train_lora.py

"""Modified from https://github.com/huggingface/diffusers/blob/main/examples/text_to_image/train_text_to_image.py
"""
#!/usr/bin/env python
# coding=utf-8
# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and

import argparse
import copy
import gc
import logging
import math
import os
import pickle
import shutil
import sys

import accelerate
import diffusers
import numpy as np
import torch
import torch.nn.functional as F
import torch.utils.checkpoint
import transformers
from accelerate import Accelerator
from accelerate.logging import get_logger
from accelerate.state import AcceleratorState
from accelerate.utils import ProjectConfiguration, set_seed
from diffusers import (DDIMScheduler, DDPMScheduler,
                       FlowMatchEulerDiscreteScheduler)
from diffusers.optimization import get_scheduler
from diffusers.training_utils import (EMAModel,
                                      _set_state_dict_into_text_encoder,
                                      cast_training_params,
                                      compute_density_for_timestep_sampling,
                                      compute_loss_weighting_for_sd3)
from diffusers.utils import check_min_version, deprecate, is_wandb_available
from diffusers.utils.import_utils import is_xformers_available
from diffusers.utils.torch_utils import is_compiled_module
from einops import rearrange
from huggingface_hub import create_repo, upload_folder
from omegaconf import OmegaConf
from packaging import version
from PIL import Image
from torch.utils.data import RandomSampler
from torch.utils.tensorboard import SummaryWriter
from torchvision import transforms
from tqdm.auto import tqdm
from transformers import (Qwen2Tokenizer, AutoTokenizer, BertModel,
                          BertTokenizer, CLIPImageProcessor,
                          CLIPVisionModelWithProjection,
                          Qwen2VLForConditionalGeneration, T5EncoderModel,
                          T5Tokenizer)
from transformers.utils import ContextManagers

import datasets

current_file_path = os.path.abspath(__file__)
project_roots = [os.path.dirname(current_file_path), os.path.dirname(os.path.dirname(current_file_path))]
for project_root in project_roots:
    sys.path.insert(0, project_root) if project_root not in sys.path else None

from easyanimate.data.bucket_sampler import (ASPECT_RATIO_512,
                                             ASPECT_RATIO_RANDOM_CROP_512,
                                             ASPECT_RATIO_RANDOM_CROP_PROB,
                                             AspectRatioBatchImageVideoSampler,
                                             RandomSampler, get_closest_ratio)
from easyanimate.data.dataset_image_video import (ImageVideoDataset,
                                                  ImageVideoSampler,
                                                  get_random_mask)
from easyanimate.models import (name_to_autoencoder_magvit,
                                name_to_transformer3d)
from easyanimate.pipeline.pipeline_easyanimate import (
    EasyAnimatePipeline, get_2d_rotary_pos_embed, get_3d_rotary_pos_embed,
    get_resize_crop_region_for_grid)
from easyanimate.pipeline.pipeline_easyanimate_inpaint import (
    EasyAnimateInpaintPipeline, add_noise_to_reference_video, resize_mask)
from easyanimate.utils import gaussian_diffusion as gd
from easyanimate.utils.discrete_sampler import DiscreteSampling
from easyanimate.utils.lora_utils import (create_network, merge_lora,
                                          unmerge_lora)
from easyanimate.utils.respace import SpacedDiffusion, space_timesteps
from easyanimate.utils.utils import get_image_to_video_latent, save_videos_grid

if is_wandb_available():
    import wandb

rotary_pos_embed_cache = {}

def _get_2d_rotary_pos_embed_cached(embed_dim, crops_coords, grid_size):
    tmp_key = (embed_dim, crops_coords, grid_size)
    print('embed_dim=%d crops_coords=%s grid_size=%s in_cache=%d cache_size=%d' % (
        embed_dim, crops_coords, grid_size, tmp_key in rotary_pos_embed_cache, len(rotary_pos_embed_cache)))
    if tmp_key not in rotary_pos_embed_cache:
        tmp_embed = get_2d_rotary_pos_embed(embed_dim, crops_coords, grid_size)
        tmp_embed_gpu = (tmp_embed[0].cuda(), tmp_embed[1].cuda())
        print('\tembed_dim=%d data_size=%s' % (embed_dim, tmp_embed[0].shape))
        rotary_pos_embed_cache[tmp_key] = tmp_embed_gpu
        return tmp_embed_gpu
    else:
        return rotary_pos_embed_cache[tmp_key]

def _get_3d_rotary_pos_embed_cached(embed_dim, crops_coords, grid_size, temporal_size):
    tmp_key = (embed_dim, crops_coords, grid_size, temporal_size)
    print('embed_dim=%d crops_coords=%s grid_size=%s in_cache=%d cache_size=%d' % (
        embed_dim, crops_coords, grid_size, tmp_key in rotary_pos_embed_cache, len(rotary_pos_embed_cache)))
    if tmp_key not in rotary_pos_embed_cache:
        tmp_embed = get_3d_rotary_pos_embed(
            embed_dim, crops_coords, grid_size=grid_size,
            temporal_size=temporal_size, use_real=True,
        )
        tmp_embed_gpu = (tmp_embed[0].cuda(), tmp_embed[1].cuda())
        print('\tembed_dim=%d data_size=%s' % (embed_dim, tmp_embed[0].shape))
        rotary_pos_embed_cache[tmp_key] = tmp_embed_gpu
        return tmp_embed_gpu
    else:
        return rotary_pos_embed_cache[tmp_key]

def encode_prompt(
    tokenizer, 
    text_encoder, 
    prompt: str,
    device: torch.device,
    dtype: torch.dtype,
    max_sequence_length = None,
    tokenizer_max_length = 256,
    add_special_tokens = False,
    enable_text_attention_mask = True,
):
    if type(tokenizer) in [BertTokenizer, T5Tokenizer]:
        if max_sequence_length is None:
            max_length = min(tokenizer.model_max_length, tokenizer_max_length)
        else:
            max_length = max_sequence_length

        text_inputs = tokenizer(
            prompt,
            padding="max_length",
            max_length=max_length,
            truncation=True,
            return_attention_mask=True,
            add_special_tokens=add_special_tokens,
            return_tensors="pt",
        )
        if device is not None:
            text_input_ids = text_inputs.input_ids.to(device)
            prompt_attention_mask = text_inputs.attention_mask.to(device)
        else:
            text_input_ids = text_inputs.input_ids
            prompt_attention_mask = text_inputs.attention_mask
        
        if enable_text_attention_mask:
            prompt_embeds = text_encoder(
                text_input_ids,
                attention_mask=prompt_attention_mask,
            )[0]
        else:
            prompt_embeds = text_encoder(
                text_input_ids
            )[0]
        prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
        prompt_attention_mask = prompt_attention_mask.to(dtype=dtype, device=device)
    else:
        max_length = tokenizer_max_length
        texts = []
        for _prompt in prompt:
            messages = [
                {
                    "role": "user",
                    "content": [{"type": "text", "text": _prompt}],
                } 
            ]
            text = tokenizer.apply_chat_template(
                messages, tokenize=False, add_generation_prompt=True
            )
            texts.append(text)
        text_inputs = tokenizer(
            text=texts,
            images=None,
            videos=None,
            padding="max_length",
            max_length=max_length,
            truncation=True,
            return_attention_mask=True,
            padding_side="right",
            return_tensors="pt",
        )
        text_inputs = text_inputs.to(text_encoder.device)

        text_input_ids = text_inputs.input_ids
        prompt_attention_mask = text_inputs.attention_mask
        if enable_text_attention_mask:
            # Inference: Generation of the output
            prompt_embeds = text_encoder(
                input_ids=text_input_ids,
                attention_mask=prompt_attention_mask,
                output_hidden_states=True).hidden_states[-2]
        else:
            raise ValueError("LLM needs attention_mask")
    return prompt_embeds, prompt_attention_mask

def get_random_downsample_ratio(sample_size, image_ratio=[], all_choices=False, rng=None):
    def _create_special_list(length):
        if length == 1:
            return [1.0]
        if length >= 2:
            first_element = 0.75
            remaining_sum = 1.0 - first_element
            other_elements_value = remaining_sum / (length - 1)
            special_list = [first_element] + [other_elements_value] * (length - 1)
            return special_list
            
    if sample_size >= 1536:
        number_list = [1, 1.25, 1.5, 2, 2.5, 3] + image_ratio 
    elif sample_size >= 1024:
        number_list = [1, 1.25, 1.5, 2] + image_ratio
    elif sample_size >= 768:
        number_list = [1, 1.25, 1.5] + image_ratio
    elif sample_size >= 512:
        number_list = [1] + image_ratio
    else:
        number_list = [1]

    if all_choices:
        return number_list

    number_list_prob = np.array(_create_special_list(len(number_list)))
    if rng is None:
        return np.random.choice(number_list, p = number_list_prob)
    else:
        return rng.choice(number_list, p = number_list_prob)

# Will error if the minimal version of diffusers is not installed. Remove at your own risks.
check_min_version("0.18.0.dev0")

logger = get_logger(__name__, log_level="INFO")

def log_validation(
    vae, text_encoder, text_encoder_2, tokenizer, tokenizer_2, transformer3d, image_encoder, image_processor, network, 
    config, args, accelerator, weight_dtype, global_step
):
    try:
        logger.info("Running validation... ")

        # Get New Transformer
        Choosen_Transformer3DModel = name_to_transformer3d[
            config['transformer_additional_kwargs'].get('transformer_type', 'Transformer3DModel')
        ]

        transformer3d_val = Choosen_Transformer3DModel.from_pretrained_2d(
            args.pretrained_model_name_or_path, subfolder="transformer",
            transformer_additional_kwargs=OmegaConf.to_container(config['transformer_additional_kwargs'])
        ).to(weight_dtype)
        transformer3d_val.load_state_dict(accelerator.unwrap_model(transformer3d).state_dict())
        
        if args.loss_type == "flow":
            scheduler = FlowMatchEulerDiscreteScheduler.from_pretrained(
                args.pretrained_model_name_or_path, 
                subfolder="scheduler"
            )
        else:
            scheduler = DDIMScheduler.from_pretrained(
                args.pretrained_model_name_or_path, 
                subfolder="scheduler"
            )

        if args.train_mode != "normal":
            pipeline = EasyAnimateInpaintPipeline(
                vae=accelerator.unwrap_model(vae).to(weight_dtype), 
                text_encoder=accelerator.unwrap_model(text_encoder),
                text_encoder_2=accelerator.unwrap_model(text_encoder_2),
                tokenizer=tokenizer,
                tokenizer_2=tokenizer_2,
                transformer=transformer3d_val,
                scheduler=scheduler,
                clip_image_encoder=image_encoder,
                clip_image_processor=image_processor,
            )
        else:
            pipeline = EasyAnimatePipeline(
                vae=accelerator.unwrap_model(vae).to(weight_dtype), 
                text_encoder=accelerator.unwrap_model(text_encoder),
                text_encoder_2=accelerator.unwrap_model(text_encoder_2),
                tokenizer=tokenizer,
                tokenizer_2=tokenizer_2,
                transformer=transformer3d_val,
                scheduler=scheduler,
            )
        pipeline = pipeline.to(weight_dtype, accelerator.device)
        pipeline = merge_lora(
            pipeline, None, 1, accelerator.device, state_dict=accelerator.unwrap_model(network).state_dict(), transformer_only=True
        )

        if args.enable_xformers_memory_efficient_attention \
            and config['transformer_additional_kwargs'].get('transformer_type', 'Transformer3DModel') == 'Transformer3DModel':
            pipeline.enable_xformers_memory_efficient_attention()

