Kohya-ss-sd-scripts/train_network_appl_weights.py

import importlib
import argparse
import gc
import math
import os
import sys
import random
import time
import json
from multiprocessing import Value
import toml

from tqdm import tqdm
import torch
from torch.nn.parallel import DistributedDataParallel as DDP

from library.ipex_interop import init_ipex

init_ipex()

from accelerate.utils import set_seed
from diffusers import DDPMScheduler
from library import model_util, sdxl_model_util, sdxl_train_util

import library.train_util as train_util
from library.train_util import (
    DreamBoothDataset,
)
import library.config_util as config_util
from library.config_util import (
    ConfigSanitizer,
    BlueprintGenerator,
)
import library.huggingface_util as huggingface_util
import library.custom_train_functions as custom_train_functions
from library.custom_train_functions import (
    apply_snr_weight,
    get_weighted_text_embeddings,
    prepare_scheduler_for_custom_training,
    scale_v_prediction_loss_like_noise_prediction,
    add_v_prediction_like_loss,
    apply_debiased_estimation,
)


class NetworkApplicationWeights(torch.nn.Module):
    def __init__(self, is_sdxl, num_networks, num_weights_for_network):
        super().__init__()
        self.is_sdxl = is_sdxl
        self.num_networks = num_networks
        self.num_weights_for_network = num_weights_for_network
        # self.weights = torch.nn.Parameter(torch.rand(sum(num_weights_for_network), requires_grad=True))
        # self.weights = torch.nn.Parameter(torch.zeros(sum(num_weights_for_network), requires_grad=True))
        self.weights = torch.nn.Parameter(torch.full((sum(num_weights_for_network),), 0.5, requires_grad=True))

    def apply_weights(self, networks):
        weight_index = 0
        for i, network in enumerate(networks):
            network_weights = self.weights[weight_index : weight_index + self.num_weights_for_network[i]]
            weight_index += self.num_weights_for_network[i]
            network.set_block_wise_weights(network_weights)

    def forward(self, networks, unet_func, unet_args):
        self.apply_weights(networks)

        # I'm not sure if this is the correct way. Is it okay not to call unet here?
        return unet_func(*unet_args)

    def print_parameters(self):
        weight_index = 0
        for i in range(self.num_networks):
            network_weights = self.weights[weight_index : weight_index + self.num_weights_for_network[i]]
            weight_index += self.num_weights_for_network[i]

            network_weights = network_weights.detach().cpu().numpy()
            weights_str = ",".join([f"{w:.3f}" for w in network_weights])
            print(f"Network {i} weights: {weights_str}")


class NetworkAppTrainer:
    def __init__(self):
        self.vae_scale_factor = 0.18215
        self.is_sdxl = False

    # TODO 他のスクリプトと共通化する
    def generate_step_logs(
        self, args: argparse.Namespace, current_loss, avr_loss, lr_scheduler, keys_scaled=None, mean_norm=None, maximum_norm=None
    ):
        logs = {"loss/current": current_loss, "loss/average": avr_loss}

        if keys_scaled is not None:
            logs["max_norm/keys_scaled"] = keys_scaled
            logs["max_norm/average_key_norm"] = mean_norm
            logs["max_norm/max_key_norm"] = maximum_norm

        lrs = lr_scheduler.get_last_lr()

        if args.network_train_text_encoder_only or len(lrs) <= 2:  # not block lr (or single block)
            if args.network_train_unet_only:
                logs["lr/unet"] = float(lrs[0])
            elif args.network_train_text_encoder_only:
                logs["lr/textencoder"] = float(lrs[0])
            else:
                logs["lr/textencoder"] = float(lrs[0])
                logs["lr/unet"] = float(lrs[-1])  # may be same to textencoder

            if (
                args.optimizer_type.lower().startswith("DAdapt".lower()) or args.optimizer_type.lower() == "Prodigy".lower()
            ):  # tracking d*lr value of unet.
                logs["lr/d*lr"] = (
                    lr_scheduler.optimizers[-1].param_groups[0]["d"] * lr_scheduler.optimizers[-1].param_groups[0]["lr"]
                )
        else:
            idx = 0
            if not args.network_train_unet_only:
                logs["lr/textencoder"] = float(lrs[0])
                idx = 1

            for i in range(idx, len(lrs)):
                logs[f"lr/group{i}"] = float(lrs[i])
                if args.optimizer_type.lower().startswith("DAdapt".lower()) or args.optimizer_type.lower() == "Prodigy".lower():
                    logs[f"lr/d*lr/group{i}"] = (
                        lr_scheduler.optimizers[-1].param_groups[i]["d"] * lr_scheduler.optimizers[-1].param_groups[i]["lr"]
                    )

        return logs

    def assert_extra_args(self, args, train_dataset_group):
        pass

    def load_target_model(self, args, weight_dtype, accelerator):
        text_encoder, vae, unet, _ = train_util.load_target_model(args, weight_dtype, accelerator)
        return model_util.get_model_version_str_for_sd1_sd2(args.v2, args.v_parameterization), text_encoder, vae, unet

    def load_tokenizer(self, args):
        tokenizer = train_util.load_tokenizer(args)
        return tokenizer

    def is_text_encoder_outputs_cached(self, args):
        return False

    def is_train_text_encoder(self, args):
        return not args.network_train_unet_only and not self.is_text_encoder_outputs_cached(args)

