Kohya-ss-sd-scripts/lumina_train.py

# training with captions

# Swap blocks between CPU and GPU:
# This implementation is inspired by and based on the work of 2kpr.
# Many thanks to 2kpr for the original concept and implementation of memory-efficient offloading.
# The original idea has been adapted and extended to fit the current project's needs.

# Key features:
# - CPU offloading during forward and backward passes
# - Use of fused optimizer and grad_hook for efficient gradient processing
# - Per-block fused optimizer instances

import argparse
import copy
import math
import os
from multiprocessing import Value
import toml

from tqdm import tqdm

import torch
from library.device_utils import init_ipex, clean_memory_on_device

init_ipex()

from accelerate.utils import set_seed
from library import (
    deepspeed_utils,
    lumina_train_util,
    lumina_util,
    strategy_base,
    strategy_lumina,
    sai_model_spec
)
from library.sd3_train_utils import FlowMatchEulerDiscreteScheduler

import library.train_util as train_util

from library.utils import setup_logging, add_logging_arguments

setup_logging()
import logging

logger = logging.getLogger(__name__)

import library.config_util as config_util

# import library.sdxl_train_util as sdxl_train_util
from library.config_util import (
    ConfigSanitizer,
    BlueprintGenerator,
)
from library.custom_train_functions import apply_masked_loss, add_custom_train_arguments


def train(args):
    train_util.verify_training_args(args)
    train_util.prepare_dataset_args(args, True)
    # sdxl_train_util.verify_sdxl_training_args(args)
    deepspeed_utils.prepare_deepspeed_args(args)
    setup_logging(args, reset=True)

    # temporary: backward compatibility for deprecated options. remove in the future
    if not args.skip_cache_check:
        args.skip_cache_check = args.skip_latents_validity_check

    # assert (
    #     not args.weighted_captions
    # ), "weighted_captions is not supported currently / weighted_captionsは現在サポートされていません"
    if args.cache_text_encoder_outputs_to_disk and not args.cache_text_encoder_outputs:
        logger.warning(
            "cache_text_encoder_outputs_to_disk is enabled, so cache_text_encoder_outputs is also enabled / cache_text_encoder_outputs_to_diskが有効になっているため、cache_text_encoder_outputsも有効になります"
        )
        args.cache_text_encoder_outputs = True

    if args.cpu_offload_checkpointing and not args.gradient_checkpointing:
        logger.warning(
            "cpu_offload_checkpointing is enabled, so gradient_checkpointing is also enabled / cpu_offload_checkpointingが有効になっているため、gradient_checkpointingも有効になります"
        )
        args.gradient_checkpointing = True

    # assert (
    #     args.blocks_to_swap is None or args.blocks_to_swap == 0
    # ) or not args.cpu_offload_checkpointing, "blocks_to_swap is not supported with cpu_offload_checkpointing / blocks_to_swapはcpu_offload_checkpointingと併用できません"

    cache_latents = args.cache_latents
    use_dreambooth_method = args.in_json is None

    if args.seed is not None:
        set_seed(args.seed)  # 乱数系列を初期化する

    # prepare caching strategy: this must be set before preparing dataset. because dataset may use this strategy for initialization.
    if args.cache_latents:
        latents_caching_strategy = strategy_lumina.LuminaLatentsCachingStrategy(
            args.cache_latents_to_disk, args.vae_batch_size, args.skip_cache_check
        )
        strategy_base.LatentsCachingStrategy.set_strategy(latents_caching_strategy)

    # データセットを準備する
    if args.dataset_class is None:
        blueprint_generator = BlueprintGenerator(
            ConfigSanitizer(True, True, args.masked_loss, True)
        )
        if args.dataset_config is not None:
            logger.info(f"Load dataset config from {args.dataset_config}")
            user_config = config_util.load_user_config(args.dataset_config)
            ignored = ["train_data_dir", "in_json"]
            if any(getattr(args, attr) is not None for attr in ignored):
                logger.warning(
                    "ignore 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:
                logger.info("Using DreamBooth method.")
                user_config = {
                    "datasets": [
                        {
                            "subsets": config_util.generate_dreambooth_subsets_config_by_subdirs(
                                args.train_data_dir, args.reg_data_dir
                            )
                        }
                    ]
                }
            else:
                logger.info("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)
        train_dataset_group, val_dataset_group = (
            config_util.generate_dataset_group_by_blueprint(blueprint.dataset_group)
        )
    else:
        train_dataset_group = train_util.load_arbitrary_dataset(args)
        val_dataset_group = None