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

        for i in range(len(args.validation_prompts)):
            with torch.no_grad():
                if args.train_mode != "normal":
                    with torch.autocast("cuda", dtype=weight_dtype):
                        if vae.cache_mag_vae:
                            video_length = int((args.video_sample_n_frames - 1) // vae.mini_batch_encoder * vae.mini_batch_encoder) + 1 if args.video_sample_n_frames != 1 else 1
                        else:
                            video_length = int(args.video_sample_n_frames // vae.mini_batch_encoder * vae.mini_batch_encoder) if args.video_sample_n_frames != 1 else 1
                        input_video, input_video_mask, clip_image = get_image_to_video_latent(None, None, video_length=video_length, sample_size=[args.video_sample_size, args.video_sample_size])
                        sample = pipeline(
                            args.validation_prompts[i], 
                            video_length = video_length,
                            negative_prompt = "bad detailed",
                            height      = args.video_sample_size,
                            width       = args.video_sample_size,
                            generator   = generator, 

                            video        = input_video,
                            mask_video   = input_video_mask,
                            clip_image   = clip_image, 
                        ).frames
                        os.makedirs(os.path.join(args.output_dir, "sample"), exist_ok=True)
                        save_videos_grid(sample, os.path.join(args.output_dir, f"sample/sample-{global_step}-{i}.gif"))

                        video_length = 1
                        input_video, input_video_mask, clip_image = get_image_to_video_latent(None, None, video_length=video_length, sample_size=[args.video_sample_size, args.video_sample_size])
                        sample = pipeline(
                            args.validation_prompts[i], 
                            video_length = 1,
                            negative_prompt = "bad detailed",
                            height      = args.video_sample_size,
                            width       = args.video_sample_size,
                            generator   = generator, 

                            video        = input_video,
                            mask_video   = input_video_mask,
                            clip_image   = clip_image, 
                        ).frames
                        os.makedirs(os.path.join(args.output_dir, "sample"), exist_ok=True)
                        save_videos_grid(sample, os.path.join(args.output_dir, f"sample/sample-{global_step}-image-{i}.gif"))
                else:
                    with torch.autocast("cuda", dtype=weight_dtype):
                        sample = pipeline(
                            args.validation_prompts[i], 
                            video_length = video_length,
                            negative_prompt = "bad detailed",
                            height      = args.video_sample_size,
                            width       = args.video_sample_size,
                            generator   = generator
                        ).frames
                        os.makedirs(os.path.join(args.output_dir, "sample"), exist_ok=True)
                        save_videos_grid(sample, os.path.join(args.output_dir, f"sample/sample-{global_step}-{i}.gif"))

                        sample = pipeline(
                            args.validation_prompts[i], 
                            video_length = 1,
                            negative_prompt = "bad detailed",
                            height      = args.video_sample_size,
                            width       = args.video_sample_size,
                            generator   = generator
                        ).frames
                        os.makedirs(os.path.join(args.output_dir, "sample"), exist_ok=True)
                        save_videos_grid(sample, os.path.join(args.output_dir, f"sample/sample-{global_step}-image-{i}.gif"))

        del pipeline
        del transformer3d_val
        gc.collect()
        torch.cuda.empty_cache()
        torch.cuda.ipc_collect()

        return None
    except Exception as e:
        gc.collect()
        torch.cuda.empty_cache()
        torch.cuda.ipc_collect()
        print(f"Eval error with info {e}")
        return None

def linear_decay(initial_value, final_value, total_steps, current_step):
    if current_step >= total_steps:
        return final_value
    current_step = max(0, current_step)
    step_size = (final_value - initial_value) / total_steps
    current_value = initial_value + step_size * current_step
    return current_value

def generate_timestep_with_lognorm(low, high, shape, device="cpu", generator=None):
    u = torch.normal(mean=0.0, std=1.0, size=shape, device=device, generator=generator)
    t = 1 / (1 + torch.exp(-u)) * (high - low) + low
    return torch.clip(t.to(torch.int32), low, high - 1)


def parse_args():
    parser = argparse.ArgumentParser(description="Simple example of a training script.")
    parser.add_argument(
        "--input_perturbation", type=float, default=0, help="The scale of input perturbation. Recommended 0.1."
    )
    parser.add_argument(
        "--pretrained_model_name_or_path",
        type=str,
        default=None,
        required=True,
        help="Path to pretrained model or model identifier from huggingface.co/models.",
    )
    parser.add_argument(
        "--revision",
        type=str,
        default=None,
        required=False,
        help="Revision of pretrained model identifier from huggingface.co/models.",
    )
    parser.add_argument(
        "--variant",
        type=str,
        default=None,
        help="Variant of the model files of the pretrained model identifier from huggingface.co/models, 'e.g.' fp16",
    )
    parser.add_argument(
        "--train_data_dir",
        type=str,
        default=None,
        help=(
            "A folder containing the training data. "
        ),
    )
    parser.add_argument(
        "--train_data_meta",
        type=str,
        default=None,
        help=(
            "A csv containing the training data. "
        ),
    )
    parser.add_argument(
        "--max_train_samples",
        type=int,
        default=None,
        help=(
            "For debugging purposes or quicker training, truncate the number of training examples to this "
            "value if set."
        ),
    )
    parser.add_argument(
        "--validation_prompts",
        type=str,
        default=None,
        nargs="+",
        help=("A set of prompts evaluated every `--validation_epochs` and logged to `--report_to`."),
    )
    parser.add_argument(
        "--output_dir",
        type=str,
        default="sd-model-finetuned",
        help="The output directory where the model predictions and checkpoints will be written.",
    )
    parser.add_argument(
        "--cache_dir",
        type=str,
        default=None,
        help="The directory where the downloaded models and datasets will be stored.",
    )
    parser.add_argument("--seed", type=int, default=None, help="A seed for reproducible training.")
    parser.add_argument(
        "--random_flip",
        action="store_true",
        help="whether to randomly flip images horizontally",
    )
    parser.add_argument(
        "--use_came",
        action="store_true",
        help="whether to use came",
    )
    parser.add_argument(
        "--multi_stream",
        action="store_true",
        help="whether to use cuda multi-stream",
    )
    parser.add_argument(
        "--train_batch_size", type=int, default=16, help="Batch size (per device) for the training dataloader."
    )
    parser.add_argument(
        "--vae_mini_batch", type=int, default=32, help="mini batch size for vae."
    )
    parser.add_argument("--num_train_epochs", type=int, default=100)
    parser.add_argument(
        "--max_train_steps",
        type=int,
        default=None,
        help="Total number of training steps to perform.  If provided, overrides num_train_epochs.",
    )
    parser.add_argument(
        "--gradient_accumulation_steps",
        type=int,
        default=1,
        help="Number of updates steps to accumulate before performing a backward/update pass.",
    )
    parser.add_argument(
        "--gradient_checkpointing",
        action="store_true",
        help="Whether or not to use gradient checkpointing to save memory at the expense of slower backward pass.",
    )
    parser.add_argument(
        "--learning_rate",
        type=float,
        default=1e-4,
        help="Initial learning rate (after the potential warmup period) to use.",
    )
    parser.add_argument(
        "--scale_lr",
        action="store_true",
        default=False,
        help="Scale the learning rate by the number of GPUs, gradient accumulation steps, and batch size.",
    )
    parser.add_argument(
        "--lr_scheduler",
        type=str,
        default="constant",
        help=(
            'The scheduler type to use. Choose between ["linear", "cosine", "cosine_with_restarts", "polynomial",'
            ' "constant", "constant_with_warmup"]'
        ),
    )
    parser.add_argument(
        "--lr_warmup_steps", type=int, default=500, help="Number of steps for the warmup in the lr scheduler."
    )
    parser.add_argument(
        "--use_8bit_adam", action="store_true", help="Whether or not to use 8-bit Adam from bitsandbytes."
    )
    parser.add_argument(
        "--allow_tf32",
        action="store_true",
        help=(
            "Whether or not to allow TF32 on Ampere GPUs. Can be used to speed up training. For more information, see"
            " https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices"
        ),
    )
    parser.add_argument("--use_ema", action="store_true", help="Whether to use EMA model.")
    parser.add_argument(
        "--non_ema_revision",
        type=str,
        default=None,
        required=False,
        help=(
            "Revision of pretrained non-ema model identifier. Must be a branch, tag or git identifier of the local or"
            " remote repository specified with --pretrained_model_name_or_path."
        ),
    )
    parser.add_argument(
        "--dataloader_num_workers",
        type=int,
        default=0,
        help=(
            "Number of subprocesses to use for data loading. 0 means that the data will be loaded in the main process."
        ),
    )
    parser.add_argument("--adam_beta1", type=float, default=0.9, help="The beta1 parameter for the Adam optimizer.")
    parser.add_argument("--adam_beta2", type=float, default=0.999, help="The beta2 parameter for the Adam optimizer.")
    parser.add_argument("--adam_weight_decay", type=float, default=1e-2, help="Weight decay to use.")
    parser.add_argument("--adam_epsilon", type=float, default=1e-08, help="Epsilon value for the Adam optimizer")
    parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.")
    parser.add_argument("--push_to_hub", action="store_true", help="Whether or not to push the model to the Hub.")
    parser.add_argument("--hub_token", type=str, default=None, help="The token to use to push to the Model Hub.")
    parser.add_argument(
        "--prediction_type",
        type=str,
        default=None,
        help="The prediction_type that shall be used for training. Choose between 'epsilon' or 'v_prediction' or leave `None`. If left to `None` the default prediction type of the scheduler: `noise_scheduler.config.prediciton_type` is chosen.",
    )
    parser.add_argument(
        "--hub_model_id",
        type=str,
        default=None,
        help="The name of the repository to keep in sync with the local `output_dir`.",
    )
    parser.add_argument(
        "--logging_dir",
        type=str,
        default="logs",
        help=(
            "[TensorBoard](https://www.tensorflow.org/tensorboard) log directory. Will default to"
            " *output_dir/runs/**CURRENT_DATETIME_HOSTNAME***."
        ),
    )
    parser.add_argument(
        "--report_model_info", action="store_true", help="Whether or not to report more info about model (such as norm, grad)."
    )
    parser.add_argument(
        "--mixed_precision",
        type=str,
        default=None,
        choices=["no", "fp16", "bf16"],
        help=(
            "Whether to use mixed precision. Choose between fp16 and bf16 (bfloat16). Bf16 requires PyTorch >="
            " 1.10.and an Nvidia Ampere GPU.  Default to the value of accelerate config of the current system or the"
            " flag passed with the `accelerate.launch` command. Use this argument to override the accelerate config."
        ),
    )
    parser.add_argument(
        "--report_to",
        type=str,
        default="tensorboard",
        help=(
            'The integration to report the results and logs to. Supported platforms are `"tensorboard"`'
            ' (default), `"wandb"` and `"comet_ml"`. Use `"all"` to report to all integrations.'
        ),
    )
    parser.add_argument("--local_rank", type=int, default=-1, help="For distributed training: local_rank")
    parser.add_argument(
        "--checkpointing_steps",
        type=int,
        default=500,
        help=(
            "Save a checkpoint of the training state every X updates. These checkpoints are only suitable for resuming"
            " training using `--resume_from_checkpoint`."
        ),
    )
    parser.add_argument(
        "--checkpoints_total_limit",
        type=int,
        default=None,
        help=("Max number of checkpoints to store."),
    )
    parser.add_argument(
        "--resume_from_checkpoint",
        type=str,
        default=None,
        help=(
            "Whether training should be resumed from a previous checkpoint. Use a path saved by"
            ' `--checkpointing_steps`, or `"latest"` to automatically select the last available checkpoint.'
        ),
    )
    parser.add_argument(
        "--enable_xformers_memory_efficient_attention", action="store_true", help="Whether or not to use xformers."
    )
    parser.add_argument("--noise_offset", type=float, default=0, help="The scale of noise offset.")
    parser.add_argument(
        "--validation_epochs",
        type=int,
        default=5,
        help="Run validation every X epochs.",
    )
    parser.add_argument(
        "--validation_steps",
        type=int,
        default=2000,
        help="Run validation every X steps.",
    )
    parser.add_argument(
        "--tracker_project_name",
        type=str,
        default="text2image-fine-tune",
        help=(
            "The `project_name` argument passed to Accelerator.init_trackers for"
            " more information see https://huggingface.co/docs/accelerate/v0.17.0/en/package_reference/accelerator#accelerate.Accelerator"
        ),
    )
    