    def cache_text_encoder_outputs_if_needed(
        self, args, accelerator, unet, vae, tokenizers, text_encoders, data_loader, weight_dtype
    ):
        for t_enc in text_encoders:
            t_enc.to(accelerator.device, dtype=weight_dtype)

    def get_text_cond(self, args, accelerator, batch, tokenizers, text_encoders, weight_dtype):
        input_ids = batch["input_ids"].to(accelerator.device)
        encoder_hidden_states = train_util.get_hidden_states(args, input_ids, tokenizers[0], text_encoders[0], weight_dtype)
        return encoder_hidden_states

    def call_unet(self, args, accelerator, unet, noisy_latents, timesteps, text_conds, batch, weight_dtype):
        noise_pred = unet(noisy_latents, timesteps, text_conds).sample
        return noise_pred

    def all_reduce_network(self, accelerator, network):
        for param in network.parameters():
            if param.grad is not None:
                param.grad = accelerator.reduce(param.grad, reduction="mean")

    def sample_images(self, accelerator, args, epoch, global_step, device, vae, tokenizer, text_encoder, unet):
        train_util.sample_images(accelerator, args, epoch, global_step, device, vae, tokenizer, text_encoder, unet)

    def train(self, args):
        session_id = random.randint(0, 2**32)
        training_started_at = time.time()
        train_util.verify_training_args(args)
        train_util.prepare_dataset_args(args, True)

        cache_latents = args.cache_latents
        use_dreambooth_method = args.in_json is None
        use_user_config = args.dataset_config is not None

        if args.seed is None:
            args.seed = random.randint(0, 2**32)
        set_seed(args.seed)

        # tokenizerは単体またはリスト、tokenizersは必ずリスト：既存のコードとの互換性のため
        tokenizer = self.load_tokenizer(args)
        tokenizers = tokenizer if isinstance(tokenizer, list) else [tokenizer]

        # データセットを準備する
        if args.dataset_class is None:
            blueprint_generator = BlueprintGenerator(ConfigSanitizer(True, True, False, True))
            if use_user_config:
                print(f"Loading dataset config from {args.dataset_config}")
                user_config = config_util.load_user_config(args.dataset_config)
                ignored = ["train_data_dir", "reg_data_dir", "in_json"]
                if any(getattr(args, attr) is not None for attr in ignored):
                    print(
                        "ignoring the following options because config file is found: {0} / 設定ファイルが利用されるため以下のオプションは無視されます: {0}".format(
                            ", ".join(ignored)
                        )
                    )
            else:
                if use_dreambooth_method:
                    print("Using DreamBooth method.")
                    user_config = {
                        "datasets": [
                            {
                                "subsets": config_util.generate_dreambooth_subsets_config_by_subdirs(
                                    args.train_data_dir, args.reg_data_dir
                                )
                            }
                        ]
                    }
                else:
                    print("Training with captions.")
                    user_config = {
                        "datasets": [
                            {
                                "subsets": [
                                    {
                                        "image_dir": args.train_data_dir,
                                        "metadata_file": args.in_json,
                                    }
                                ]
                            }
                        ]
                    }

            blueprint = blueprint_generator.generate(user_config, args, tokenizer=tokenizer)
            train_dataset_group = config_util.generate_dataset_group_by_blueprint(blueprint.dataset_group)
        else:
            # use arbitrary dataset class
            train_dataset_group = train_util.load_arbitrary_dataset(args, tokenizer)

        current_epoch = Value("i", 0)
        current_step = Value("i", 0)
        ds_for_collator = train_dataset_group if args.max_data_loader_n_workers == 0 else None
        collator = train_util.collator_class(current_epoch, current_step, ds_for_collator)

        if args.debug_dataset:
            train_util.debug_dataset(train_dataset_group)
            return
        if len(train_dataset_group) == 0:
            print(
                "No data found. Please verify arguments (train_data_dir must be the parent of folders with images) / 画像がありません。引数指定を確認してください（train_data_dirには画像があるフォルダではなく、画像があるフォルダの親フォルダを指定する必要があります）"
            )
            return

        if cache_latents:
            assert (
                train_dataset_group.is_latent_cacheable()
            ), "when caching latents, either color_aug or random_crop cannot be used / latentをキャッシュするときはcolor_augとrandom_cropは使えません"

        self.assert_extra_args(args, train_dataset_group)

        # acceleratorを準備する
        print("preparing accelerator")
        accelerator = train_util.prepare_accelerator(args)
        is_main_process = accelerator.is_main_process

        # mixed precisionに対応した型を用意しておき適宜castする
        weight_dtype, save_dtype = train_util.prepare_dtype(args)
        vae_dtype = torch.float32 if args.no_half_vae else weight_dtype

        # モデルを読み込む
        model_version, text_encoder, vae, unet = self.load_target_model(args, weight_dtype, accelerator)

        # text_encoder is List[CLIPTextModel] or CLIPTextModel
        text_encoders = text_encoder if isinstance(text_encoder, list) else [text_encoder]

        # モデルに xformers とか memory efficient attention を組み込む
        train_util.replace_unet_modules(unet, args.mem_eff_attn, args.xformers, args.sdpa)
        if torch.__version__ >= "2.0.0":  # PyTorch 2.0.0 以上対応のxformersなら以下が使える
            vae.set_use_memory_efficient_attention_xformers(args.xformers)