    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)

    train_dataset_group.verify_bucket_reso_steps(16)  # TODO これでいいか確認

    if args.debug_dataset:
        if args.cache_text_encoder_outputs:
            strategy_base.TextEncoderOutputsCachingStrategy.set_strategy(
                strategy_lumina.LuminaTextEncoderOutputsCachingStrategy(
                    args.cache_text_encoder_outputs_to_disk,
                    args.text_encoder_batch_size,
                    args.skip_cache_check,
                    False,
                )
            )
        strategy_base.TokenizeStrategy.set_strategy(
            strategy_lumina.LuminaTokenizeStrategy(args.system_prompt)
        )

        train_dataset_group.set_current_strategies()
        train_util.debug_dataset(train_dataset_group, True)
        return
    if len(train_dataset_group) == 0:
        logger.error(
            "No data found. Please verify the metadata file and train_data_dir option. / 画像がありません。メタデータおよびtrain_data_dirオプションを確認してください。"
        )
        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"

    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"

    # acceleratorを準備する
    logger.info("prepare accelerator")
    accelerator = train_util.prepare_accelerator(args)

    # 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)

    # モデルを読み込む

    # load VAE for caching latents
    ae = None
    if cache_latents:
        ae = lumina_util.load_ae(
            args.ae, weight_dtype, "cpu", args.disable_mmap_load_safetensors
        )
        ae.to(accelerator.device, dtype=weight_dtype)
        ae.requires_grad_(False)
        ae.eval()

        train_dataset_group.new_cache_latents(ae, accelerator)

        ae.to("cpu")  # if no sampling, vae can be deleted
        clean_memory_on_device(accelerator.device)

        accelerator.wait_for_everyone()

    # prepare tokenize strategy
    if args.gemma2_max_token_length is None:
        gemma2_max_token_length = 256
    else:
        gemma2_max_token_length = args.gemma2_max_token_length

    lumina_tokenize_strategy = strategy_lumina.LuminaTokenizeStrategy(
        args.system_prompt, gemma2_max_token_length
    )
    strategy_base.TokenizeStrategy.set_strategy(lumina_tokenize_strategy)

    # load gemma2 for caching text encoder outputs
    gemma2 = lumina_util.load_gemma2(
        args.gemma2, weight_dtype, "cpu", args.disable_mmap_load_safetensors
    )
    gemma2.eval()
    gemma2.requires_grad_(False)

    text_encoding_strategy = strategy_lumina.LuminaTextEncodingStrategy()
    strategy_base.TextEncodingStrategy.set_strategy(text_encoding_strategy)

    # cache text encoder outputs
    sample_prompts_te_outputs = None
    if args.cache_text_encoder_outputs:
        # Text Encodes are eval and no grad here
        gemma2.to(accelerator.device)

        text_encoder_caching_strategy = (
            strategy_lumina.LuminaTextEncoderOutputsCachingStrategy(
                args.cache_text_encoder_outputs_to_disk,
                args.text_encoder_batch_size,
                False,
                False,
            )
        )
        strategy_base.TextEncoderOutputsCachingStrategy.set_strategy(
            text_encoder_caching_strategy
        )

        with accelerator.autocast():
            train_dataset_group.new_cache_text_encoder_outputs([gemma2], accelerator)