    parser.add_argument(
        "--rank",
        type=int,
        default=128,
        help=("The dimension of the LoRA update matrices."),
    )
    parser.add_argument(
        "--network_alpha",
        type=int,
        default=64,
        help=("The dimension of the LoRA update matrices."),
    )
    parser.add_argument(
        "--train_text_encoder",
        action="store_true",
        help="Whether to train the text encoder. If set, the text encoder should be float32 precision.",
    )
    parser.add_argument(
        "--snr_loss", action="store_true", help="Whether or not to use snr_loss."
    )
    parser.add_argument(
        "--uniform_sampling", action="store_true", help="Whether or not to use uniform_sampling."
    )
    parser.add_argument(
        "--loss_type", 
        type=str,
        default="sigma",
        help=(
            'The format of training data. Support `"sigma"`'
            ' (default), `"ddpm"`, `"flow"`.'
        ),
    )
    parser.add_argument(
        "--enable_text_encoder_in_dataloader", action="store_true", help="Whether or not to use text encoder in dataloader."
    )
    parser.add_argument(
        "--enable_bucket", action="store_true", help="Whether enable bucket sample in datasets."
    )
    parser.add_argument(
        "--random_ratio_crop", action="store_true", help="Whether enable random ratio crop sample in datasets."
    )
    parser.add_argument(
        "--random_frame_crop", action="store_true", help="Whether enable random frame crop sample in datasets."
    )
    parser.add_argument(
        "--random_hw_adapt", action="store_true", help="Whether enable random adapt height and width in datasets."
    )
    parser.add_argument(
        "--training_with_video_token_length", action="store_true", help="The training stage of the model in training.",
    )
    parser.add_argument(
        "--noise_share_in_frames", action="store_true", help="Whether enable noise share in frames."
    )
    parser.add_argument(
        "--noise_share_in_frames_ratio", type=float, default=0.5, help="Noise share ratio.",
    )
    parser.add_argument(
        "--motion_sub_loss", action="store_true", help="Whether enable motion sub loss."
    )
    parser.add_argument(
        "--motion_sub_loss_ratio", type=float, default=0.25, help="The ratio of motion sub loss."
    )
    parser.add_argument(
        "--keep_all_node_same_token_length",
        action="store_true", 
        help="Reference of the length token.",
    )
    parser.add_argument(
        "--train_sampling_steps",
        type=int,
        default=1000,
        help="Run train_sampling_steps.",
    )
    parser.add_argument(
        "--token_sample_size",
        type=int,
        default=512,
        help="Sample size of the token.",
    )
    parser.add_argument(
        "--video_sample_size",
        type=int,
        default=512,
        help="Sample size of the video.",
    )
    parser.add_argument(
        "--image_sample_size",
        type=int,
        default=512,
        help="Sample size of the video.",
    )
    parser.add_argument(
        "--video_sample_stride",
        type=int,
        default=4,
        help="Sample stride of the video.",
    )
    parser.add_argument(
        "--video_sample_n_frames",
        type=int,
        default=17,
        help="Num frame of video.",
    )
    parser.add_argument(
        "--video_repeat",
        type=int,
        default=0,
        help="Num of repeat video.",
    )
    parser.add_argument(
        "--image_repeat_in_forward",
        type=int,
        default=0,
        help="Num of repeat image in forward.",
    )
    parser.add_argument(
        "--config_path",
        type=str,
        default=None,
        help=(
            "The config of the model in training."
        ),
    )
    parser.add_argument(
        "--transformer_path",
        type=str,
        default=None,
        help=("If you want to load the weight from other transformers, input its path."),
    )
    parser.add_argument(
        "--vae_path",
        type=str,
        default=None,
        help=("If you want to load the weight from other vaes, input its path."),
    )
    parser.add_argument("--save_state", action="store_true", help="Whether or not to save state.")

    parser.add_argument(
        '--tokenizer_max_length', 
        type=int,
        default=256,
        help='Max length of tokenizer'
    )
    parser.add_argument(
        "--use_deepspeed", action="store_true", help="Whether or not to use deepspeed."
    )
    parser.add_argument(
        "--low_vram", action="store_true", help="Whether enable low_vram mode."
    )
    parser.add_argument(
        "--train_mode",
        type=str,
        default="normal",
        help=(
            'The format of training data. Support `"normal"`'
            ' (default), `"inpaint"`.'
        ),
    )
    parser.add_argument(
        "--abnormal_norm_clip_start",
        type=int,
        default=1000,
        help=(
            'When do we start doing additional processing on abnormal gradients. '
        ),
    )
    parser.add_argument(
        "--initial_grad_norm_ratio",
        type=int,
        default=5,
        help=(
            'The initial gradient is relative to the multiple of the max_grad_norm. '
        ),
    )

    parser.add_argument(
        "--weighting_scheme",
        type=str,
        default="none",
        choices=["sigma_sqrt", "logit_normal", "mode", "cosmap", "none"],
        help=('We default to the "none" weighting scheme for uniform sampling and uniform loss'),
    )
    parser.add_argument(
        "--logit_mean", type=float, default=0.0, help="mean to use when using the `'logit_normal'` weighting scheme."
    )
    parser.add_argument(
        "--logit_std", type=float, default=1.0, help="std to use when using the `'logit_normal'` weighting scheme."
    )
    parser.add_argument(
        "--mode_scale",
        type=float,
        default=1.29,
        help="Scale of mode weighting scheme. Only effective when using the `'mode'` as the `weighting_scheme`.",
    )

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

    # default to using the same revision for the non-ema model if not specified
    if args.non_ema_revision is None:
        args.non_ema_revision = args.revision

    return args


def main():
    args = parse_args()

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

    if args.non_ema_revision is not None:
        deprecate(
            "non_ema_revision!=None",
            "0.15.0",
            message=(
                "Downloading 'non_ema' weights from revision branches of the Hub is deprecated. Please make sure to"
                " use `--variant=non_ema` instead."
            ),
        )
    logging_dir = os.path.join(args.output_dir, args.logging_dir)

    config = OmegaConf.load(args.config_path)
    accelerator_project_config = ProjectConfiguration(project_dir=args.output_dir, logging_dir=logging_dir)

    accelerator = Accelerator(
        gradient_accumulation_steps=args.gradient_accumulation_steps,
        mixed_precision=args.mixed_precision,
        log_with=args.report_to,
        project_config=accelerator_project_config,
    )
    if accelerator.is_main_process:
        writer = SummaryWriter(log_dir=logging_dir)

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

    # If passed along, set the training seed now.
    if args.seed is not None:
        set_seed(args.seed)
        rng = np.random.default_rng(np.random.PCG64(args.seed + accelerator.process_index))
        torch_rng = torch.Generator(accelerator.device).manual_seed(args.seed + accelerator.process_index)
    else:
        rng = None
        torch_rng = None
    index_rng = np.random.default_rng(np.random.PCG64(43))
    print(f"Init rng with seed {args.seed + accelerator.process_index}. Process_index is {accelerator.process_index}")

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

    # For mixed precision training we cast all non-trainable weigths (vae, non-lora text_encoder and non-lora transformer3d) to half-precision
    # as these weights are only used for inference, keeping weights in full precision is not required.
    weight_dtype = torch.float32
    if accelerator.mixed_precision == "fp16":
        weight_dtype = torch.float16
        args.mixed_precision = accelerator.mixed_precision
    elif accelerator.mixed_precision == "bf16":
        weight_dtype = torch.bfloat16
        args.mixed_precision = accelerator.mixed_precision

    # Load scheduler, tokenizer and models.
    if args.loss_type == "ddpm":
        noise_scheduler = DDPMScheduler.from_pretrained(args.pretrained_model_name_or_path, subfolder="scheduler")
    elif args.loss_type == "flow":
        noise_scheduler = FlowMatchEulerDiscreteScheduler.from_pretrained(args.pretrained_model_name_or_path, subfolder="scheduler")
    else:
        train_diffusion = SpacedDiffusion(
            use_timesteps=space_timesteps(1000, str(args.train_sampling_steps)), betas=gd.get_named_beta_schedule("linear", 1000),
            model_mean_type=(gd.ModelMeanType.EPSILON), model_var_type=((gd.ModelVarType.LEARNED_RANGE)),
            loss_type=gd.LossType.MSE, snr=args.snr_loss, return_startx=False,
        )

    if config['text_encoder_kwargs'].get('enable_multi_text_encoder', False):
        print("Init BertTokenizer")
        tokenizer = BertTokenizer.from_pretrained(
            args.pretrained_model_name_or_path, subfolder="tokenizer", revision=args.revision
        )
        if config['text_encoder_kwargs'].get('replace_t5_to_llm', False):
            print("Init LLM Processor")
            tokenizer_2 = Qwen2Tokenizer.from_pretrained(
                os.path.join(args.pretrained_model_name_or_path, "tokenizer_2"), revision=args.revision
            )
        else:
            print("Init T5Tokenizer")
            tokenizer_2 = T5Tokenizer.from_pretrained(
                args.pretrained_model_name_or_path, subfolder="tokenizer_2", revision=args.revision
            )
    else:
        if config['text_encoder_kwargs'].get('replace_t5_to_llm', False):
            print("Init LLM Processor")
            tokenizer = Qwen2Tokenizer.from_pretrained(
                os.path.join(args.pretrained_model_name_or_path, "tokenizer"), revision=args.revision
            )
        else:
            print("Init T5Tokenizer")
            tokenizer = T5Tokenizer.from_pretrained(
                args.pretrained_model_name_or_path, subfolder="tokenizer", revision=args.revision
            )
        tokenizer_2 = None

    def deepspeed_zero_init_disabled_context_manager():
        """
        returns either a context list that includes one that will disable zero.Init or an empty context list
        """
        deepspeed_plugin = AcceleratorState().deepspeed_plugin if accelerate.state.is_initialized() else None
        if deepspeed_plugin is None:
            return []

        return [deepspeed_plugin.zero3_init_context_manager(enable=False)]

