# ================================================ # HunyuanDiT training scripts (with captions) import argparse import math import os from multiprocessing import Value from typing import List 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 diffusers import DDPMScheduler from library import deepspeed_utils, sdxl_model_util 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, ) import library.custom_train_functions as custom_train_functions from library.custom_train_functions import ( apply_snr_weight, prepare_scheduler_for_custom_training, scale_v_prediction_loss_like_noise_prediction, add_v_prediction_like_loss, apply_debiased_estimation, apply_masked_loss, ) import library.hunyuan_utils as hunyuan_utils UNET_NUM_BLOCKS_FOR_BLOCK_LR = 23 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) 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" assert ( not args.train_text_encoder or not args.cache_text_encoder_outputs ), "cache_text_encoder_outputs is not supported when training text encoder / text encoderを学習するときはcache_text_encoder_outputsはサポートされていません" if args.block_lr: block_lrs = [float(lr) for lr in args.block_lr.split(",")] assert ( len(block_lrs) == UNET_NUM_BLOCKS_FOR_BLOCK_LR ), f"block_lr must have {UNET_NUM_BLOCKS_FOR_BLOCK_LR} values / block_lrは{UNET_NUM_BLOCKS_FOR_BLOCK_LR}個の値を指定してください" else: block_lrs = None 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 tokenizer1, tokenizer2 = hunyuan_utils.load_tokenizers() # Prepare datasets 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, tokenizer=[tokenizer1, tokenizer2] ) train_dataset_group = config_util.generate_dataset_group_by_blueprint( blueprint.dataset_group ) else: train_dataset_group = train_util.load_arbitrary_dataset( args, [tokenizer1, tokenizer2] ) 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(32) if args.debug_dataset: 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" # Prepare accelerator logger.info("prepare accelerator") accelerator = train_util.prepare_accelerator(args) # Prepare types that supports mixed precision and casts as needed. weight_dtype, save_dtype = train_util.prepare_dtype(args) vae_dtype = torch.float32 if args.no_half_vae else weight_dtype # Load models ( load_stable_diffusion_format, text_encoder1, text_encoder2, vae, hydit, logit_scale, ckpt_info, ) = hunyuan_utils.load_target_model( args, accelerator, "hydit", weight_dtype, args.use_extra_cond ) if args.use_extra_cond: hydit_version = "v1.1" else: hydit_version = "v1.2" # verify load/save model formats if load_stable_diffusion_format: src_stable_diffusion_ckpt = args.pretrained_model_name_or_path src_diffusers_model_path = None else: src_stable_diffusion_ckpt = None src_diffusers_model_path = args.pretrained_model_name_or_path if args.save_model_as is None: save_stable_diffusion_format = load_stable_diffusion_format use_safetensors = args.use_safetensors else: save_stable_diffusion_format = ( args.save_model_as.lower() == "ckpt" or args.save_model_as.lower() == "safetensors" ) use_safetensors = args.use_safetensors or ( "safetensors" in args.save_model_as.lower() ) # assert save_stable_diffusion_format, "save_model_as must be ckpt or safetensors / save_model_asはckptかsafetensorsである必要があります" # Setting the flag for using Diffusers version of xformers function def set_diffusers_xformers_flag(model, valid): def fn_recursive_set_mem_eff(module: torch.nn.Module): if hasattr(module, "set_use_memory_efficient_attention_xformers"): module.set_use_memory_efficient_attention_xformers(valid) for child in module.children(): fn_recursive_set_mem_eff(child) fn_recursive_set_mem_eff(model) # Integrate xformers and memory efficient attention into the model if args.diffusers_xformers: # もうU-Netを独自にしたので動かないけどVAEのxformersは動くはず accelerator.print("Use xformers by Diffusers") # set_diffusers_xformers_flag(hydit, True) set_diffusers_xformers_flag(vae, True) else: # The Windows version of xformers may not be able to train with float, so there is a need to enable settings that don't use xformers. accelerator.print("Disable Diffusers' xformers") train_util.replace_unet_modules( hydit, args.mem_eff_attn, args.xformers, args.sdpa ) if ( torch.