        # cache sample prompt's embeddings to free text encoder's memory
        if args.sample_prompts is not None:
            logger.info(
                f"cache Text Encoder outputs for sample prompt: {args.sample_prompts}"
            )

            text_encoding_strategy: strategy_lumina.LuminaTextEncodingStrategy = (
                strategy_base.TextEncodingStrategy.get_strategy()
            )

            prompts = train_util.load_prompts(args.sample_prompts)
            sample_prompts_te_outputs = {}  # key: prompt, value: text encoder outputs
            with accelerator.autocast(), torch.no_grad():
                for prompt_dict in prompts:
                    for i, p in enumerate([
                        prompt_dict.get("prompt", ""),
                        prompt_dict.get("negative_prompt", ""),
                    ]):
                        if p not in sample_prompts_te_outputs:
                            logger.info(f"cache Text Encoder outputs for prompt: {p}")
                            tokens_and_masks = lumina_tokenize_strategy.tokenize(p, i == 1)  # i == 1 means negative prompt
                            sample_prompts_te_outputs[p] = (
                                text_encoding_strategy.encode_tokens(
                                    lumina_tokenize_strategy,
                                    [gemma2],
                                    tokens_and_masks,
                                )
                            )

        accelerator.wait_for_everyone()

        # now we can delete Text Encoders to free memory
        gemma2 = None
        clean_memory_on_device(accelerator.device)

    # load lumina
    nextdit = lumina_util.load_lumina_model(
        args.pretrained_model_name_or_path,
        weight_dtype,
        torch.device("cpu"),
        disable_mmap=args.disable_mmap_load_safetensors,
        use_flash_attn=args.use_flash_attn,
    )

    if args.gradient_checkpointing:
        nextdit.enable_gradient_checkpointing(
            cpu_offload=args.cpu_offload_checkpointing
        )

    nextdit.requires_grad_(True)

    # block swap

    # backward compatibility
    # if args.blocks_to_swap is None:
    #     blocks_to_swap = args.double_blocks_to_swap or 0
    #     if args.single_blocks_to_swap is not None:
    #         blocks_to_swap += args.single_blocks_to_swap // 2
    #     if blocks_to_swap > 0:
    #         logger.warning(
    #             "double_blocks_to_swap and single_blocks_to_swap are deprecated. Use blocks_to_swap instead."
    #             " / double_blocks_to_swapとsingle_blocks_to_swapは非推奨です。blocks_to_swapを使ってください。"
    #         )
    #         logger.info(
    #             f"double_blocks_to_swap={args.double_blocks_to_swap} and single_blocks_to_swap={args.single_blocks_to_swap} are converted to blocks_to_swap={blocks_to_swap}."
    #         )
    #         args.blocks_to_swap = blocks_to_swap
    #     del blocks_to_swap

    # is_swapping_blocks = args.blocks_to_swap is not None and args.blocks_to_swap > 0
    # if is_swapping_blocks:
    #     # Swap blocks between CPU and GPU to reduce memory usage, in forward and backward passes.
    #     # This idea is based on 2kpr's great work. Thank you!
    #     logger.info(f"enable block swap: blocks_to_swap={args.blocks_to_swap}")
    #     flux.enable_block_swap(args.blocks_to_swap, accelerator.device)

    if not cache_latents:
        # load VAE here if not cached
        ae = lumina_util.load_ae(args.ae, weight_dtype, "cpu")
        ae.requires_grad_(False)
        ae.eval()
        ae.to(accelerator.device, dtype=weight_dtype)

    training_models = []
    params_to_optimize = []
    training_models.append(nextdit)
    name_and_params = list(nextdit.named_parameters())
    # single param group for now
    params_to_optimize.append(
        {"params": [p for _, p in name_and_params], "lr": args.learning_rate}
    )
    param_names = [[n for n, _ in name_and_params]]

    # calculate number of trainable parameters
    n_params = 0
    for group in params_to_optimize:
        for p in group["params"]:
            n_params += p.numel()

    accelerator.print(f"number of trainable parameters: {n_params}")

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

    if args.blockwise_fused_optimizers:
        # fused backward pass: https://pytorch.org/tutorials/intermediate/optimizer_step_in_backward_tutorial.html
        # Instead of creating an optimizer for all parameters as in the tutorial, we create an optimizer for each block of parameters.
        # This balances memory usage and management complexity.