    # Currently Accelerate doesn't know how to handle multiple models under Deepspeed ZeRO stage 3.
    # For this to work properly all models must be run through `accelerate.prepare`. But accelerate
    # will try to assign the same optimizer with the same weights to all models during
    # `deepspeed.initialize`, which of course doesn't work.
    #
    # For now the following workaround will partially support Deepspeed ZeRO-3, by excluding the 2
    # frozen models from being partitioned during `zero.Init` which gets called during
    # `from_pretrained` So CLIPTextModel and AutoencoderKL will not enjoy the parameter sharding
    # across multiple gpus and only UNet2DConditionModel will get ZeRO sharded.
    with ContextManagers(deepspeed_zero_init_disabled_context_manager()):
        if config['text_encoder_kwargs'].get('enable_multi_text_encoder', False):
            text_encoder = BertModel.from_pretrained(
                args.pretrained_model_name_or_path, subfolder="text_encoder", revision=args.revision, variant=args.variant,
                torch_dtype=weight_dtype
            )
            if config['text_encoder_kwargs'].get('replace_t5_to_llm', False):
                text_encoder_2 = Qwen2VLForConditionalGeneration.from_pretrained(
                    os.path.join(args.pretrained_model_name_or_path, "text_encoder_2"), revision=args.revision, variant=args.variant,
                    torch_dtype=weight_dtype,
                )
            else:
                text_encoder_2 = T5EncoderModel.from_pretrained(
                    args.pretrained_model_name_or_path, subfolder="text_encoder_2", revision=args.revision, variant=args.variant,
                    torch_dtype=weight_dtype,
                )
        else:
            if config['text_encoder_kwargs'].get('replace_t5_to_llm', False):
                text_encoder = Qwen2VLForConditionalGeneration.from_pretrained(
                    os.path.join(args.pretrained_model_name_or_path, "text_encoder"), revision=args.revision, variant=args.variant,
                    torch_dtype=weight_dtype,
                )
            else:
                text_encoder = T5EncoderModel.from_pretrained(
                    args.pretrained_model_name_or_path, subfolder="text_encoder", revision=args.revision, variant=args.variant,
                    torch_dtype=weight_dtype
                )
            text_encoder_2 = None

        # Get Vae
        Choosen_AutoencoderKL = name_to_autoencoder_magvit[
            config['vae_kwargs'].get('vae_type', 'AutoencoderKL')
        ]
        vae = Choosen_AutoencoderKL.from_pretrained(
            args.pretrained_model_name_or_path, subfolder="vae", revision=args.revision, variant=args.variant,
            vae_additional_kwargs=OmegaConf.to_container(config['vae_kwargs'])
        )

    # Get Transformer
    Choosen_Transformer3DModel = name_to_transformer3d[
        config['transformer_additional_kwargs'].get('transformer_type', 'Transformer3DModel')
    ]
    transformer3d = Choosen_Transformer3DModel.from_pretrained_2d(
        args.pretrained_model_name_or_path, subfolder="transformer",
        transformer_additional_kwargs=OmegaConf.to_container(config['transformer_additional_kwargs'])
    )

    # Get Image encoder
    if args.train_mode != "normal" and config['transformer_additional_kwargs'].get('enable_clip_in_inpaint', True):
        image_encoder = CLIPVisionModelWithProjection.from_pretrained(
            args.pretrained_model_name_or_path, subfolder="image_encoder"
        )
        image_processor = CLIPImageProcessor.from_pretrained(
            args.pretrained_model_name_or_path, subfolder="image_encoder"
        )
    else:
        image_encoder = None
        image_processor = None

    # Freeze vae and text_encoder and set transformer3d to trainable
    vae.requires_grad_(False)
    text_encoder.requires_grad_(False)
    if config['text_encoder_kwargs'].get('enable_multi_text_encoder', False):
        text_encoder_2.requires_grad_(False)
    transformer3d.requires_grad_(False)
    if args.train_mode != "normal" and config['transformer_additional_kwargs'].get('enable_clip_in_inpaint', True):
        image_encoder.requires_grad_(False)

    # Lora will work with this...
    network = create_network(
        1.0,
        args.rank,
        args.network_alpha,
        text_encoder,
        transformer3d,
        neuron_dropout=None,
        add_lora_in_attn_temporal=True,
    )
    network.apply_to(text_encoder, transformer3d, args.train_text_encoder and not args.training_with_video_token_length, True)

    # Load transformer and vae from path if it needs.
    if args.transformer_path is not None:
        print(f"From checkpoint: {args.transformer_path}")
        if args.transformer_path.endswith("safetensors"):
            from safetensors.torch import load_file, safe_open
            state_dict = load_file(args.transformer_path)
        else:
            state_dict = torch.load(args.transformer_path, map_location="cpu")
        state_dict = state_dict["state_dict"] if "state_dict" in state_dict else state_dict

        m, u = transformer3d.load_state_dict(state_dict, strict=False)
        print(f"missing keys: {len(m)}, unexpected keys: {len(u)}")

    if args.vae_path is not None:
        print(f"From checkpoint: {args.vae_path}")
        if args.vae_path.endswith("safetensors"):
            from safetensors.torch import load_file, safe_open
            state_dict = load_file(args.vae_path)
        else:
            state_dict = torch.load(args.vae_path, map_location="cpu")
        state_dict = state_dict["state_dict"] if "state_dict" in state_dict else state_dict

        m, u = vae.load_state_dict(state_dict, strict=False)
        print(f"missing keys: {len(m)}, unexpected keys: {len(u)}")

    if args.enable_xformers_memory_efficient_attention \
        and config['transformer_additional_kwargs'].get('transformer_type', 'Transformer3DModel') == 'Transformer3DModel':
        if is_xformers_available():
            import xformers

            xformers_version = version.parse(xformers.__version__)
            if xformers_version == version.parse("0.0.16"):
                logger.warn(
                    "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
                )
            transformer3d.enable_xformers_memory_efficient_attention()
        else:
            raise ValueError("xformers is not available. Make sure it is installed correctly")

    # `accelerate` 0.16.0 will have better support for customized saving
    if version.parse(accelerate.__version__) >= version.parse("0.16.0"):
        # create custom saving & loading hooks so that `accelerator.save_state(...)` serializes in a nice format
        def save_model_hook(models, weights, output_dir):
            if accelerator.is_main_process:
                safetensor_save_path = os.path.join(output_dir, f"lora_diffusion_pytorch_model.safetensors")
                save_model(safetensor_save_path, accelerator.unwrap_model(models[-1]))

                with open(os.path.join(output_dir, "sampler_pos_start.pkl"), 'wb') as file:
                    pickle.dump([batch_sampler.sampler._pos_start, first_epoch], file)

        def load_model_hook(models, input_dir):
            pkl_path = os.path.join(input_dir, "sampler_pos_start.pkl")
            if os.path.exists(pkl_path):
                with open(pkl_path, 'rb') as file:
                    loaded_number, _ = pickle.load(file)
                    batch_sampler.sampler._pos_start = max(loaded_number - args.dataloader_num_workers * accelerator.num_processes * 2, 0)
                print(f"Load pkl from {pkl_path}. Get loaded_number = {loaded_number}.")

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

    if args.gradient_checkpointing:
        transformer3d.enable_gradient_checkpointing()

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

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

    # Initialize the optimizer
    if args.use_8bit_adam:
        try:
            import bitsandbytes as bnb
        except ImportError:
            raise ImportError(
                "Please install bitsandbytes to use 8-bit Adam. You can do so by running `pip install bitsandbytes`"
            )

        optimizer_cls = bnb.optim.AdamW8bit
    elif args.use_came:
        try:
            from came_pytorch import CAME
        except:
            raise ImportError(
                "Please install came_pytorch to use CAME. You can do so by running `pip install came_pytorch`"
            )

        optimizer_cls = CAME
    else:
        optimizer_cls = torch.optim.AdamW

    logging.info("Add network parameters")
    trainable_params = list(filter(lambda p: p.requires_grad, network.parameters()))
    trainable_params_optim = network.prepare_optimizer_params(args.learning_rate / 2, args.learning_rate, args.learning_rate)

    # Init optimizer
    if args.use_came:
        optimizer = optimizer_cls(
            trainable_params_optim,
            lr=args.learning_rate,
            betas=(0.9, 0.999, 0.9999), 
            eps=(1e-30, 1e-16)
        )
    else:
        optimizer = optimizer_cls(
            trainable_params_optim,
            lr=args.learning_rate,
            betas=(args.adam_beta1, args.adam_beta2),
            weight_decay=args.adam_weight_decay,
            eps=args.adam_epsilon,
        )

    # Get the training dataset
    sample_n_frames_bucket_interval = vae.mini_batch_encoder if vae.quant_conv is None or vae.quant_conv.weight.ndim==5 else 4

    # Get the dataset
    train_dataset = ImageVideoDataset(
        args.train_data_meta, args.train_data_dir,
        video_sample_size=args.video_sample_size, video_sample_stride=args.video_sample_stride, video_sample_n_frames=args.video_sample_n_frames, 
        video_repeat=args.video_repeat, 
        image_sample_size=args.image_sample_size,
        enable_bucket=args.enable_bucket, enable_inpaint=True if args.train_mode != "normal" else False,
    )
    
    if args.enable_bucket:
        aspect_ratio_sample_size = {key : [x / 512 * args.video_sample_size for x in ASPECT_RATIO_512[key]] for key in ASPECT_RATIO_512.keys()}
        batch_sampler_generator = torch.Generator().manual_seed(args.seed)
        batch_sampler = AspectRatioBatchImageVideoSampler(
            sampler=RandomSampler(train_dataset, generator=batch_sampler_generator), dataset=train_dataset.dataset, 
            batch_size=args.train_batch_size, train_folder = args.train_data_dir, drop_last=True,
            aspect_ratios=aspect_ratio_sample_size,
        )
        