        # # 差分追加学習のためにモデルを読み込む
        sys.path.append(os.path.dirname(__file__))
        # accelerator.print("import network module:", args.network_module)
        # network_module = importlib.import_module(args.network_module)

        if args.base_weights is not None:
            # base_weights が指定されている場合は、指定された重みを読み込みマージする
            # currently 1st network_module is used for merging weights
            network_module = importlib.import_module(args.network_module[0])
            for i, weight_path in enumerate(args.base_weights):
                if args.base_weights_multiplier is None or len(args.base_weights_multiplier) <= i:
                    multiplier = 1.0
                else:
                    multiplier = args.base_weights_multiplier[i]

                accelerator.print(f"merging module: {weight_path} with multiplier {multiplier}")

                module, weights_sd = network_module.create_network_from_weights(
                    multiplier, weight_path, vae, text_encoder, unet, for_inference=True
                )
                module.merge_to(text_encoder, unet, weights_sd, weight_dtype, accelerator.device if args.lowram else "cpu")

            accelerator.print(f"all weights merged: {', '.join(args.base_weights)}")

        # 学習を準備する
        if cache_latents:
            vae.to(accelerator.device, dtype=vae_dtype)
            vae.requires_grad_(False)
            vae.eval()
            with torch.no_grad():
                train_dataset_group.cache_latents(vae, args.vae_batch_size, args.cache_latents_to_disk, accelerator.is_main_process)
            vae.to("cpu")
            if torch.cuda.is_available():
                torch.cuda.empty_cache()
            gc.collect()

            accelerator.wait_for_everyone()

        # 必要ならテキストエンコーダーの出力をキャッシュする: Text Encoderはcpuまたはgpuへ移される
        # cache text encoder outputs if needed: Text Encoder is moved to cpu or gpu
        self.cache_text_encoder_outputs_if_needed(
            args, accelerator, unet, vae, tokenizers, text_encoders, train_dataset_group, weight_dtype
        )

        # prepare network
        networks = []
        for network_module_name, network_weight in zip(args.network_module, args.network_weights):
            accelerator.print("import network module:", network_module_name)
            network_module = importlib.import_module(network_module_name)

            # currently network_kwargs is not supported
            network, _ = network_module.create_network_from_weights(1, network_weight, vae, text_encoder, unet)  # , **net_kwargs)
            if hasattr(network, "prepare_network"):
                network.prepare_network(args)

            # do not support Text Encoder only LoRA
            network.apply_to(text_encoder, unet, network.has_text_encoder_block(), True)  # train_unet)

            info = network.load_weights(network_weight)
            accelerator.print(f"load network weights from {network_weight}: {info}")

            assert hasattr(network, "set_block_wise_weights"), "network should have set_block_wise_weights method"

            networks.append(network)

        if args.gradient_checkpointing:
            unet.enable_gradient_checkpointing()
            for t_enc in text_encoders:
                t_enc.gradient_checkpointing_enable()
            del t_enc

        # 学習に必要なクラスを準備する
        accelerator.print("prepare optimizer, data loader etc.")

        network_application = NetworkApplicationWeights(
            self.is_sdxl, len(networks), [network.get_number_of_blocks() for network in networks]
        )
        trainable_params = network_application.parameters()

        train_text_encoder = any([network.has_text_encoder_block() for network in networks])
        print(f"train_text_encoder: {train_text_encoder}")

        optimizer_name, optimizer_args, optimizer = train_util.get_optimizer(args, trainable_params)

        # dataloaderを準備する
        # DataLoaderのプロセス数：0はメインプロセスになる
        n_workers = min(args.max_data_loader_n_workers, os.cpu_count() - 1)  # cpu_count-1 ただし最大で指定された数まで

        train_dataloader = torch.utils.data.DataLoader(
            train_dataset_group,
            batch_size=1,
            shuffle=True,
            collate_fn=collator,
            num_workers=n_workers,
            persistent_workers=args.persistent_data_loader_workers,
        )

        # 学習ステップ数を計算する
        if args.max_train_epochs is not None:
            args.max_train_steps = args.max_train_epochs * math.ceil(
                len(train_dataloader) / accelerator.num_processes / args.gradient_accumulation_steps
            )
            accelerator.print(
                f"override steps. steps for {args.max_train_epochs} epochs is / 指定エポックまでのステップ数: {args.max_train_steps}"
            )

        # データセット側にも学習ステップを送信
        train_dataset_group.set_max_train_steps(args.max_train_steps)

        # lr schedulerを用意する
        lr_scheduler = train_util.get_scheduler_fix(args, optimizer, accelerator.num_processes)

        # # 実験的機能：勾配も含めたfp16/bf16学習を行う　モデル全体をfp16/bf16にする
        # if args.full_fp16:
        #     assert (
        #         args.mixed_precision == "fp16"
        #     ), "full_fp16 requires mixed precision='fp16' / full_fp16を使う場合はmixed_precision='fp16'を指定してください。"
        #     accelerator.print("enable full fp16 training.")
        #     network.to(weight_dtype)
        # elif args.full_bf16:
        #     assert (
        #         args.mixed_precision == "bf16"
        #     ), "full_bf16 requires mixed precision='bf16' / full_bf16を使う場合はmixed_precision='bf16'を指定してください。"
        #     accelerator.print("enable full bf16 training.")
        #     network.to(weight_dtype)

        unet_weight_dtype = te_weight_dtype = weight_dtype
        # Experimental Feature: Put base model into fp8 to save vram
        if args.fp8_base:
            assert torch.__version__ >= "2.1.0", "fp8_base requires torch>=2.1.0 / fp8を使う場合はtorch>=2.1.0が必要です。"
            assert (
                args.mixed_precision != "no"
            ), "fp8_base requires mixed precision='fp16' or 'bf16' / fp8を使う場合はmixed_precision='fp16'または'bf16'が必要です。"
            accelerator.print("enable fp8 training.")
            unet_weight_dtype = torch.float8_e4m3fn
            te_weight_dtype = torch.float8_e4m3fn

        unet.requires_grad_(False)
        unet.to(dtype=unet_weight_dtype)
        for t_enc in text_encoders:
            t_enc.requires_grad_(False)