        # split params into groups. currently different learning rates are not supported
        grouped_params = []
        param_group = {}
        for group in params_to_optimize:
            named_parameters = list(nextdit.named_parameters())
            assert len(named_parameters) == len(
                group["params"]
            ), "number of parameters does not match"
            for p, np in zip(group["params"], named_parameters):
                # determine target layer and block index for each parameter
                block_type = "other"  # double, single or other
                if np[0].startswith("double_blocks"):
                    block_index = int(np[0].split(".")[1])
                    block_type = "double"
                elif np[0].startswith("single_blocks"):
                    block_index = int(np[0].split(".")[1])
                    block_type = "single"
                else:
                    block_index = -1

                param_group_key = (block_type, block_index)
                if param_group_key not in param_group:
                    param_group[param_group_key] = []
                param_group[param_group_key].append(p)

        block_types_and_indices = []
        for param_group_key, param_group in param_group.items():
            block_types_and_indices.append(param_group_key)
            grouped_params.append({"params": param_group, "lr": args.learning_rate})

            num_params = 0
            for p in param_group:
                num_params += p.numel()
            accelerator.print(f"block {param_group_key}: {num_params} parameters")

        # prepare optimizers for each group
        optimizers = []
        for group in grouped_params:
            _, _, optimizer = train_util.get_optimizer(args, trainable_params=[group])
            optimizers.append(optimizer)
        optimizer = optimizers[0]  # avoid error in the following code

        logger.info(
            f"using {len(optimizers)} optimizers for blockwise fused optimizers"
        )

        if train_util.is_schedulefree_optimizer(optimizers[0], args):
            raise ValueError(
                "Schedule-free optimizer is not supported with blockwise fused optimizers"
            )
        optimizer_train_fn = lambda: None  # dummy function
        optimizer_eval_fn = lambda: None  # dummy function
    else:
        _, _, optimizer = train_util.get_optimizer(
            args, trainable_params=params_to_optimize
        )
        optimizer_train_fn, optimizer_eval_fn = train_util.get_optimizer_train_eval_fn(
            optimizer, args
        )

    # prepare dataloader
    # strategies are set here because they cannot be referenced in another process. Copy them with the dataset
    # some strategies can be None
    train_dataset_group.set_current_strategies()

    # DataLoaderのプロセス数：0 は persistent_workers が使えないので注意
    n_workers = min(
        args.max_data_loader_n_workers, os.cpu_count()
    )  # cpu_count or max_data_loader_n_workers
    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
    if args.blockwise_fused_optimizers:
        # prepare lr schedulers for each optimizer
        lr_schedulers = [
            train_util.get_scheduler_fix(args, optimizer, accelerator.num_processes)
            for optimizer in optimizers
        ]
        lr_scheduler = lr_schedulers[0]  # avoid error in the following code
    else:
        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.")
        nextdit.to(weight_dtype)
        if gemma2 is not None:
            gemma2.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.")
        nextdit.to(weight_dtype)
        if gemma2 is not None:
            gemma2.to(weight_dtype)

    # if we don't cache text encoder outputs, move them to device
    if not args.cache_text_encoder_outputs:
        gemma2.to(accelerator.device)

    clean_memory_on_device(accelerator.device)

    is_swapping_blocks = args.blocks_to_swap is not None and args.blocks_to_swap > 0

    if args.deepspeed:
        ds_model = deepspeed_utils.prepare_deepspeed_model(args, nextdit=nextdit)
        # most of ZeRO stage uses optimizer partitioning, so we have to prepare optimizer and ds_model at the same time. # pull/1139#issuecomment-1986790007
        ds_model, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
            ds_model, optimizer, train_dataloader, lr_scheduler
        )
        training_models = [ds_model]

    else:
        # accelerator does some magic
        # if we doesn't swap blocks, we can move the model to device
        nextdit = accelerator.prepare(
            nextdit, device_placement=[not is_swapping_blocks]
        )
        if is_swapping_blocks:
            accelerator.unwrap_model(nextdit).move_to_device_except_swap_blocks(
                accelerator.device
            )  # reduce peak memory usage
        optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
            optimizer, train_dataloader, lr_scheduler
        )