        # Get the frame length at different resolutions according to token_length
        def get_length_to_frame_num(token_length):
            if args.image_sample_size > args.video_sample_size:
                sample_sizes = list(range(args.video_sample_size, args.image_sample_size + 1, 128))

                if sample_sizes[-1] != args.image_sample_size:
                    sample_sizes.append(args.image_sample_size)
            else:
                sample_sizes = [args.image_sample_size]
            
            if vae.cache_mag_vae:
                length_to_frame_num = {
                    sample_size: min(token_length / sample_size / sample_size, args.video_sample_n_frames) // sample_n_frames_bucket_interval * sample_n_frames_bucket_interval + 1 for sample_size in sample_sizes
                }
            else:
                length_to_frame_num = {
                    sample_size: min(token_length / sample_size / sample_size, args.video_sample_n_frames) // sample_n_frames_bucket_interval * sample_n_frames_bucket_interval for sample_size in sample_sizes
                }

            return length_to_frame_num

        def collate_fn(examples):
            # Get token length
            target_token_length = args.video_sample_n_frames * args.token_sample_size * args.token_sample_size
            length_to_frame_num = get_length_to_frame_num(target_token_length)

            # Create new output
            new_examples                 = {}
            new_examples["target_token_length"] = target_token_length
            new_examples["pixel_values"] = []
            new_examples["text"]         = []
            # Used in Inpaint mode
            if args.train_mode != "normal":
                new_examples["mask_pixel_values"] = []
                new_examples["mask"] = []
                new_examples["clip_pixel_values"] = []

            # Get downsample ratio in image and videos
            pixel_value     = examples[0]["pixel_values"]
            data_type       = examples[0]["data_type"]
            f, h, w, c      = np.shape(pixel_value)
            if data_type == 'image':
                random_downsample_ratio = 1 if not args.random_hw_adapt else get_random_downsample_ratio(args.image_sample_size, image_ratio=[args.image_sample_size / args.video_sample_size], rng=rng)

                aspect_ratio_sample_size = {key : [x / 512 * args.image_sample_size / random_downsample_ratio for x in ASPECT_RATIO_512[key]] for key in ASPECT_RATIO_512.keys()}
                aspect_ratio_random_crop_sample_size = {key : [x / 512 * args.image_sample_size / random_downsample_ratio for x in ASPECT_RATIO_RANDOM_CROP_512[key]] for key in ASPECT_RATIO_RANDOM_CROP_512.keys()}
                
                batch_video_length = args.video_sample_n_frames + sample_n_frames_bucket_interval
            else:
                if args.random_hw_adapt:
                    if args.training_with_video_token_length:
                        local_min_size = np.min(np.array([np.mean(np.array([np.shape(example["pixel_values"])[1], np.shape(example["pixel_values"])[2]])) for example in examples]))
                        # The video will be resized to a lower resolution than its own.
                        choice_list = [length for length in list(length_to_frame_num.keys()) if length < local_min_size * 1.25]
                        if len(choice_list) == 0:
                            choice_list = list(length_to_frame_num.keys())
                        if rng is None:
                            local_video_sample_size = np.random.choice(choice_list)
                        else:
                            local_video_sample_size = rng.choice(choice_list)
                        batch_video_length = length_to_frame_num[local_video_sample_size]
                        random_downsample_ratio = args.video_sample_size / local_video_sample_size
                    else:
                        random_downsample_ratio = get_random_downsample_ratio(
                                args.video_sample_size, rng=rng)
                        batch_video_length = args.video_sample_n_frames + sample_n_frames_bucket_interval
                else:
                    random_downsample_ratio = 1
                    batch_video_length = args.video_sample_n_frames + sample_n_frames_bucket_interval

                aspect_ratio_sample_size = {key : [x / 512 * args.video_sample_size / random_downsample_ratio for x in ASPECT_RATIO_512[key]] for key in ASPECT_RATIO_512.keys()}
                aspect_ratio_random_crop_sample_size = {key : [x / 512 * args.video_sample_size / random_downsample_ratio for x in ASPECT_RATIO_RANDOM_CROP_512[key]] for key in ASPECT_RATIO_RANDOM_CROP_512.keys()}

            closest_size, closest_ratio = get_closest_ratio(h, w, ratios=aspect_ratio_sample_size)
            closest_size = [int(x / 16) * 16 for x in closest_size]
            if args.random_ratio_crop:
                if rng is None:
                    random_sample_size = aspect_ratio_random_crop_sample_size[
                        np.random.choice(list(aspect_ratio_random_crop_sample_size.keys()), p = ASPECT_RATIO_RANDOM_CROP_PROB)
                    ]
                else:
                    random_sample_size = aspect_ratio_random_crop_sample_size[
                        rng.choice(list(aspect_ratio_random_crop_sample_size.keys()), p = ASPECT_RATIO_RANDOM_CROP_PROB)
                    ]
                random_sample_size = [int(x / 16) * 16 for x in random_sample_size]

            for example in examples:
                if args.random_ratio_crop:
                    # To 0~1
                    pixel_values = torch.from_numpy(example["pixel_values"]).permute(0, 3, 1, 2).contiguous()
                    pixel_values = pixel_values / 255.

                    # Get adapt hw for resize
                    b, c, h, w = pixel_values.size()
                    th, tw = random_sample_size
                    if th / tw > h / w:
                        nh = int(th)
                        nw = int(w / h * nh)
                    else:
                        nw = int(tw)
                        nh = int(h / w * nw)
                    
                    transform = transforms.Compose([
                        transforms.Resize([nh, nw]),
                        transforms.CenterCrop([int(x) for x in random_sample_size]),
                        transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True),
                    ])
                else:
                    # To 0~1
                    pixel_values = torch.from_numpy(example["pixel_values"]).permute(0, 3, 1, 2).contiguous()
                    pixel_values = pixel_values / 255.

                    # Get adapt hw for resize
                    closest_size = list(map(lambda x: int(x), closest_size))
                    if closest_size[0] / h > closest_size[1] / w:
                        resize_size = closest_size[0], int(w * closest_size[0] / h)
                    else:
                        resize_size = int(h * closest_size[1] / w), closest_size[1]
                    
                    transform = transforms.Compose([
                        transforms.Resize(resize_size, interpolation=transforms.InterpolationMode.BILINEAR),  # Image.BICUBIC
                        transforms.CenterCrop(closest_size),
                        transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True),
                    ])
                new_examples["pixel_values"].append(transform(pixel_values))
                new_examples["text"].append(example["text"])

                # Magvae needs the number of frames to be 4n + 1.
                if vae.cache_mag_vae:
                    batch_video_length = int(
                        min(
                            batch_video_length,
                            (len(pixel_values) - 1) // sample_n_frames_bucket_interval * sample_n_frames_bucket_interval + 1, 
                        )
                    )
                else:
                    batch_video_length = int(
                        min(
                            batch_video_length,
                            len(pixel_values) // sample_n_frames_bucket_interval * sample_n_frames_bucket_interval, 
                        )
                    )
                if batch_video_length == 0:
                    batch_video_length = 1

                # Used in Inpaint mode
                if args.train_mode != "normal":
                    mask = get_random_mask(new_examples["pixel_values"][-1].size())
                    mask_pixel_values = new_examples["pixel_values"][-1] * (1 - mask) + torch.ones_like(new_examples["pixel_values"][-1]) * -1 * mask
                    new_examples["mask_pixel_values"].append(mask_pixel_values)
                    new_examples["mask"].append(mask)

                    def get_random_clip_index(low, high):
                        if high - low <= 1.1:
                            return low
                        values = np.arange(low, high)
                        probabilities = np.ones(len(values)) * 0.5 / (len(values) - 1)
                        probabilities[0] = 0.5
                        return np.random.choice(values, p=probabilities)

                    clip_index = get_random_clip_index(0, len(new_examples["pixel_values"][-1]))
                    clip_pixel_values = new_examples["pixel_values"][-1][clip_index].permute(1, 2, 0).contiguous()
                    clip_pixel_values = (clip_pixel_values * 0.5 + 0.5) * 255
                    new_examples["clip_pixel_values"].append(clip_pixel_values)
  
            # Limit the number of frames to the same
            new_examples["pixel_values"] = torch.stack([example[:batch_video_length] for example in new_examples["pixel_values"]])
            if args.train_mode != "normal":
                new_examples["mask_pixel_values"] = torch.stack([example[:batch_video_length] for example in new_examples["mask_pixel_values"]])
                new_examples["mask"] = torch.stack([example[:batch_video_length] for example in new_examples["mask"]])
                new_examples["clip_pixel_values"] = torch.stack([example for example in new_examples["clip_pixel_values"]])

            # Encode prompts when enable_text_encoder_in_dataloader=True
            if args.enable_text_encoder_in_dataloader:
                if config['text_encoder_kwargs'].get('enable_multi_text_encoder', False):
                    prompt_embeds, prompt_attention_mask = \
                        encode_prompt(
                            tokenizer, text_encoder, new_examples['text'], None, 
                            tokenizer_max_length=args.tokenizer_max_length, dtype=weight_dtype,
                            enable_text_attention_mask=transformer3d.config.get("enable_text_attention_mask", True)
                        )
                    prompt_embeds_2, prompt_attention_mask_2 = \
                        encode_prompt(
                            tokenizer_2, text_encoder_2, new_examples['text'], None, 
                            tokenizer_max_length=args.tokenizer_max_length, dtype=weight_dtype,
                            enable_text_attention_mask=transformer3d.config.get("enable_text_attention_mask", True)
                        )
                    new_examples['prompt_embeds'] = prompt_embeds
                    new_examples['prompt_attention_mask'] = prompt_attention_mask
                    new_examples['prompt_embeds_2'] = prompt_embeds_2
                    new_examples['prompt_attention_mask_2'] = prompt_attention_mask_2
                else:
                    prompt_embeds, prompt_attention_mask = \
                        encode_prompt(
                            tokenizer, text_encoder, new_examples['text'], None, 
                            tokenizer_max_length=args.tokenizer_max_length, dtype=weight_dtype,
                                    add_special_tokens=True,
                            enable_text_attention_mask=transformer3d.config.get("enable_text_attention_mask", True)
                        )
                    new_examples['prompt_embeds'] = prompt_embeds
                    new_examples['prompt_attention_mask'] = prompt_attention_mask

            return new_examples
        
        # DataLoaders creation:
        train_dataloader = torch.utils.data.DataLoader(
            train_dataset,
            batch_sampler=batch_sampler,
            collate_fn=collate_fn,
            persistent_workers=True if args.dataloader_num_workers != 0 else False,
            num_workers=args.dataloader_num_workers,
        )
    else:
        # DataLoaders creation:
        batch_sampler_generator = torch.Generator().manual_seed(args.seed)
        batch_sampler = ImageVideoSampler(RandomSampler(train_dataset, generator=batch_sampler_generator), train_dataset, args.train_batch_size)
        train_dataloader = torch.utils.data.DataLoader(
            train_dataset,
            batch_sampler=batch_sampler, 
            persistent_workers=True if args.dataloader_num_workers != 0 else False,
            num_workers=args.dataloader_num_workers,
        )