            # in case of cpu, dtype is already set to fp32 because cpu does not support fp8/fp16/bf16
            if t_enc.device.type != "cpu":
                t_enc.to(dtype=te_weight_dtype)
                # nn.Embedding not support FP8
                t_enc.text_model.embeddings.to(dtype=(weight_dtype if te_weight_dtype != weight_dtype else te_weight_dtype))

        for network in networks:
            network.requires_grad_(False)
            network.to(dtype=weight_dtype).to(accelerator.device)

        # acceleratorがなんかよろしくやってくれるらしい / accelerator will do something good
        # if train_unet:
        unet = accelerator.prepare(unet)
        # else:
        # unet.to(accelerator.device, dtype=unet_weight_dtype)  # move to device because unet is not prepared by accelerator
        if train_text_encoder:
            if len(text_encoders) > 1:
                text_encoder = text_encoders = [accelerator.prepare(t_enc) for t_enc in text_encoders]
            else:
                text_encoder = accelerator.prepare(text_encoder)
                text_encoders = [text_encoder]
        # else:
        #     pass  # if text_encoder is not trained, no need to prepare. and device and dtype are already set

        network_application, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
            network_application, optimizer, train_dataloader, lr_scheduler
        )

        if args.gradient_checkpointing:
            # according to TI example in Diffusers, train is required
            unet.train()
            for t_enc in text_encoders:
                t_enc.train()

                # set top parameter requires_grad = True for gradient checkpointing works
                if train_text_encoder:
                    t_enc.text_model.embeddings.requires_grad_(True)

        else:
            unet.eval()
            for t_enc in text_encoders:
                t_enc.eval()

        for network in networks:
            network.eval()

        del t_enc

        # accelerator.unwrap_model(network_application).prepare_grad_etc(text_encoder, unet)

        if not cache_latents:  # キャッシュしない場合はVAEを使うのでVAEを準備する
            vae.requires_grad_(False)
            vae.eval()
            vae.to(accelerator.device, dtype=vae_dtype)

        # # 実験的機能：勾配も含めたfp16学習を行う　PyTorchにパッチを当ててfp16でのgrad scaleを有効にする
        # if args.full_fp16:
        #     train_util.patch_accelerator_for_fp16_training(accelerator)

        # resumeする
        train_util.resume_from_local_or_hf_if_specified(accelerator, args)

        # epoch数を計算する
        num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
        num_train_epochs = math.ceil(args.max_train_steps / num_update_steps_per_epoch)
        if (args.save_n_epoch_ratio is not None) and (args.save_n_epoch_ratio > 0):
            args.save_every_n_epochs = math.floor(num_train_epochs / args.save_n_epoch_ratio) or 1

        # 学習する
        # TODO: find a way to handle total batch size when there are multiple datasets
        total_batch_size = args.train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps

        accelerator.print("running training / 学習開始")
        accelerator.print(f"  num train images * repeats / 学習画像の数×繰り返し回数: {train_dataset_group.num_train_images}")
        accelerator.print(f"  num reg images / 正則化画像の数: {train_dataset_group.num_reg_images}")
        accelerator.print(f"  num batches per epoch / 1epochのバッチ数: {len(train_dataloader)}")
        accelerator.print(f"  num epochs / epoch数: {num_train_epochs}")
        accelerator.print(
            f"  batch size per device / バッチサイズ: {', '.join([str(d.batch_size) for d in train_dataset_group.datasets])}"
        )
        # accelerator.print(f"  total train batch size (with parallel & distributed & accumulation) / 総バッチサイズ（並列学習、勾配合計含む）: {total_batch_size}")
        accelerator.print(f"  gradient accumulation steps / 勾配を合計するステップ数 = {args.gradient_accumulation_steps}")
        accelerator.print(f"  total optimization steps / 学習ステップ数: {args.max_train_steps}")

        progress_bar = tqdm(range(args.max_train_steps), smoothing=0, disable=not accelerator.is_local_main_process, desc="steps")
        global_step = 0

        noise_scheduler = DDPMScheduler(
            beta_start=0.00085, beta_end=0.012, beta_schedule="scaled_linear", num_train_timesteps=1000, clip_sample=False
        )
        prepare_scheduler_for_custom_training(noise_scheduler, accelerator.device)
        if args.zero_terminal_snr:
            custom_train_functions.fix_noise_scheduler_betas_for_zero_terminal_snr(noise_scheduler)

        if accelerator.is_main_process:
            init_kwargs = {}
            if args.wandb_run_name:
                init_kwargs["wandb"] = {"name": args.wandb_run_name}
            if args.log_tracker_config is not None:
                init_kwargs = toml.load(args.log_tracker_config)
            accelerator.init_trackers(
                "network_train" if args.log_tracker_name is None else args.log_tracker_name, init_kwargs=init_kwargs
            )

        loss_recorder = train_util.LossRecorder()
        del train_dataset_group

        # # callback for step start
        # if hasattr(accelerator.unwrap_model(network_application), "on_step_start"):
        #     on_step_start = accelerator.unwrap_model(network_application).on_step_start
        # else:
        #     on_step_start = lambda *args, **kwargs: None