    # 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:
        # During deepseed training, accelerate not handles fp16/bf16|mixed precision directly via scaler. Let deepspeed engine do.
        # -> But we think it's ok to patch accelerator even if deepspeed is enabled.
        train_util.patch_accelerator_for_fp16_training(accelerator)

    # resumeする
    train_util.resume_from_local_or_hf_if_specified(accelerator, args)

    if args.fused_backward_pass:
        # use fused optimizer for backward pass: other optimizers will be supported in the future
        import library.adafactor_fused

        library.adafactor_fused.patch_adafactor_fused(optimizer)

        for param_group, param_name_group in zip(optimizer.param_groups, param_names):
            for parameter, param_name in zip(param_group["params"], param_name_group):
                if parameter.requires_grad:

                    def create_grad_hook(p_name, p_group):
                        def grad_hook(tensor: torch.Tensor):
                            if accelerator.sync_gradients and args.max_grad_norm != 0.0:
                                accelerator.clip_grad_norm_(tensor, args.max_grad_norm)
                            optimizer.step_param(tensor, p_group)
                            tensor.grad = None

                        return grad_hook

                    parameter.register_post_accumulate_grad_hook(
                        create_grad_hook(param_name, param_group)
                    )

    elif args.blockwise_fused_optimizers:
        # prepare for additional optimizers and lr schedulers
        for i in range(1, len(optimizers)):
            optimizers[i] = accelerator.prepare(optimizers[i])
            lr_schedulers[i] = accelerator.prepare(lr_schedulers[i])

        # counters are used to determine when to step the optimizer
        global optimizer_hooked_count
        global num_parameters_per_group
        global parameter_optimizer_map

        optimizer_hooked_count = {}
        num_parameters_per_group = [0] * len(optimizers)
        parameter_optimizer_map = {}

        for opt_idx, optimizer in enumerate(optimizers):
            for param_group in optimizer.param_groups:
                for parameter in param_group["params"]:
                    if parameter.requires_grad:

                        def grad_hook(parameter: torch.Tensor):
                            if accelerator.sync_gradients and args.max_grad_norm != 0.0:
                                accelerator.clip_grad_norm_(
                                    parameter, args.max_grad_norm
                                )

                            i = parameter_optimizer_map[parameter]
                            optimizer_hooked_count[i] += 1
                            if optimizer_hooked_count[i] == num_parameters_per_group[i]:
                                optimizers[i].step()
                                optimizers[i].zero_grad(set_to_none=True)

                        parameter.register_post_accumulate_grad_hook(grad_hook)
                        parameter_optimizer_map[parameter] = opt_idx
                        num_parameters_per_group[opt_idx] += 1

    # 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
    # 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 examples / サンプル数: {train_dataset_group.num_train_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 = FlowMatchEulerDiscreteScheduler(
        num_train_timesteps=1000, shift=args.discrete_flow_shift
    )
    noise_scheduler_copy = copy.deepcopy(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(
            "finetuning" if args.log_tracker_name is None else args.log_tracker_name,
            config=train_util.get_sanitized_config_or_none(args),
            init_kwargs=init_kwargs,
        )

    if is_swapping_blocks:
        accelerator.unwrap_model(nextdit).prepare_block_swap_before_forward()

    # For --sample_at_first
    optimizer_eval_fn()
    lumina_train_util.sample_images(
        accelerator,
        args,
        0,
        global_step,
        nextdit,
        ae,
        gemma2,
        sample_prompts_te_outputs,
    )
    optimizer_train_fn()
    if len(accelerator.trackers) > 0:
        # log empty object to commit the sample images to wandb
        accelerator.log({}, step=0)

    loss_recorder = train_util.LossRecorder()
    epoch = 0  # avoid error when max_train_steps is 0
    for epoch in range(num_train_epochs):
        accelerator.print(f"\nepoch {epoch+1}/{num_train_epochs}")
        current_epoch.value = epoch + 1

        for m in training_models:
            m.train()

        for step, batch in enumerate(train_dataloader):
            current_step.value = global_step

            if args.blockwise_fused_optimizers:
                optimizer_hooked_count = {
                    i: 0 for i in range(len(optimizers))
                }  # reset counter for each step

            with accelerator.accumulate(*training_models):
                if "latents" in batch and batch["latents"] is not None:
                    latents = batch["latents"].to(
                        accelerator.device, dtype=weight_dtype
                    )
                else:
                    with torch.no_grad():
                        # encode images to latents. images are [-1, 1]
                        latents = ae.encode(batch["images"].to(ae.dtype)).to(
                            accelerator.device, dtype=weight_dtype
                        )