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

    lr_scheduler = get_scheduler(
        args.lr_scheduler,
        optimizer=optimizer,
        num_warmup_steps=args.lr_warmup_steps * accelerator.num_processes,
        num_training_steps=args.max_train_steps * accelerator.num_processes,
    )

    # Prepare everything with our `accelerator`.
    network, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
        network, optimizer, train_dataloader, lr_scheduler
    )

    # Move text_encode and vae to gpu and cast to weight_dtype
    vae.to(accelerator.device, dtype=weight_dtype)
    if args.train_mode != "normal" and config['transformer_additional_kwargs'].get('enable_clip_in_inpaint', True):
        image_encoder.to(accelerator.device, dtype=weight_dtype)
    transformer3d.to(accelerator.device, dtype=weight_dtype)
    if not args.enable_text_encoder_in_dataloader:
        text_encoder.to(accelerator.device)
        if config['text_encoder_kwargs'].get('enable_multi_text_encoder', False):
            text_encoder_2.to(accelerator.device)

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

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

    # Function for unwrapping if model was compiled with `torch.compile`.
    def unwrap_model(model):
        model = accelerator.unwrap_model(model)
        model = model._orig_mod if is_compiled_module(model) else model
        return model

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

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

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

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

            initial_global_step = global_step

            pkl_path = os.path.join(os.path.join(args.output_dir, path), "sampler_pos_start.pkl")
            if os.path.exists(pkl_path):
                with open(pkl_path, 'rb') as file:
                    _, first_epoch = pickle.load(file)
            else:
                first_epoch = global_step // num_update_steps_per_epoch
            print(f"Load pkl from {pkl_path}. Get first_epoch = {first_epoch}.")

            from safetensors.torch import load_file, safe_open
            state_dict = load_file(os.path.join(os.path.join(args.output_dir, path), "lora_diffusion_pytorch_model.safetensors"))
            m, u = accelerator.unwrap_model(network).load_state_dict(state_dict, strict=False)
            print(f"missing keys: {len(m)}, unexpected keys: {len(u)}")

            accelerator.print(f"Resuming from checkpoint {path}")
            accelerator.load_state(os.path.join(args.output_dir, path))
    else:
        initial_global_step = 0

    # function for saving/removing
    def save_model(ckpt_file, unwrapped_nw):
        os.makedirs(args.output_dir, exist_ok=True)
        accelerator.print(f"\nsaving checkpoint: {ckpt_file}")
        unwrapped_nw.save_weights(ckpt_file, weight_dtype, None)

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

    if args.multi_stream and args.train_mode != "normal":
        # create extra cuda streams to speedup inpaint vae computation
        vae_stream_1 = torch.cuda.Stream()
        vae_stream_2 = torch.cuda.Stream()
    else:
        vae_stream_1 = None
        vae_stream_2 = None

    idx_sampling = DiscreteSampling(args.train_sampling_steps, uniform_sampling=args.uniform_sampling)

    for epoch in range(first_epoch, args.num_train_epochs):
        train_loss = 0.0
        batch_sampler.sampler.generator = torch.Generator().manual_seed(args.seed + epoch)
        for step, batch in enumerate(train_dataloader):
            # Data batch sanity check
            if epoch == first_epoch and step == 0:
                pixel_values, texts = batch['pixel_values'].cpu(), batch['text']
                pixel_values = rearrange(pixel_values, "b f c h w -> b c f h w")
                if pixel_values.ndim==4:
                    pixel_values = pixel_values.unsqueeze(2)
                os.makedirs(os.path.join(args.output_dir, "sanity_check"), exist_ok=True)
                for idx, (pixel_value, text) in enumerate(zip(pixel_values, texts)):
                    pixel_value = pixel_value[None, ...]
                    gif_name = '-'.join(text.replace('/', '').split()[:10]) if not text == '' else f'{global_step}-{idx}'
                    save_videos_grid(pixel_value, f"{args.output_dir}/sanity_check/{gif_name[:10]}.gif", rescale=True)
                if args.train_mode != "normal":
                    clip_pixel_values, mask_pixel_values, texts = batch['clip_pixel_values'].cpu(), batch['mask_pixel_values'].cpu(), batch['text']
                    mask_pixel_values = rearrange(mask_pixel_values, "b f c h w -> b c f h w")
                    for idx, (clip_pixel_value, pixel_value, text) in enumerate(zip(clip_pixel_values, mask_pixel_values, texts)):
                        pixel_value = pixel_value[None, ...]
                        Image.fromarray(np.uint8(clip_pixel_value)).save(f"{args.output_dir}/sanity_check/clip_{gif_name[:10] if not text == '' else f'{global_step}-{idx}'}.png")
                        save_videos_grid(pixel_value, f"{args.output_dir}/sanity_check/mask_{gif_name[:10] if not text == '' else f'{global_step}-{idx}'}.gif", rescale=True)

            with accelerator.accumulate(transformer3d):
                # Convert images to latent space
                pixel_values = batch["pixel_values"].to(weight_dtype)

                # Increase the batch size when the length of the latent sequence of the current sample is small
                if args.training_with_video_token_length:
                    if args.video_sample_n_frames * args.token_sample_size * args.token_sample_size // 16 >= pixel_values.size()[1] * pixel_values.size()[3] * pixel_values.size()[4]:
                        pixel_values = torch.tile(pixel_values, (4, 1, 1, 1, 1))
                        if args.enable_text_encoder_in_dataloader:
                            batch['prompt_embeds'] = torch.tile(batch['prompt_embeds'], (4, 1, 1))
                            batch['prompt_attention_mask'] = torch.tile(batch['prompt_attention_mask'], (4, 1))
                            if config['text_encoder_kwargs'].get('enable_multi_text_encoder', False):
                                batch['prompt_embeds_2'] = torch.tile(batch['prompt_embeds_2'], (4, 1, 1))
                                batch['prompt_attention_mask_2'] = torch.tile(batch['prompt_attention_mask_2'], (4, 1))
                        else:
                            batch['text'] = batch['text'] * 4
                    elif args.video_sample_n_frames * args.token_sample_size * args.token_sample_size // 4 >= pixel_values.size()[1] * pixel_values.size()[3] * pixel_values.size()[4]:
                        pixel_values = torch.tile(pixel_values, (2, 1, 1, 1, 1))
                        if args.enable_text_encoder_in_dataloader:
                            batch['prompt_embeds'] = torch.tile(batch['prompt_embeds'], (2, 1, 1))
                            batch['prompt_attention_mask'] = torch.tile(batch['prompt_attention_mask'], (2, 1))
                            if config['text_encoder_kwargs'].get('enable_multi_text_encoder', False):
                                batch['prompt_embeds_2'] = torch.tile(batch['prompt_embeds_2'], (2, 1, 1))
                                batch['prompt_attention_mask_2'] = torch.tile(batch['prompt_attention_mask_2'], (2, 1))
                        else:
                            batch['text'] = batch['text'] * 2

                if args.train_mode != "normal":
                    clip_pixel_values = batch["clip_pixel_values"]
                    mask_pixel_values = batch["mask_pixel_values"].to(weight_dtype)
                    mask = batch["mask"].to(weight_dtype)
                    # Increase the batch size when the length of the latent sequence of the current sample is small
                    if args.training_with_video_token_length:
                        if args.video_sample_n_frames * args.token_sample_size * args.token_sample_size // 16 >= pixel_values.size()[1] * pixel_values.size()[3] * pixel_values.size()[4]:
                            clip_pixel_values = torch.tile(clip_pixel_values, (4, 1, 1, 1))
                            mask_pixel_values = torch.tile(mask_pixel_values, (4, 1, 1, 1, 1))
                            mask = torch.tile(mask, (4, 1, 1, 1, 1))
                        elif args.video_sample_n_frames * args.token_sample_size * args.token_sample_size // 4 >= pixel_values.size()[1] * pixel_values.size()[3] * pixel_values.size()[4]:
                            clip_pixel_values = torch.tile(clip_pixel_values, (2, 1, 1, 1))
                            mask_pixel_values = torch.tile(mask_pixel_values, (2, 1, 1, 1, 1))
                            mask = torch.tile(mask, (2, 1, 1, 1, 1))

                # Random crop number of frames to adapt different frames of video.
                if args.random_frame_crop:
                    def _create_special_list(length):
                        if length == 1:
                            return [1.0]
                        if length >= 2:
                            last_element = 0.90
                            remaining_sum = 1.0 - last_element
                            other_elements_value = remaining_sum / (length - 1)
                            special_list = [other_elements_value] * (length - 1) + [last_element]
                            return special_list
                        
                    if vae.cache_mag_vae:
                        select_frames = [_tmp for _tmp in list(range(sample_n_frames_bucket_interval + 1, args.video_sample_n_frames + sample_n_frames_bucket_interval, sample_n_frames_bucket_interval))]
                    else:
                        select_frames = [_tmp for _tmp in list(range(sample_n_frames_bucket_interval, args.video_sample_n_frames + sample_n_frames_bucket_interval, sample_n_frames_bucket_interval))]
                    select_frames_prob = np.array(_create_special_list(len(select_frames)))
                    
                    if rng is None:
                        temp_n_frames = np.random.choice(select_frames, p = select_frames_prob)
                    else:
                        temp_n_frames = rng.choice(select_frames, p = select_frames_prob)

                    pixel_values = pixel_values[:, :temp_n_frames, :, :]

                    if args.train_mode != "normal":
                        mask_pixel_values = mask_pixel_values[:, :temp_n_frames, :, :]
                        mask = mask[:, :temp_n_frames, :, :]
                    
                # Keep all node same token length to accelerate the traning when resolution grows.
                if args.keep_all_node_same_token_length:
                    if args.token_sample_size > 256:
                        numbers_list = list(range(256, args.token_sample_size + 1, 128))

                        if numbers_list[-1] != args.token_sample_size:
                            numbers_list.append(args.token_sample_size)
                    else:
                        numbers_list = [256]
                    numbers_list = [_number * _number * args.video_sample_n_frames for _number in  numbers_list]
            
                    actual_token_length = index_rng.choice(numbers_list)
                    if vae.cache_mag_vae:
                        actual_video_length = (min(
                                actual_token_length / pixel_values.size()[-1] / pixel_values.size()[-2], args.video_sample_n_frames
                        ) - 1) // sample_n_frames_bucket_interval * sample_n_frames_bucket_interval + 1
                    else:
                        actual_video_length = min(
                            actual_token_length / pixel_values.size()[-1] / pixel_values.size()[-2], args.video_sample_n_frames
                        ) // sample_n_frames_bucket_interval * sample_n_frames_bucket_interval
                    actual_video_length = int(max(actual_video_length, 1))

                    pixel_values = pixel_values[:, :actual_video_length, :, :]
                    if args.train_mode != "normal":
                        mask_pixel_values = mask_pixel_values[:, :actual_video_length, :, :]
                        mask = mask[:, :actual_video_length, :, :]