        # function for saving/removing
        def save_model(ckpt_name, unwrapped_nw, steps, epoch_no, force_sync_upload=False):
            os.makedirs(args.output_dir, exist_ok=True)
            ckpt_file = os.path.join(args.output_dir, ckpt_name)
            accelerator.print(f"\nsaving checkpoint: {ckpt_file}")

            state_dict = unwrapped_nw.state_dict()
            if os.path.splitext(ckpt_file)[1] == ".safetensors":
                from safetensors.torch import save_file

                save_file(state_dict, ckpt_file)
            else:
                torch.save(state_dict, ckpt_file)

            # print parameters
            unwrapped_nw.print_parameters()

        def remove_model(old_ckpt_name):
            old_ckpt_file = os.path.join(args.output_dir, old_ckpt_name)
            if os.path.exists(old_ckpt_file):
                accelerator.print(f"removing old checkpoint: {old_ckpt_file}")
                os.remove(old_ckpt_file)

        # For --sample_at_first
        accelerator.unwrap_model(network_application).apply_weights(networks)
        accelerator.unwrap_model(network_application).print_parameters()

        self.sample_images(accelerator, args, 0, global_step, accelerator.device, vae, tokenizer, text_encoder, unet)

        # training loop
        for epoch in range(num_train_epochs):
            accelerator.print(f"\nepoch {epoch+1}/{num_train_epochs}")
            current_epoch.value = epoch + 1

            # accelerator.unwrap_model(network_application).on_epoch_start(text_encoder, unet)

            for step, batch in enumerate(train_dataloader):
                current_step.value = global_step
                with accelerator.accumulate(network_application):
                    # on_step_start(text_encoder, unet)

                    with torch.no_grad():
                        if "latents" in batch and batch["latents"] is not None:
                            latents = batch["latents"].to(accelerator.device)
                        else:
                            # latentに変換
                            latents = vae.encode(batch["images"].to(dtype=vae_dtype)).latent_dist.sample()

                            # NaNが含まれていれば警告を表示し0に置き換える
                            if torch.any(torch.isnan(latents)):
                                accelerator.print("NaN found in latents, replacing with zeros")
                                latents = torch.nan_to_num(latents, 0, out=latents)
                        latents = latents * self.vae_scale_factor

                    with torch.set_grad_enabled(train_text_encoder), accelerator.autocast():
                        # Get the text embedding for conditioning
                        if args.weighted_captions:
                            text_encoder_conds = get_weighted_text_embeddings(
                                tokenizer,
                                text_encoder,
                                batch["captions"],
                                accelerator.device,
                                args.max_token_length // 75 if args.max_token_length else 1,
                                clip_skip=args.clip_skip,
                            )
                        else:
                            text_encoder_conds = self.get_text_cond(
                                args, accelerator, batch, tokenizers, text_encoders, weight_dtype
                            )

                    # Sample noise, sample a random timestep for each image, and add noise to the latents,
                    # with noise offset and/or multires noise if specified
                    noise, noisy_latents, timesteps = train_util.get_noise_noisy_latents_and_timesteps(
                        args, noise_scheduler, latents
                    )

                    # ensure the hidden state will require grad
                    if args.gradient_checkpointing:
                        for x in noisy_latents:
                            x.requires_grad_(True)
                        for t in text_encoder_conds:
                            t.requires_grad_(True)

                    # Predict the noise residual
                    with accelerator.autocast():
                        unet_func = self.call_unet
                        unet_args = (
                            args,
                            accelerator,
                            unet,
                            noisy_latents.requires_grad_(True),
                            timesteps,
                            text_encoder_conds,
                            batch,
                            weight_dtype,
                        )
                        noise_pred = network_application(networks, unet_func, unet_args)

                    if args.v_parameterization:
                        # v-parameterization training
                        target = noise_scheduler.get_velocity(latents, noise, timesteps)
                    else:
                        target = noise

                    loss = torch.nn.functional.mse_loss(noise_pred.float(), target.float(), reduction="none")
                    loss = loss.mean([1, 2, 3])

                    loss_weights = batch["loss_weights"]  # 各sampleごとのweight
                    loss = loss * loss_weights

                    if args.min_snr_gamma:
                        loss = apply_snr_weight(loss, timesteps, noise_scheduler, args.min_snr_gamma, args.v_parameterization)
                    if args.scale_v_pred_loss_like_noise_pred:
                        loss = scale_v_prediction_loss_like_noise_prediction(loss, timesteps, noise_scheduler)
                    if args.v_pred_like_loss:
                        loss = add_v_prediction_like_loss(loss, timesteps, noise_scheduler, args.v_pred_like_loss)
                    if args.debiased_estimation_loss:
                        loss = apply_debiased_estimation(loss, timesteps, noise_scheduler)

                    loss = loss.mean()  # 平均なのでbatch_sizeで割る必要なし

                    # use sum of parameter values as loss
                    weights_loss = 0
                    for param in network_application.parameters():
                        # weights_loss += param.abs().sum()
                        # we add more weight for negative values. because we want to keep the weights positive
                        weights_loss += param.abs().sum() + param[param < 0].abs().sum() * 10
                    loss = loss + weights_loss * args.application_loss_weight

                    accelerator.backward(loss)
                    if accelerator.sync_gradients:
                        self.all_reduce_network(accelerator, network_application)  # sync DDP grad manually
                        # if args.max_grad_norm != 0.0:
                        #     params_to_clip = accelerator.unwrap_model(network_application).get_trainable_params()
                        #     accelerator.clip_grad_norm_(params_to_clip, args.max_grad_norm)

                    optimizer.step()
                    lr_scheduler.step()
                    optimizer.zero_grad(set_to_none=True)