__version__ >= "2.0.0" ): # The following can be used with xformers compatible with PyTorch 2.0.0 and above. vae.set_use_memory_efficient_attention_xformers(args.xformers) # Prepare vae latents 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") clean_memory_on_device(accelerator.device) accelerator.wait_for_everyone() # Prepare for learning: Get the model into a proper state if args.gradient_checkpointing: hydit.enable_gradient_checkpointing() train_hydit = args.learning_rate != 0 train_text_encoder1 = False train_text_encoder2 = False if args.train_text_encoder: raise NotImplementedError( "Training text encoder is not supported yet for HunyuanDiT" ) else: text_encoder1.to(weight_dtype) text_encoder2.to(weight_dtype) text_encoder1.requires_grad_(False) text_encoder2.requires_grad_(False) text_encoder1.eval() text_encoder2.eval() # Cache the output of Textencoder if args.cache_text_encoder_outputs: raise NotImplementedError( "Caching text encoder outputs in HunyuanDiT is not supported yet" ) # TODO: We just copy the code from sdxl_train.py, need to rewrite `cache_text_encoder_outputs` # for supporting SDXL and HunyuanDiT at the same time. # Text Encodes are eval and no grad with torch.no_grad(), accelerator.autocast(): train_dataset_group.cache_text_encoder_outputs( (tokenizer1, tokenizer2), (text_encoder1, text_encoder2), accelerator.device, None, args.cache_text_encoder_outputs_to_disk, accelerator.is_main_process, ) accelerator.wait_for_everyone() if not cache_latents: vae.requires_grad_(False) vae.eval() vae.to(accelerator.device, dtype=vae_dtype) hydit.requires_grad_(train_hydit) if not train_hydit: hydit.to( accelerator.device, dtype=weight_dtype ) # because of hydit is not prepared training_models = [] params_to_optimize = [] if train_hydit: training_models.append(hydit) if block_lrs is None: params_to_optimize.append( {"params": list(hydit.parameters()), "lr": args.learning_rate} ) else: raise NotImplementedError("block_lr is not supported yet for HunyuanDiT") if train_text_encoder1: training_models.append(text_encoder1) params_to_optimize.append( { "params": list(text_encoder1.parameters()), "lr": args.learning_rate_te1 or args.learning_rate, } ) if train_text_encoder2: training_models.append(text_encoder2) params_to_optimize.append( { "params": list(text_encoder2.parameters()), "lr": args.learning_rate_te2 or args.learning_rate, } ) # 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"train hydit: {train_hydit}, text_encoder1: {train_text_encoder1}, text_encoder2: {train_text_encoder2}" ) accelerator.print(f"number of models: {len(training_models)}") accelerator.print(f"number of trainable parameters: {n_params}") # Prepare the tools necessary for training accelerator.print("prepare optimizer, data loader etc.") if args.fused_optimizer_groups: # 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 group of parameters. # This balances memory usage and management complexity. # calculate total number of parameters n_total_params = sum(len(params["params"]) for params in params_to_optimize) params_per_group = math.ceil(n_total_params / args.fused_optimizer_groups) # split params into groups, keeping the learning rate the same for all params in a group # this will increase the number of groups if the learning rate is different for different params (e.g. U-Net and text encoders) grouped_params = [] param_group = [] param_group_lr = -1 for group in params_to_optimize: lr = group["lr"] for p in group["params"]: # if the learning rate is different for different params, start a new group if lr != param_group_lr: if param_group: grouped_params.append( {"params": param_group, "lr": param_group_lr} ) param_group = [] param_group_lr = lr param_group.append(p) # if the group has enough parameters, start a