{global_step}")
                        pipeline.save_pretrained(
                            save_path, parameter_group=args.parameter_group)
                        logger.info(f"Saved state to {save_path}")

            logs = {"loss": loss.detach().item(
            ), "lr": lr_scheduler.get_last_lr()[0]}
            progress_bar.set_postfix(**logs)
            accelerator.log(logs, step=global_step)

            if global_step >= args.max_train_steps:
                break

        if accelerator.is_main_process:
            if args.validation_prompt is not None and global_step % args.validation_steps == 0:
                logger.info(
                    f"Running validation... \n Generating {args.num_validation_images} images with prompt:"
                    f" {args.validation_prompt}."
                )
                # create pipeline
                pipeline = CustomDiffusionPipeline.from_pretrained(
                    args.pretrained_model_name_or_path,
                    unet=accelerator.unwrap_model(unet),
                    text_encoder=accelerator.unwrap_model(text_encoder),
                    tokenizer=tokenizer,
                    revision=args.revision,
                    modifier_token_id=modifier_token_id,
                )
                pipeline.scheduler = DPMSolverMultistepScheduler.from_config(
                    pipeline.scheduler.config)
                pipeline = pipeline.to(accelerator.device)
                pipeline.set_progress_bar_config(disable=True)

                # run inference
                generator = torch.Generator(
                    device=accelerator.device).manual_seed(args.seed)
                images = [
                    pipeline(args.validation_prompt, num_inference_steps=25,
                             generator=generator, eta=1.).images[0]
                    for _ in range(args.num_validation_images)
                ]

                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=f"{i}: {args.validation_prompt}")
                                    for i, image in enumerate(images)
                                ]
                            }
                        )

                del pipeline
                torch.cuda.empty_cache()

    accelerator.wait_for_everyone()
    if accelerator.is_main_process:
        unet = unet.to(torch.float32)
        pipeline = CustomDiffusionPipeline.from_pretrained(
            args.pretrained_model_name_or_path,
            unet=accelerator.unwrap_model(unet),
            text_encoder=accelerator.unwrap_model(text_encoder),
            tokenizer=tokenizer,
            revision=args.revision,
            modifier_token_id=modifier_token_id,
        )
        save_path = os.path.join(args.output_dir, "delta.bin")
        pipeline.save_pretrained(
            save_path, parameter_group=args.parameter_group)

        # run inference
        if args.validation_prompt and args.num_validation_images > 0:
            pipeline.scheduler = DPMSolverMultistepScheduler.from_config(
                pipeline.scheduler.config)
            pipeline = pipeline.to(accelerator.device)
            pipeline.set_progress_bar_config(disable=True)

            # run inference
            generator = torch.Generator(
                device=accelerator.device).manual_seed(args.seed)
            images = [
                pipeline(args.validation_prompt, num_inference_steps=25,
                         generator=generator, eta=1.).images[0]
                for _ in range(args.num_validation_images)
            ]

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

        if args.push_to_hub:
            save_model_card(
                repo_id,
                images=images,
                base_model=args.pretrained_model_name_or_path,
                prompt=args.instance_prompt,
                repo_folder=args.output_dir,
            )
            api = HfApi(token=args.hub_token)
            api.upload_folder(
                repo_id=repo_id,
                folder_path=args.output_dir,
                path_in_repo='.',
                repo_type='model'
            )

    accelerator.end_training()


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