                # if args.scale_weight_norms:
                #     keys_scaled, mean_norm, maximum_norm = accelerator.unwrap_model(network_application).apply_max_norm_regularization(
                #         args.scale_weight_norms, accelerator.device
                #     )
                #     max_mean_logs = {"Keys Scaled": keys_scaled, "Average key norm": mean_norm}
                # else:
                #     keys_scaled, mean_norm, maximum_norm = None, None, None

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

                    self.sample_images(accelerator, args, None, global_step, accelerator.device, vae, tokenizer, text_encoder, unet)

                    # 指定ステップごとにモデルを保存
                    if args.save_every_n_steps is not None and global_step % args.save_every_n_steps == 0:
                        accelerator.wait_for_everyone()
                        if accelerator.is_main_process:
                            ckpt_name = train_util.get_step_ckpt_name(args, "." + args.save_model_as, global_step)
                            save_model(ckpt_name, accelerator.unwrap_model(network_application), global_step, epoch)

                            if args.save_state:
                                train_util.save_and_remove_state_stepwise(args, accelerator, global_step)

                            remove_step_no = train_util.get_remove_step_no(args, global_step)
                            if remove_step_no is not None:
                                remove_ckpt_name = train_util.get_step_ckpt_name(args, "." + args.save_model_as, remove_step_no)
                                remove_model(remove_ckpt_name)

                current_loss = loss.detach().item()
                loss_recorder.add(epoch=epoch, step=step, loss=current_loss)
                avr_loss: float = loss_recorder.moving_average
                logs = {"avr_loss": avr_loss}  # , "lr": lr_scheduler.get_last_lr()[0]}

                logs["ap_loss"] = weights_loss.detach().item()

                progress_bar.set_postfix(**logs)

                # if args.scale_weight_norms:
                #     progress_bar.set_postfix(**{**max_mean_logs, **logs})

                if args.logging_dir is not None:
                    # logs = self.generate_step_logs(args, current_loss, avr_loss, lr_scheduler, keys_scaled, mean_norm, maximum_norm)
                    accelerator.log(logs, step=global_step)

                if global_step >= args.max_train_steps:
                    break

            if args.logging_dir is not None:
                logs = {"loss/epoch": loss_recorder.moving_average}
                accelerator.log(logs, step=epoch + 1)

            accelerator.wait_for_everyone()

            # 指定エポックごとにモデルを保存
            if args.save_every_n_epochs is not None:
                saving = (epoch + 1) % args.save_every_n_epochs == 0 and (epoch + 1) < num_train_epochs
                if is_main_process and saving:
                    ckpt_name = train_util.get_epoch_ckpt_name(args, "." + args.save_model_as, epoch + 1)
                    save_model(ckpt_name, accelerator.unwrap_model(network_application), global_step, epoch + 1)

                    remove_epoch_no = train_util.get_remove_epoch_no(args, epoch + 1)
                    if remove_epoch_no is not None:
                        remove_ckpt_name = train_util.get_epoch_ckpt_name(args, "." + args.save_model_as, remove_epoch_no)
                        remove_model(remove_ckpt_name)

                    if args.save_state:
                        train_util.save_and_remove_state_on_epoch_end(args, accelerator, epoch + 1)

            self.sample_images(accelerator, args, epoch + 1, global_step, accelerator.device, vae, tokenizer, text_encoder, unet)

            # end of epoch

        # metadata["ss_epoch"] = str(num_train_epochs)
        # metadata["ss_training_finished_at"] = str(time.time())

        if is_main_process:
            network_application = accelerator.unwrap_model(network_application)

        accelerator.end_training()

        if is_main_process and args.save_state:
            train_util.save_state_on_train_end(args, accelerator)

        if is_main_process:
            ckpt_name = train_util.get_last_ckpt_name(args, "." + args.save_model_as)
            save_model(ckpt_name, network_application, global_step, num_train_epochs, force_sync_upload=True)

            print("model saved.")


def setup_parser() -> argparse.ArgumentParser:
    parser = argparse.ArgumentParser()

    train_util.add_sd_models_arguments(parser)
    train_util.add_dataset_arguments(parser, True, True, True)
    train_util.add_training_arguments(parser, True)
    train_util.add_optimizer_arguments(parser)
    config_util.add_config_arguments(parser)
    custom_train_functions.add_custom_train_arguments(parser)

    # parser.add_argument(
    #     "--no_metadata", action="store_true", help="do not save metadata in output model / メタデータを出力先モデルに保存しない"
    # )
    parser.add_argument(
        "--save_model_as",
        type=str,
        default="safetensors",
        choices=[None, "ckpt", "pt", "safetensors"],
        help="format to save the model (default is .safetensors) / モデル保存時の形式（デフォルトはsafetensors）",
    )