                    # 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)

                text_encoder_outputs_list = batch.get("text_encoder_outputs_list", None)
                if text_encoder_outputs_list is not None:
                    text_encoder_conds = text_encoder_outputs_list
                else:
                    # not cached or training, so get from text encoders
                    tokens_and_masks = batch["input_ids_list"]
                    with torch.no_grad():
                        input_ids = [
                            ids.to(accelerator.device)
                            for ids in batch["input_ids_list"]
                        ]
                        text_encoder_conds = text_encoding_strategy.encode_tokens(
                            lumina_tokenize_strategy,
                            [gemma2],
                            input_ids,
                        )
                        if args.full_fp16:
                            text_encoder_conds = [
                                c.to(weight_dtype) for c in text_encoder_conds
                            ]

                # TODO support some features for noise implemented in get_noise_noisy_latents_and_timesteps

                # Sample noise that we'll add to the latents
                noise = torch.randn_like(latents)

                # get noisy model input and timesteps
                noisy_model_input, timesteps, sigmas = (
                    lumina_train_util.get_noisy_model_input_and_timesteps(
                        args,
                        noise_scheduler_copy,
                        latents,
                        noise,
                        accelerator.device,
                        weight_dtype,
                    )
                )
                # call model
                gemma2_hidden_states, input_ids, gemma2_attn_mask = text_encoder_conds

                with accelerator.autocast():
                    # YiYi notes: divide it by 1000 for now because we scale it by 1000 in the transformer model (we should not keep it but I want to keep the inputs same for the model for testing)
                    model_pred = nextdit(
                        x=noisy_model_input,  # image latents (B, C, H, W)
                        t=1 - timesteps / 1000,  # timesteps需要除以1000来匹配模型预期
                        cap_feats=gemma2_hidden_states,  # Gemma2的hidden states作为caption features
                        cap_mask=gemma2_attn_mask.to(
                            dtype=torch.int32
                        ),  # Gemma2的attention mask
                    )
                # apply model prediction type
                model_pred, weighting = lumina_train_util.apply_model_prediction_type(
                    args, model_pred, noisy_model_input, sigmas
                )

                # flow matching loss
                target = latents - noise

                # calculate loss
                huber_c = train_util.get_huber_threshold_if_needed(
                    args, timesteps, noise_scheduler
                )
                loss = train_util.conditional_loss(
                    model_pred.float(), target.float(), args.loss_type, "none", huber_c
                )
                if weighting is not None:
                    loss = loss * weighting
                if args.masked_loss or (
                    "alpha_masks" in batch and batch["alpha_masks"] is not None
                ):
                    loss = apply_masked_loss(loss, batch)
                loss = loss.mean([1, 2, 3])

                loss_weights = batch["loss_weights"]  # 各sampleごとのweight
                loss = loss * loss_weights
                loss = loss.mean()

                # backward
                accelerator.backward(loss)

                if not (args.fused_backward_pass or args.blockwise_fused_optimizers):
                    if accelerator.sync_gradients and args.max_grad_norm != 0.0:
                        params_to_clip = []
                        for m in training_models:
                            params_to_clip.extend(m.parameters())
                        accelerator.clip_grad_norm_(params_to_clip, args.max_grad_norm)

                    optimizer.step()
                    lr_scheduler.step()
                    optimizer.zero_grad(set_to_none=True)
                else:
                    # optimizer.step() and optimizer.zero_grad() are called in the optimizer hook
                    lr_scheduler.step()
                    if args.blockwise_fused_optimizers:
                        for i in range(1, len(optimizers)):
                            lr_schedulers[i].step()