                # Make the inpaint latents to be zeros.
                if args.train_mode != "normal":
                    t2v_flag = [(_mask == 1).all() for _mask in mask]
                    new_t2v_flag = []
                    for _mask in t2v_flag:
                        if _mask and np.random.rand() < 0.90:
                            new_t2v_flag.append(0)
                        else:
                            new_t2v_flag.append(1)
                    t2v_flag = torch.from_numpy(np.array(new_t2v_flag)).to(accelerator.device, dtype=weight_dtype)

                # Reduce the vram by offload vae and text encoders
                if args.low_vram:
                    torch.cuda.empty_cache()
                    vae.to(accelerator.device)
                    if not args.enable_text_encoder_in_dataloader:
                        text_encoder.to('cpu')
                        if text_encoder_2 is not None:
                            text_encoder_2.to('cpu')

                with torch.no_grad():
                    video_length = pixel_values.shape[1]

                    # This way is quicker when batch grows up
                    def _batch_encode_vae(pixel_values):
                        if vae.quant_conv is None or vae.quant_conv.weight.ndim==5:
                            pixel_values = rearrange(pixel_values, "b f c h w -> b c f h w")
                            bs = args.vae_mini_batch
                            new_pixel_values = []
                            for i in range(0, pixel_values.shape[0], bs):
                                pixel_values_bs = pixel_values[i : i + bs]
                                pixel_values_bs = vae.encode(pixel_values_bs)[0]
                                pixel_values_bs = pixel_values_bs.sample()
                                new_pixel_values.append(pixel_values_bs)
                            latents = torch.cat(new_pixel_values, dim = 0)
                        else:
                            pixel_values = rearrange(pixel_values, "b f c h w -> (b f) c h w")
                            bs = args.vae_mini_batch
                            new_pixel_values = []
                            for i in range(0, pixel_values.shape[0], bs):
                                pixel_values_bs = pixel_values[i : i + bs]
                                pixel_values_bs = vae.encode(pixel_values_bs.to(dtype=weight_dtype)).latent_dist
                                pixel_values_bs = pixel_values_bs.sample()
                                new_pixel_values.append(pixel_values_bs)
                            latents = torch.cat(new_pixel_values, dim = 0)
                            latents = rearrange(latents, "(b f) c h w -> b c f h w", f=video_length)
                        return latents
                
                    # Encode latents.
                    if vae_stream_1 is not None:
                        vae_stream_1.wait_stream(torch.cuda.current_stream())
                        with torch.cuda.stream(vae_stream_1):
                            latents = _batch_encode_vae(pixel_values)
                    else:
                        latents = _batch_encode_vae(pixel_values)
                    latents = latents * vae.config.scaling_factor

                    if args.train_mode != "normal":
                        # Encode masks.
                        if config['transformer_additional_kwargs'].get('resize_inpaint_mask_directly', False):
                            mask = rearrange(mask, "b f c h w -> b c f h w")
                            mask = 1 - mask
                            mask = resize_mask(mask, latents, vae.cache_mag_vae)
                        else:
                            mask = torch.tile(mask, [1, 1, 3, 1, 1])
                            if vae_stream_2 is not None:
                                vae_stream_2.wait_stream(torch.cuda.current_stream())
                                with torch.cuda.stream(vae_stream_2):
                                    mask = _batch_encode_vae(mask)
                            else:
                                mask = _batch_encode_vae(mask)

                        if unwrap_model(transformer3d).config.add_noise_in_inpaint_model:
                            mask_pixel_values = add_noise_to_reference_video(mask_pixel_values)
                        # Encode inpaint latents.
                        mask_latents = _batch_encode_vae(mask_pixel_values)
                        if vae_stream_2 is not None:
                            torch.cuda.current_stream().wait_stream(vae_stream_2) 

                        inpaint_latents = torch.concat([mask, mask_latents], dim=1)
                        inpaint_latents = t2v_flag[:, None, None, None, None] * inpaint_latents
                        inpaint_latents = inpaint_latents * vae.config.scaling_factor

                        # Encode Vision CLIP latents
                        with torch.no_grad():
                            if config['transformer_additional_kwargs'].get('enable_clip_in_inpaint', True):
                                clip_encoder_hidden_states = []
                                for clip_pixel_value in clip_pixel_values:
                                    image = Image.fromarray(np.uint8(clip_pixel_value.cpu().numpy()))
                                    inputs = image_processor(images=image, return_tensors="pt")
                                    inputs["pixel_values"] = inputs["pixel_values"].to(accelerator.device, dtype=weight_dtype)
                                    if config['text_encoder_kwargs'].get('enable_multi_text_encoder', False):
                                        outputs = image_encoder(**inputs).last_hidden_state[0, 1:]
                                    else:
                                        outputs = image_encoder(**inputs).image_embeds[0]
                                    clip_encoder_hidden_states.append(outputs)
                                clip_encoder_hidden_states = torch.stack(clip_encoder_hidden_states)
                                bsc = clip_encoder_hidden_states.shape[0]

                                if config['text_encoder_kwargs'].get('enable_multi_text_encoder', False):
                                    clip_attention_mask = torch.ones([bsc, unwrap_model(transformer3d).n_query]) if np.random.rand() < 0.50 else torch.zeros([bsc, unwrap_model(transformer3d).n_query])
                                else:
                                    clip_attention_mask = torch.ones([bsc, 8]) if np.random.rand() < 0.50 else torch.zeros([bsc, 8])
                                clip_attention_mask = clip_attention_mask.to(accelerator.device, dtype=weight_dtype)
                            else:
                                clip_encoder_hidden_states = None
                                clip_attention_mask = None
                        
                # wait for latents = vae.encode(pixel_values) to complete
                if vae_stream_1 is not None:
                    torch.cuda.current_stream().wait_stream(vae_stream_1)

                # Reduce the vram by offload vae and text encoders
                if args.low_vram:
                    vae.to('cpu')
                    torch.cuda.empty_cache()
                    if not args.enable_text_encoder_in_dataloader:
                        text_encoder.to(accelerator.device)
                        if text_encoder_2 is not None:
                            text_encoder_2.to(accelerator.device)

                # Get text embeds
                if args.enable_text_encoder_in_dataloader:
                    if config['text_encoder_kwargs'].get('enable_multi_text_encoder', False):
                        prompt_embeds = batch['prompt_embeds'].to(device=latents.device)
                        prompt_attention_mask = batch['prompt_attention_mask'].to(device=latents.device)
                        prompt_embeds_2 = batch['prompt_embeds_2'].to(device=latents.device)
                        prompt_attention_mask_2 = batch['prompt_attention_mask_2'].to(device=latents.device)
                    else:
                        prompt_embeds = batch['prompt_embeds'].to(device=latents.device)
                        prompt_attention_mask = batch['prompt_attention_mask'].to(device=latents.device)
                        prompt_embeds_2 = None
                        prompt_attention_mask_2 = None
                else:
                    if config['text_encoder_kwargs'].get('enable_multi_text_encoder', False):
                        with torch.no_grad():
                            prompt_embeds, prompt_attention_mask = \
                                encode_prompt(
                                    tokenizer, text_encoder, batch['text'], latents.device, 
                                    tokenizer_max_length=args.tokenizer_max_length, dtype=weight_dtype,
                                    enable_text_attention_mask=unwrap_model(transformer3d).config.get("enable_text_attention_mask", True)
                                )
                            prompt_embeds_2, prompt_attention_mask_2 = \
                                encode_prompt(
                                    tokenizer_2, text_encoder_2, batch['text'], latents.device, 
                                    tokenizer_max_length=args.tokenizer_max_length, dtype=weight_dtype,
                                    enable_text_attention_mask=unwrap_model(transformer3d).config.get("enable_text_attention_mask", True)
                                )

                    else:
                        with torch.no_grad():
                            prompt_embeds, prompt_attention_mask = \
                                encode_prompt(
                                    tokenizer, text_encoder, batch['text'], latents.device, 
                                    tokenizer_max_length=args.tokenizer_max_length, dtype=weight_dtype,
                                    add_special_tokens=True,
                                    enable_text_attention_mask=unwrap_model(transformer3d).config.get("enable_text_attention_mask", True)
                                )

                            prompt_embeds_2 = None
                            prompt_attention_mask_2 = None

                # Reduce the vram by offload vae and text encoders
                if args.low_vram and not args.enable_text_encoder_in_dataloader:
                    text_encoder.to('cpu')
                    if text_encoder_2 is not None:
                        text_encoder_2.to('cpu')
                    torch.cuda.empty_cache()

                # Get Noise
                bsz = latents.shape[0]
                if args.noise_share_in_frames:
                    def generate_noise(bs, channel, length, height, width, ratio=0.5, generator=None, device="cuda", dtype=None):
                        noise = torch.randn(bs, channel, length, height, width, generator=generator, device=device, dtype=dtype)
                        for i in range(1, length):
                            noise[:, :, i, :, :] = ratio * noise[:, :, i - 1, :, :] + (1 - ratio) * noise[:, :, i, :, :]
                        return noise
                    noise = generate_noise(*latents.size(), ratio=args.noise_share_in_frames_ratio, device=latents.device, generator=torch_rng, dtype=weight_dtype)
                else:
                    noise = torch.randn(latents.size(), device=latents.device, generator=torch_rng, dtype=weight_dtype)
                
                # Sample a random timestep for each image
                # timesteps = generate_timestep_with_lognorm(0, args.train_sampling_steps, (bsz,), device=latents.device, generator=torch_rng)
                # timesteps = torch.randint(0, args.train_sampling_steps, (bsz,), device=latents.device, generator=torch_rng)
                timesteps = idx_sampling(bsz, generator=torch_rng, device=latents.device)
                timesteps = timesteps.long()
                
                if args.loss_type != "sigma":
                    # Create image_rotary_emb, style embedding & time ids
                    height, width = batch["pixel_values"].size()[-2], batch["pixel_values"].size()[-1]

                    # Get 3D rope or 2d rope
                    grid_height = height // 8 // unwrap_model(transformer3d).config.patch_size
                    grid_width = width // 8 // unwrap_model(transformer3d).config.patch_size
                    if unwrap_model(transformer3d).config.get("time_position_encoding_type", "2d_rope") == "3d_rope":
                        base_size_width = 720 // 8 // unwrap_model(transformer3d).config.patch_size
                        base_size_height = 480 // 8 // unwrap_model(transformer3d).config.patch_size
                        grid_crops_coords = get_resize_crop_region_for_grid(
                            (grid_height, grid_width), base_size_width, base_size_height
                        )
                        image_rotary_emb = _get_3d_rotary_pos_embed_cached(
                            unwrap_model(transformer3d).config.attention_head_dim,
                            grid_crops_coords, (grid_height, grid_width), latents.size()[2],
                        )
                    else:
                        base_size = 512 // 8 // unwrap_model(transformer3d).config.patch_size
                        grid_crops_coords = get_resize_crop_region_for_grid(
                            (grid_height, grid_width), base_size, base_size
                        )
                        image_rotary_emb = _get_2d_rotary_pos_embed_cached(
                            unwrap_model(transformer3d).config.attention_head_dim,
                            grid_crops_coords, (grid_height, grid_width)
                        )