    # parser.add_argument("--unet_lr", type=float, default=None, help="learning rate for U-Net / U-Netの学習率")
    # parser.add_argument("--text_encoder_lr", type=float, default=None, help="learning rate for Text Encoder / Text Encoderの学習率")

    parser.add_argument(
        "--network_weights",
        type=str,
        nargs="+",
        default=None,
        help="pretrained weights for network / 学習するネットワークの初期重み",
    )
    parser.add_argument(
        "--network_module", type=str, nargs="+", default=None, help="network module to train / 学習対象のネットワークのモジュール"
    )
    # parser.add_argument(
    #     "--network_dim",
    #     type=int,
    #     default=None,
    #     help="network dimensions (depends on each network) / モジュールの次元数（ネットワークにより定義は異なります）",
    # )
    # parser.add_argument(
    #     "--network_alpha",
    #     type=float,
    #     default=1,
    #     help="alpha for LoRA weight scaling, default 1 (same as network_dim for same behavior as old version) / LoRaの重み調整のalpha値、デフォルト1（旧バージョンと同じ動作をするにはnetwork_dimと同じ値を指定）",
    # )
    # parser.add_argument(
    #     "--network_dropout",
    #     type=float,
    #     default=None,
    #     help="Drops neurons out of training every step (0 or None is default behavior (no dropout), 1 would drop all neurons) / 訓練時に毎ステップでニューロンをdropする（0またはNoneはdropoutなし、1は全ニューロンをdropout）",
    # )
    # parser.add_argument(
    #     "--network_args",
    #     type=str,
    #     default=None,
    #     nargs="*",
    #     help="additional arguments for network (key=value) / ネットワークへの追加の引数",
    # )
    # parser.add_argument(
    #     "--network_train_unet_only", action="store_true", help="only training U-Net part / U-Net関連部分のみ学習する"
    # )
    # parser.add_argument(
    #     "--network_train_text_encoder_only",
    #     action="store_true",
    #     help="only training Text Encoder part / Text Encoder関連部分のみ学習する",
    # )
    # parser.add_argument(
    #     "--training_comment",
    #     type=str,
    #     default=None,
    #     help="arbitrary comment string stored in metadata / メタデータに記録する任意のコメント文字列",
    # )
    parser.add_argument(
        "--dim_from_weights",
        action="store_true",
        help="automatically determine dim (rank) from network_weights / dim (rank)をnetwork_weightsで指定した重みから自動で決定する",
    )
    parser.add_argument(
        "--scale_weight_norms",
        type=float,
        default=None,
        help="Scale the weight of each key pair to help prevent overtraing via exploding gradients. (1 is a good starting point) / 重みの値をスケーリングして勾配爆発を防ぐ（1が初期値としては適当）",
    )
    parser.add_argument(
        "--base_weights",
        type=str,
        default=None,
        nargs="*",
        help="network weights to merge into the model before training / 学習前にあらかじめモデルにマージするnetworkの重みファイル",
    )
    parser.add_argument(
        "--base_weights_multiplier",
        type=float,
        default=None,
        nargs="*",
        help="multiplier for network weights to merge into the model before training / 学習前にあらかじめモデルにマージするnetworkの重みの倍率",
    )
    parser.add_argument(
        "--no_half_vae",
        action="store_true",
        help="do not use fp16/bf16 VAE in mixed precision (use float VAE) / mixed precisionでも fp16/bf16 VAEを使わずfloat VAEを使う",
    )

    parser.add_argument(
        "--application_loss_weight", type=float, default=0.0001, help="weight for application loss / application lossの重み"
    )

    sdxl_train_util.add_sdxl_training_arguments(parser)

    return parser


class SdxlNetworkAppTrainer(NetworkAppTrainer):
    def __init__(self):
        super().__init__()
        self.vae_scale_factor = sdxl_model_util.VAE_SCALE_FACTOR
        self.is_sdxl = True

    def assert_extra_args(self, args, train_dataset_group):
        super().assert_extra_args(args, train_dataset_group)
        sdxl_train_util.verify_sdxl_training_args(args)

        if args.cache_text_encoder_outputs:
            assert (
                train_dataset_group.is_text_encoder_output_cacheable()
            ), "when caching Text Encoder output, either caption_dropout_rate, shuffle_caption, token_warmup_step or caption_tag_dropout_rate cannot be used / Text Encoderの出力をキャッシュするときはcaption_dropout_rate, shuffle_caption, token_warmup_step, caption_tag_dropout_rateは使えません"

        assert (
            args.network_train_unet_only or not args.cache_text_encoder_outputs
        ), "network for Text Encoder cannot be trained with caching Text Encoder outputs / Text Encoderの出力をキャッシュしながらText Encoderのネットワークを学習することはできません"

        train_dataset_group.verify_bucket_reso_steps(32)

    def load_target_model(self, args, weight_dtype, accelerator):
        (
            load_stable_diffusion_format,
            text_encoder1,
            text_encoder2,
            vae,
            unet,
            logit_scale,
            ckpt_info,
        ) = sdxl_train_util.load_target_model(args, accelerator, sdxl_model_util.MODEL_VERSION_SDXL_BASE_V1_0, weight_dtype)

        self.load_stable_diffusion_format = load_stable_diffusion_format
        self.logit_scale = logit_scale
        self.ckpt_info = ckpt_info

        return sdxl_model_util.MODEL_VERSION_SDXL_BASE_V1_0, [text_encoder1, text_encoder2], vae, unet

    def load_tokenizer(self, args):
        tokenizer = sdxl_train_util.load_tokenizers(args)
        return tokenizer

    def is_text_encoder_outputs_cached(self, args):
        return args.cache_text_encoder_outputs

    def cache_text_encoder_outputs_if_needed(
        self, args, accelerator, unet, vae, tokenizers, text_encoders, dataset: train_util.DatasetGroup, weight_dtype
    ):
        if args.cache_text_encoder_outputs:
            if not args.lowram:
                # メモリ消費を減らす
                print("move vae and unet to cpu to save memory")
                org_vae_device = vae.device
                org_unet_device = unet.device
                vae.to("cpu")
                unet.to("cpu")
                if torch.cuda.is_available():
                    torch.cuda.empty_cache()