new group if len(param_group) == params_per_group: grouped_params.append({"params": param_group, "lr": param_group_lr}) param_group = [] param_group_lr = -1 if param_group: grouped_params.append({"params": param_group, "lr": param_group_lr}) # 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 fused optimizer groups") else: _, _, optimizer = train_util.get_optimizer( args, trainable_params=params_to_optimize ) # Prepare the DataLoader # Note that the number of DataLoader processes: 0 cannot use persistent_workers 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, ) # Calculate the number of training steps 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}" ) # Send the training steps to the dataset side train_dataset_group.set_max_train_steps(args.max_train_steps) # Prepare a learning rate scheduler if args.fused_optimizer_groups: # 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 ) # Experimental Feature: Conducting fp16/bf16 learning, including gradients, converting the entire model to fp16/bf16. 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.") hydit.to(weight_dtype) text_encoder1.to(weight_dtype) text_encoder2.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.") hydit.to(weight_dtype) text_encoder1.to(weight_dtype) text_encoder2.to(weight_dtype) if args.deepspeed: ds_model = deepspeed_utils.prepare_deepspeed_model( args, hydit=hydit if train_hydit else None, text_encoder1=text_encoder1 if train_text_encoder1 else None, text_encoder2=text_encoder2 if train_text_encoder2 else None, ) # 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がなんかよろしくやってくれるらしい if train_hydit: hydit = accelerator.prepare(hydit) if train_text_encoder1: text_encoder1 = accelerator.prepare(text_encoder1) if train_text_encoder2: text_encoder2 = accelerator.prepare(text_encoder2) optimizer, train_dataloader, lr_scheduler = accelerator.prepare( optimizer, train_dataloader, lr_scheduler ) # TextEncoderの出力をキャッシュするときにはCPUへ移動する if args.cache_text_encoder_outputs: # move Text Encoders for sampling images. Text Encoder doesn't work on CPU with fp16 text_encoder1.to("cpu", dtype=torch.float32) text_encoder2.to("cpu", dtype=torch.float32) clean_memory_on_device(accelerator.device) else: # make sure Text Encoders are on GPU text_encoder1.to(accelerator.device) text_encoder2.to(accelerator.device) # 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 in optimizer.param_groups: for parameter in param_group["params"]: if parameter.requires_grad: def __grad_hook(tensor: torch.Tensor, param_group=param_group): if accelerator.sync_gradients and args.max_grad_norm != 0.0: accelerator.clip_grad_norm_(tensor, args.max_grad_norm) optimizer.step_param(tensor, param_group) tensor.grad = None parameter.register_post_accumulate_grad_hook(__grad_hook) elif args.fused_optimizer_groups: # 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 optimizer_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(optimizer_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 train_noise_scheduler = DDPMScheduler( beta_start=0.00085, beta_end=args.beta_end, beta_schedule="scaled_linear", num_train_timesteps=1000, clip_sample=False, steps_offset=1, ) prepare_scheduler_for_custom_training(train_noise_scheduler, accelerator.device) if args.zero_terminal_snr: custom_train_functions.fix_noise_scheduler_betas_for_zero_terminal_snr( train_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, ) # For --sample_at_first sdxl_train_util.sample_images( accelerator, args, 0, global_step, accelerator.device, vae, [tokenizer1, tokenizer2], [text_encoder1, text_encoder2], hydit, ) loss_recorder = train_util.LossRecorder() 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.fused_optimizer_groups: 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).to(dtype=weight_dtype) ) else: with torch.no_grad(): # latentに変換 latents = ( vae.encode(batch["images"].to(vae_dtype)) .latent_dist.sample() .to(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) latents = latents * sdxl_model_util.VAE_SCALE_FACTOR 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.set_grad_enabled(args.train_text_encoder): input_ids1 = input_ids1.to(accelerator.device) input_ids2 = input_ids2.to(accelerator.device) encoder_hidden_states1, mask1, encoder_hidden_states2, mask2 = ( hunyuan_utils.hunyuan_get_hidden_states( args.max_token_length, input_ids1, input_ids2, tokenizer1, tokenizer2, text_encoder1, text_encoder2, None if not args.full_fp16 else weight_dtype, accelerator=accelerator, ) ) logger.debug("encoder_hidden_states1", encoder_hidden_states1.shape) logger.debug("encoder_hidden_states2", encoder_hidden_states2.shape) else: raise NotImplementedError # 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, huber_c = ( train_util.get_noise_noisy_latents_and_timesteps( args, train_noise_scheduler, latents ) ) noisy_latents = noisy_latents.to( weight_dtype ) # TODO check why noisy_latents is not weight_dtype B, C, H, W = noisy_latents.shape if args.use_extra_cond: # get size embeddings orig_size = batch["original_sizes_hw"] crop_size = batch["crop_top_lefts"] target_size = batch["target_sizes_hw"] style = torch.as_tensor([0] * B, device=accelerator.device) image_meta_size = torch.concat([orig_size, target_size, crop_size]) else: style = None image_meta_size = None # RoPE embeddings freqs_cis_img = hunyuan_utils.calc_rope(H * 8, W * 8, 2, 88) # Predict the noise residual with accelerator.autocast(): noise_pred = hydit( noisy_latents, timesteps, encoder_hidden_states=encoder_hidden_states1, text_embedding_mask=mask1, encoder_hidden_states_t5=encoder_hidden_states2, text_embedding_mask_t5=mask2, image_meta_size=image_meta_size, style=style, cos_cis_img=freqs_cis_img[0], sin_cis_img=freqs_cis_img[1], ) # `noise_pred` has 8 channels. The first four channels are used for the noise prediction, and the # last four channels are used for the variance prediction. During inference, we found that the # predicted variance has imperceptible affect on the quality of the generated images. Therefore, we # only use the first four channels for the noise prediction. See the following link for details. # https://github.com/Tencent/HunyuanDiT/blob/5657364143e44ac90f72aeb47b81bd505a95665d/hydit/diffusion/gaussian_diffusion.py#L562 noise_pred, _ = noise_pred.chunk(2, dim=1) if args.v_parameterization: # v-parameterization training target = train_noise_scheduler.get_velocity( latents, noise, timesteps ) else: target = noise if ( args.min_snr_gamma or args.scale_v_pred_loss_like_noise_pred or args.v_pred_like_loss or args.debiased_estimation_loss or args.masked_loss ): # do not mean over batch dimension for snr weight or scale v-pred loss loss = train_util.conditional_loss( noise_pred.float(), target.float(), reduction="none", loss_type=args.loss_type, huber_c=huber_c, ) 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]) if args.min_snr_gamma: loss = apply_snr_weight( loss, timesteps, train_noise_scheduler, args.min_snr_gamma ) if args.scale_v_pred_loss_like_noise_pred: loss = scale_v_prediction_loss_like_noise_prediction( loss, timesteps, train_noise_scheduler ) if args.v_pred_like_loss: loss = add_v_prediction_like_loss( loss, timesteps, train_noise_scheduler, args.v_pred_like_loss, ) if args.debiased_estimation_loss: loss = apply_debiased_estimation( loss, timesteps, train_noise_scheduler ) loss = loss.mean() # mean over batch dimension else: loss = train_util.conditional_loss( noise_pred.float(), target.float(), reduction="mean", loss_type=args.loss_type, huber_c=huber_c, ) accelerator.backward(loss) if not (args.fused_backward_pass or args.fused_optimizer_groups): 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.fused_optimizer_groups: 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 sdxl_train_util.sample_images( accelerator, args, None, global_step, accelerator.device, vae, [tokenizer1, tokenizer2], [text_encoder1, text_encoder2], hydit, ) # 