                    # Get other hunyuan params
                    style = torch.tensor([0], device=latents.device)
                    target_size = (height, width)
                    add_time_ids = list((1024, 1024) + target_size + (0, 0))
                    add_time_ids = torch.tensor([add_time_ids], dtype=prompt_embeds.dtype)
                    
                    # To latents.device
                    prompt_embeds = prompt_embeds.to(device=latents.device)
                    prompt_attention_mask = prompt_attention_mask.to(device=latents.device)
                    if prompt_embeds_2 is not None:
                        prompt_embeds_2 = prompt_embeds_2.to(device=latents.device)
                        prompt_attention_mask_2 = prompt_attention_mask_2.to(device=latents.device)
                    add_time_ids = add_time_ids.to(dtype=prompt_embeds.dtype, device=latents.device).repeat(
                        bsz, 1
                    )
                    style = style.to(device=latents.device).repeat(bsz)

                    if args.loss_type == "ddpm":
                        # Add noise
                        noisy_latents = noise_scheduler.add_noise(latents, noise, timesteps)
                        if noise_scheduler.config.prediction_type == "epsilon":
                            target = noise
                        elif noise_scheduler.config.prediction_type == "v_prediction":
                            target = noise_scheduler.get_velocity(latents, noise, timesteps)
                        else:
                            raise ValueError(f"Unknown prediction type {noise_scheduler.config.prediction_type}")
                    else:
                        def get_sigmas(timesteps, n_dim=4, dtype=torch.float32):
                            sigmas = noise_scheduler.sigmas.to(device=accelerator.device, dtype=dtype)
                            schedule_timesteps = noise_scheduler.timesteps.to(accelerator.device)
                            timesteps = timesteps.to(accelerator.device)
                            step_indices = [(schedule_timesteps == t).nonzero().item() for t in timesteps]

                            sigma = sigmas[step_indices].flatten()
                            while len(sigma.shape) < n_dim:
                                sigma = sigma.unsqueeze(-1)
                            return sigma

                        u = compute_density_for_timestep_sampling(
                            weighting_scheme=args.weighting_scheme,
                            batch_size=bsz,
                            logit_mean=args.logit_mean,
                            logit_std=args.logit_std,
                            mode_scale=args.mode_scale,
                        )
                        indices = (u * noise_scheduler.config.num_train_timesteps).long()
                        timesteps = noise_scheduler.timesteps[indices].to(device=latents.device)

                        # Add noise according to flow matching.
                        # zt = (1 - texp) * x + texp * z1
                        sigmas = get_sigmas(timesteps, n_dim=latents.ndim, dtype=latents.dtype)
                        noisy_latents = (1.0 - sigmas) * latents + sigmas * noise

                        # Add noise
                        target = noise - latents

                    # Predict the noise residual
                    noise_pred = transformer3d(
                        noisy_latents,
                        timesteps.to(noisy_latents.dtype),
                        encoder_hidden_states=prompt_embeds,
                        text_embedding_mask=prompt_attention_mask,
                        encoder_hidden_states_t5=prompt_embeds_2,
                        text_embedding_mask_t5=prompt_attention_mask_2,
                        image_meta_size=add_time_ids,
                        style=style,
                        image_rotary_emb=image_rotary_emb,
                        inpaint_latents=inpaint_latents if args.train_mode != "normal" else None,
                        clip_encoder_hidden_states=clip_encoder_hidden_states if args.train_mode != "normal" else None,
                        clip_attention_mask=clip_attention_mask if args.train_mode != "normal" else None,
                        return_dict=False
                    )[0]
                    if noise_pred.size()[1] != vae.config.latent_channels:
                        noise_pred, _ = noise_pred.chunk(2, dim=1)

                    def custom_mse_loss(noise_pred, target, weighting=None, threshold=50):
                        noise_pred = noise_pred.float()
                        target = target.float()
                        diff = noise_pred - target
                        mse_loss = F.mse_loss(noise_pred, target, reduction='none')
                        mask = (diff.abs() <= threshold).float()
                        masked_loss = mse_loss * mask
                        if weighting is not None:
                            masked_loss = masked_loss * weighting
                        final_loss = masked_loss.mean()
                        return final_loss
                    
                    if args.loss_type == "ddpm":
                        loss = custom_mse_loss(noise_pred.float(), target.float())
                    else:
                        weighting = compute_loss_weighting_for_sd3(weighting_scheme=args.weighting_scheme, sigmas=sigmas)
                        loss = custom_mse_loss(noise_pred.float(), target.float(), weighting.float())
                        loss = loss.mean()

                    if args.motion_sub_loss and noise_pred.size()[2] > 2:
                        gt_sub_noise = noise_pred[:, :, 1:].float() - noise_pred[:, :, :-1].float()
                        pre_sub_noise = target[:, :, 1:].float() - target[:, :, :-1].float()
                        sub_loss = F.mse_loss(gt_sub_noise, pre_sub_noise, reduction="mean")
                        loss = loss * (1 - args.motion_sub_loss_ratio) + sub_loss * args.motion_sub_loss_ratio
                else:
                    added_cond_kwargs = {"resolution": None, "aspect_ratio": None}
                    if unwrap_model(transformer3d).config.sample_size == 128:
                        bs, height, width = bsz, batch["pixel_values"].size()[-2], batch["pixel_values"].size()[-1]
                        resolution = torch.tensor([height, width]).repeat(bs, 1)
                        aspect_ratio = torch.tensor([float(height / width)]).repeat(bs, 1)
                        resolution = resolution.to(accelerator.device, dtype=weight_dtype)
                        aspect_ratio = aspect_ratio.to(accelerator.device, dtype=weight_dtype)
                        added_cond_kwargs = {"resolution": resolution, "aspect_ratio": aspect_ratio}

                    loss_term = train_diffusion.training_losses(
                        transformer3d, 
                        latents, 
                        timesteps, 
                        noise=noise,
                        model_kwargs=dict(
                            encoder_hidden_states=prompt_embeds, 
                            encoder_attention_mask=prompt_attention_mask, 
                            added_cond_kwargs=added_cond_kwargs, 
                            inpaint_latents=inpaint_latents if args.train_mode != "normal" else None,
                            clip_encoder_hidden_states=clip_encoder_hidden_states if args.train_mode != "normal" else None,
                            clip_attention_mask=clip_attention_mask if args.train_mode != "normal" else None,
                            return_dict=False
                        )
                    )
                    loss = loss_term['loss'].mean()

                # Gather the losses across all processes for logging (if we use distributed training).
                avg_loss = accelerator.gather(loss.repeat(args.train_batch_size)).mean()
                train_loss += avg_loss.item() / args.gradient_accumulation_steps

                # Backpropagate
                accelerator.backward(loss)
                if accelerator.sync_gradients:
                    if not args.use_deepspeed:
                        trainable_params_grads = [p.grad for p in trainable_params if p.grad is not None]
                        trainable_params_total_norm = torch.norm(torch.stack([torch.norm(g.detach(), 2) for g in trainable_params_grads]), 2)
                        max_grad_norm = linear_decay(args.max_grad_norm * args.initial_grad_norm_ratio, args.max_grad_norm, args.abnormal_norm_clip_start, global_step)
                        if trainable_params_total_norm / max_grad_norm > 5 and global_step > args.abnormal_norm_clip_start:
                            actual_max_grad_norm = max_grad_norm / min((trainable_params_total_norm / max_grad_norm), 10)
                        else:
                            actual_max_grad_norm = max_grad_norm
                    else:
                        actual_max_grad_norm = args.max_grad_norm

                    if not args.use_deepspeed and args.report_model_info and accelerator.is_main_process:
                        if trainable_params_total_norm > 1 and global_step > args.abnormal_norm_clip_start:
                            for name, param in transformer3d.named_parameters():
                                if param.requires_grad:
                                    writer.add_scalar(f'gradients/before_clip_norm/{name}', param.grad.norm(), global_step=global_step)

                    norm_sum = accelerator.clip_grad_norm_(trainable_params, actual_max_grad_norm)
                    if not args.use_deepspeed and args.report_model_info and accelerator.is_main_process:
                        writer.add_scalar(f'gradients/norm_sum', norm_sum, global_step=global_step)
                        writer.add_scalar(f'gradients/actual_max_grad_norm', actual_max_grad_norm, global_step=global_step)
                optimizer.step()
                lr_scheduler.step()
                optimizer.zero_grad()

            # Checks if the accelerator has performed an optimization step behind the scenes
            if accelerator.sync_gradients:
                progress_bar.update(1)
                global_step += 1
                accelerator.log({"train_loss": train_loss}, step=global_step)
                train_loss = 0.0

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

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

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

                                for removing_checkpoint in removing_checkpoints:
                                    removing_checkpoint = os.path.join(args.output_dir, removing_checkpoint)
                                    shutil.rmtree(removing_checkpoint)
                        if not args.save_state:
                            safetensor_save_path = os.path.join(args.output_dir, f"checkpoint-{global_step}.safetensors")
                            save_model(safetensor_save_path, accelerator.unwrap_model(network))
                            logger.info(f"Saved safetensor to {safetensor_save_path}")
                        else:
                            accelerator_save_path = os.path.join(args.output_dir, f"checkpoint-{global_step}")
                            accelerator.save_state(accelerator_save_path)
                            logger.info(f"Saved state to {accelerator_save_path}")

                if accelerator.is_main_process:
                    if args.validation_prompts is not None and global_step % args.validation_steps == 0:
                        log_validation(
                            vae,
                            text_encoder,
                            text_encoder_2,
                            tokenizer,
                            tokenizer_2,
                            transformer3d,
                            image_encoder,
                            image_processor,
                            network,
                            config,
                            args,
                            accelerator,
                            weight_dtype,
                            global_step,
                        )

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

            if global_step >= args.max_train_steps:
                break

        if accelerator.is_main_process:
            if args.validation_prompts is not None and epoch % args.validation_epochs == 0:
                log_validation(
                    vae,
                    text_encoder,
                    text_encoder_2,
                    tokenizer,
                    tokenizer_2,
                    transformer3d,
                    image_encoder,
                    image_processor,
                    network,
                    config,
                    args,
                    accelerator,
                    weight_dtype,
                    global_step,
                )

    # Create the pipeline using the trained modules and save it.
    accelerator.wait_for_everyone()
    if accelerator.is_main_process:
        safetensor_save_path = os.path.join(args.output_dir, f"checkpoint-{global_step}.safetensors")
        accelerator_save_path = os.path.join(args.output_dir, f"checkpoint-{global_step}")
        save_model(safetensor_save_path, accelerator.unwrap_model(network))
        if args.save_state:
            accelerator.save_state(accelerator_save_path)
        logger.info(f"Saved state to {accelerator_save_path}")

    accelerator.end_training()


if __name__ == "__main__":
    main()