            # When TE is not be trained, it will not be prepared so we need to use explicit autocast
            with accelerator.autocast():
                dataset.cache_text_encoder_outputs(
                    tokenizers,
                    text_encoders,
                    accelerator.device,
                    weight_dtype,
                    args.cache_text_encoder_outputs_to_disk,
                    accelerator.is_main_process,
                )

            text_encoders[0].to("cpu", dtype=torch.float32)  # Text Encoder doesn't work with fp16 on CPU
            text_encoders[1].to("cpu", dtype=torch.float32)
            if torch.cuda.is_available():
                torch.cuda.empty_cache()

            if not args.lowram:
                print("move vae and unet back to original device")
                vae.to(org_vae_device)
                unet.to(org_unet_device)
        else:
            # Text Encoderから毎回出力を取得するので、GPUに乗せておく
            text_encoders[0].to(accelerator.device, dtype=weight_dtype)
            text_encoders[1].to(accelerator.device, dtype=weight_dtype)

    def get_text_cond(self, args, accelerator, batch, tokenizers, text_encoders, weight_dtype):
        if "text_encoder_outputs1_list" not in batch or batch["text_encoder_outputs1_list"] is None:
            input_ids1 = batch["input_ids"]
            input_ids2 = batch["input_ids2"]
            with torch.enable_grad():
                # Get the text embedding for conditioning
                # TODO support weighted captions
                # if args.weighted_captions:
                #     encoder_hidden_states = get_weighted_text_embeddings(
                #         tokenizer,
                #         text_encoder,
                #         batch["captions"],
                #         accelerator.device,
                #         args.max_token_length // 75 if args.max_token_length else 1,
                #         clip_skip=args.clip_skip,
                #     )
                # else:
                input_ids1 = input_ids1.to(accelerator.device)
                input_ids2 = input_ids2.to(accelerator.device)
                encoder_hidden_states1, encoder_hidden_states2, pool2 = train_util.get_hidden_states_sdxl(
                    args.max_token_length,
                    input_ids1,
                    input_ids2,
                    tokenizers[0],
                    tokenizers[1],
                    text_encoders[0],
                    text_encoders[1],
                    None if not args.full_fp16 else weight_dtype,
                    accelerator=accelerator,
                )
        else:
            encoder_hidden_states1 = batch["text_encoder_outputs1_list"].to(accelerator.device).to(weight_dtype)
            encoder_hidden_states2 = batch["text_encoder_outputs2_list"].to(accelerator.device).to(weight_dtype)
            pool2 = batch["text_encoder_pool2_list"].to(accelerator.device).to(weight_dtype)

            # # verify that the text encoder outputs are correct
            # ehs1, ehs2, p2 = train_util.get_hidden_states_sdxl(
            #     args.max_token_length,
            #     batch["input_ids"].to(text_encoders[0].device),
            #     batch["input_ids2"].to(text_encoders[0].device),
            #     tokenizers[0],
            #     tokenizers[1],
            #     text_encoders[0],
            #     text_encoders[1],
            #     None if not args.full_fp16 else weight_dtype,
            # )
            # b_size = encoder_hidden_states1.shape[0]
            # assert ((encoder_hidden_states1.to("cpu") - ehs1.to(dtype=weight_dtype)).abs().max() > 1e-2).sum() <= b_size * 2
            # assert ((encoder_hidden_states2.to("cpu") - ehs2.to(dtype=weight_dtype)).abs().max() > 1e-2).sum() <= b_size * 2
            # assert ((pool2.to("cpu") - p2.to(dtype=weight_dtype)).abs().max() > 1e-2).sum() <= b_size * 2
            # print("text encoder outputs verified")

        return encoder_hidden_states1, encoder_hidden_states2, pool2

    def call_unet(self, args, accelerator, unet, noisy_latents, timesteps, text_conds, batch, weight_dtype):
        noisy_latents = noisy_latents.to(weight_dtype)  # TODO check why noisy_latents is not weight_dtype

        # get size embeddings
        orig_size = batch["original_sizes_hw"]
        crop_size = batch["crop_top_lefts"]
        target_size = batch["target_sizes_hw"]
        embs = sdxl_train_util.get_size_embeddings(orig_size, crop_size, target_size, accelerator.device).to(weight_dtype)

        # concat embeddings
        encoder_hidden_states1, encoder_hidden_states2, pool2 = text_conds
        vector_embedding = torch.cat([pool2, embs], dim=1).to(weight_dtype)
        text_embedding = torch.cat([encoder_hidden_states1, encoder_hidden_states2], dim=2).to(weight_dtype)

        noise_pred = unet(noisy_latents, timesteps, text_embedding, vector_embedding)
        return noise_pred

    def sample_images(self, accelerator, args, epoch, global_step, device, vae, tokenizer, text_encoder, unet):
        sdxl_train_util.sample_images(accelerator, args, epoch, global_step, device, vae, tokenizer, text_encoder, unet)


if __name__ == "__main__":
    parser = setup_parser()

    args = parser.parse_args()
    args = train_util.read_config_from_file(args, parser)

    trainer = SdxlNetworkAppTrainer()
    trainer.train(args)