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: src_path = ( src_stable_diffusion_ckpt if save_stable_diffusion_format else src_diffusers_model_path ) hunyuan_utils.save_hydit_model_on_epoch_end_or_stepwise( args, False, accelerator, src_path, save_stable_diffusion_format, use_safetensors, save_dtype, epoch, num_train_epochs, global_step, accelerator.unwrap_model(text_encoder1), accelerator.unwrap_model(text_encoder2), accelerator.unwrap_model(hydit), vae, logit_scale, ckpt_info, hydit_version, ) current_loss = loss.detach().item() # 平均なのでbatch sizeは関係ないはず if args.logging_dir is not None: logs = {"loss": current_loss} if block_lrs is None: train_util.append_lr_to_logs( logs, lr_scheduler, args.optimizer_type, including_unet=train_hydit, ) else: append_block_lr_to_logs( block_lrs, logs, lr_scheduler, args.optimizer_type ) # U-Net is included in block_lrs 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 args.logging_dir is not None: logs = {"loss/epoch": loss_recorder.moving_average} accelerator.log(logs, step=epoch + 1) accelerator.wait_for_everyone() if args.save_every_n_epochs is not None: if accelerator.is_main_process: src_path = ( src_stable_diffusion_ckpt if save_stable_diffusion_format else src_diffusers_model_path ) hunyuan_utils.save_hydit_model_on_epoch_end_or_stepwise( args, True, accelerator, src_path, save_stable_diffusion_format, use_safetensors, save_dtype, epoch, num_train_epochs, global_step, accelerator.unwrap_model(text_encoder1), accelerator.unwrap_model(text_encoder2), accelerator.unwrap_model(hydit), vae, logit_scale, ckpt_info, hydit_version, ) sdxl_train_util.sample_images( accelerator, args, epoch + 1, global_step, accelerator.device, vae, [tokenizer1, tokenizer2], [text_encoder1, text_encoder2], hydit, ) is_main_process = accelerator.is_main_process # if is_main_process: hydit = accelerator.unwrap_model(hydit) text_encoder1 = accelerator.unwrap_model(text_encoder1) text_encoder2 = accelerator.unwrap_model(text_encoder2) accelerator.end_training() 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: src_path = ( src_stable_diffusion_ckpt if save_stable_diffusion_format else src_diffusers_model_path ) hunyuan_utils.save_hydit_model_on_train_end( args, src_path, save_stable_diffusion_format, use_safetensors, save_dtype, epoch, global_step, text_encoder1, text_encoder2, hydit, vae, logit_scale, ckpt_info, hydit_version, ) logger.info("model saved.") def setup_parser() -> argparse.ArgumentParser: parser = argparse.ArgumentParser() add_logging_arguments(parser) train_util.add_sd_models_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) custom_train_functions.add_custom_train_arguments(parser) sdxl_train_util.add_sdxl_training_arguments(parser) hunyuan_utils.add_hydit_arguments(parser) parser.add_argument( "--learning_rate_te1", type=float, default=None, help="learning rate for text encoder 1 (CLIP) / text encoder 1 (ViT-L)の学習率", ) parser.add_argument( "--learning_rate_te2", type=float, default=None, help="learning rate for text encoder 2 (mT5) / text encoder 2 (BiG-G)の学習率", ) parser.add_argument( "--diffusers_xformers", action="store_true", help="use xformers by diffusers / Diffusersでxformersを使用する", ) parser.add_argument( "--train_text_encoder", action="store_true", help="train text encoder / text encoderも学習する", ) 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( "--block_lr", type=str, default=None, help=f"learning rates for each block of HunyuanDiT, comma-separated, {UNET_NUM_BLOCKS_FOR_BLOCK_LR} values / " + f"U-Netの各ブロックの学習率、カンマ区切り、{UNET_NUM_BLOCKS_FOR_BLOCK_LR}個の値", ) parser.add_argument( "--fused_optimizer_groups", type=int, default=None, help="number of optimizers for fused backward pass and optimizer step / fused backward passとoptimizer stepのためのoptimizer数", ) return parser if __name__ == "__main__": parser = setup_parser() args = parser.parse_args() print(args) train_util.verify_command_line_training_args(args) args = train_util.read_config_from_file(args, parser) train(args)