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

                optimizer_eval_fn()
                lumina_train_util.sample_images(
                    accelerator,
                    args,
                    None,
                    global_step,
                    nextdit,
                    ae,
                    gemma2,
                    sample_prompts_te_outputs,
                )

                # 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:
                        lumina_train_util.save_lumina_model_on_epoch_end_or_stepwise(
                            args,
                            False,
                            accelerator,
                            save_dtype,
                            epoch,
                            num_train_epochs,
                            global_step,
                            accelerator.unwrap_model(nextdit),
                        )
                optimizer_train_fn()

            current_loss = loss.detach().item()  # 平均なのでbatch sizeは関係ないはず
            if len(accelerator.trackers) > 0:
                logs = {"loss": current_loss}
                train_util.append_lr_to_logs(
                    logs, lr_scheduler, args.optimizer_type, including_unet=True
                )

                accelerator.log(logs, step=global_step)

            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]}
            progress_bar.set_postfix(**logs)

            if global_step >= args.max_train_steps:
                break

        if len(accelerator.trackers) > 0:
            logs = {"loss/epoch": loss_recorder.moving_average}
            accelerator.log(logs, step=epoch + 1)

        accelerator.wait_for_everyone()

        optimizer_eval_fn()
        if args.save_every_n_epochs is not None:
            if accelerator.is_main_process:
                lumina_train_util.save_lumina_model_on_epoch_end_or_stepwise(
                    args,
                    True,
                    accelerator,
                    save_dtype,
                    epoch,
                    num_train_epochs,
                    global_step,
                    accelerator.unwrap_model(nextdit),
                )

        lumina_train_util.sample_images(
            accelerator,
            args,
            epoch + 1,
            global_step,
            nextdit,
            ae,
            gemma2,
            sample_prompts_te_outputs,
        )
        optimizer_train_fn()

    is_main_process = accelerator.is_main_process
    # if is_main_process:
    nextdit = accelerator.unwrap_model(nextdit)

    accelerator.end_training()
    optimizer_eval_fn()

    if args.save_state or args.save_state_on_train_end:
        train_util.save_state_on_train_end(args, accelerator)

    del accelerator  # この後メモリを使うのでこれは消す

    if is_main_process:
        lumina_train_util.save_lumina_model_on_train_end(
            args, save_dtype, epoch, global_step, nextdit
        )
        logger.info("model saved.")


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

    add_logging_arguments(parser)
    train_util.add_sd_models_arguments(parser)  # TODO split this
    sai_model_spec.add_model_spec_arguments(parser)
    train_util.add_dataset_arguments(parser, True, True, True)
    train_util.add_training_arguments(parser, False)
    train_util.add_masked_loss_arguments(parser)
    deepspeed_utils.add_deepspeed_arguments(parser)
    train_util.add_sd_saving_arguments(parser)
    train_util.add_optimizer_arguments(parser)
    config_util.add_config_arguments(parser)
    add_custom_train_arguments(parser)  # TODO remove this from here
    train_util.add_dit_training_arguments(parser)
    lumina_train_util.add_lumina_train_arguments(parser)

    parser.add_argument(
        "--mem_eff_save",
        action="store_true",
        help="[EXPERIMENTAL] use memory efficient custom model saving method / メモリ効率の良い独自のモデル保存方法を使う",
    )

    parser.add_argument(
        "--fused_optimizer_groups",
        type=int,
        default=None,
        help="**this option is not working** will be removed in the future / このオプションは動作しません。将来削除されます",
    )
    parser.add_argument(
        "--blockwise_fused_optimizers",
        action="store_true",
        help="enable blockwise optimizers for fused backward pass and optimizer step / fused backward passとoptimizer step のためブロック単位のoptimizerを有効にする",
    )
    parser.add_argument(
        "--skip_latents_validity_check",
        action="store_true",
        help="[Deprecated] use 'skip_cache_check' instead / 代わりに 'skip_cache_check' を使用してください",
    )
    parser.add_argument(
        "--cpu_offload_checkpointing",
        action="store_true",
        help="[EXPERIMENTAL] enable offloading of tensors to CPU during checkpointing / チェックポイント時にテンソルをCPUにオフロードする",
    )
    return parser


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

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

    train(args)