2302 lines
108 KiB
Python
2302 lines
108 KiB
Python
"""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 gc
|
|
import logging
|
|
import math
|
|
import os
|
|
import pickle
|
|
import random
|
|
import shutil
|
|
import sys
|
|
|
|
import accelerate
|
|
import diffusers
|
|
import numpy as np
|
|
import torch
|
|
import torch.nn.functional as F
|
|
import torchaudio
|
|
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, FlowMatchEulerDiscreteScheduler
|
|
from diffusers.optimization import get_scheduler
|
|
from diffusers.training_utils import (EMAModel,
|
|
compute_density_for_timestep_sampling,
|
|
compute_loss_weighting_for_sd3)
|
|
from diffusers.utils import check_min_version, deprecate, is_wandb_available
|
|
from diffusers.utils.torch_utils import is_compiled_module
|
|
from einops import rearrange
|
|
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.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)), os.path.dirname(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 videox_fun.data import (ASPECT_RATIO_512, ASPECT_RATIO_RANDOM_CROP_512,
|
|
ASPECT_RATIO_RANDOM_CROP_PROB,
|
|
AspectRatioBatchImageVideoSampler,
|
|
ImageVideoDataset, ImageVideoSampler,
|
|
RandomSampler, VideoSpeechDataset,
|
|
get_closest_ratio, get_random_mask)
|
|
from videox_fun.models import (AutoencoderKLMOVAAudio, AutoencoderKLWan,
|
|
AutoTokenizer, MOVADualTowerConditionalBridge,
|
|
MOVAModel, UMT5EncoderModel,
|
|
WanAudioTransformer3DModel,
|
|
WanTransformer3DModel)
|
|
from videox_fun.pipeline import MOVAPipeline
|
|
from videox_fun.utils.discrete_sampler import DiscreteSampling
|
|
from videox_fun.utils.utils import (calculate_dimensions, get_image_latent,
|
|
get_image_to_video_latent,
|
|
save_videos_grid,
|
|
save_videos_with_audio_grid)
|
|
|
|
if is_wandb_available():
|
|
import wandb
|
|
|
|
|
|
def filter_kwargs(cls, kwargs):
|
|
import inspect
|
|
sig = inspect.signature(cls.__init__)
|
|
valid_params = set(sig.parameters.keys()) - {'self', 'cls'}
|
|
filtered_kwargs = {k: v for k, v in kwargs.items() if k in valid_params}
|
|
return filtered_kwargs
|
|
|
|
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 basic_clean(text):
|
|
"""Clean text following MOVA pipeline convention."""
|
|
import html
|
|
|
|
import ftfy
|
|
text = ftfy.fix_text(text)
|
|
text = html.unescape(html.unescape(text))
|
|
return text.strip()
|
|
|
|
|
|
def whitespace_clean(text):
|
|
"""Clean whitespace following MOVA pipeline convention."""
|
|
import re
|
|
text = re.sub(r"\s+", " ", text)
|
|
text = text.strip()
|
|
return text
|
|
|
|
# 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, audio_vae, text_encoder, tokenizer, transformer, transformer_2, transformer_audio, dual_tower_bridge, mova_model, args, accelerator, weight_dtype, global_step):
|
|
try:
|
|
# Unwrap models if needed
|
|
if type(mova_model).__name__ == 'DistributedDataParallel':
|
|
mova_model = accelerator.unwrap_model(mova_model)
|
|
|
|
print(f"[Validation] boundary_type={args.boundary_type}, transformer={transformer is not None}, transformer_2={transformer_2 is not None}")
|
|
|
|
# For boundary_type == "low" or "high", we need to load the missing transformer for validation
|
|
temp_transformer = None
|
|
temp_transformer_2 = None
|
|
|
|
if args.boundary_type == "high" and transformer is None:
|
|
# Load transformer (low noise) for validation
|
|
print("Loading transformer (low noise) for validation...")
|
|
temp_transformer = WanTransformer3DModel.from_pretrained(
|
|
args.pretrained_model_name_or_path,
|
|
subfolder="video_dit_2",
|
|
low_cpu_mem_usage=True,
|
|
torch_dtype=weight_dtype,
|
|
).to(accelerator.device)
|
|
mova_model.set_module(temp_transformer, "transformer")
|
|
print(f"[Validation] After loading: transformer device={temp_transformer.device}")
|
|
elif args.boundary_type == "low" and transformer_2 is None:
|
|
# Load transformer_2 (high noise) for validation
|
|
print("Loading transformer_2 (high noise) for validation...")
|
|
temp_transformer_2 = WanTransformer3DModel.from_pretrained(
|
|
args.pretrained_model_name_or_path,
|
|
subfolder="video_dit",
|
|
low_cpu_mem_usage=True,
|
|
torch_dtype=weight_dtype,
|
|
).to(accelerator.device)
|
|
mova_model.set_module(temp_transformer_2, "transformer_2")
|
|
print(f"[Validation] After loading: transformer_2 device={temp_transformer_2.device}")
|
|
|
|
# Use the correct transformers for validation
|
|
val_transformer = transformer if transformer is not None else temp_transformer
|
|
val_transformer_2 = transformer_2 if transformer_2 is not None else temp_transformer_2
|
|
|
|
with torch.no_grad(), torch.cuda.amp.autocast(dtype=weight_dtype), torch.cuda.device(device=accelerator.device):
|
|
logger.info("Running validation... ")
|
|
scheduler = FlowMatchEulerDiscreteScheduler.from_pretrained(
|
|
args.pretrained_model_name_or_path,
|
|
subfolder="scheduler"
|
|
)
|
|
|
|
pipeline = MOVAPipeline(
|
|
vae=vae,
|
|
audio_vae=audio_vae,
|
|
text_encoder=text_encoder,
|
|
tokenizer=tokenizer,
|
|
scheduler=scheduler,
|
|
transformer=val_transformer,
|
|
transformer_2=val_transformer_2,
|
|
transformer_audio=transformer_audio,
|
|
dual_tower_bridge=dual_tower_bridge,
|
|
audio_vae_type="dac",
|
|
)
|
|
pipeline.mova_model = mova_model
|
|
pipeline = pipeline.to(accelerator.device)
|
|
|
|
if args.seed is None:
|
|
generator = None
|
|
else:
|
|
rank_seed = args.seed + accelerator.process_index
|
|
generator = torch.Generator(device=accelerator.device).manual_seed(rank_seed)
|
|
logger.info(f"Rank {accelerator.process_index} using seed: {rank_seed}")
|
|
|
|
for i in range(len(args.validation_prompts)):
|
|
# I2V mode: load input image (MOVA only supports I2V)
|
|
if args.validation_paths is None or i >= len(args.validation_paths):
|
|
raise ValueError(
|
|
f"MOVA only supports I2V (image-to-video). "
|
|
f"Please provide --validation_paths for each validation prompt. "
|
|
f"Missing path for prompt {i}: {args.validation_prompts[i]}"
|
|
)
|
|
|
|
start_image = Image.open(args.validation_paths[i])
|
|
width, height = start_image.width, start_image.height
|
|
width, height = calculate_dimensions(args.video_sample_size * args.video_sample_size, width / height)
|
|
|
|
logger.info(f"Running I2V validation with image: {args.validation_paths[i]} ({width}x{height})")
|
|
|
|
output = pipeline(
|
|
prompt=args.validation_prompts[i],
|
|
image=start_image,
|
|
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指",
|
|
height=height,
|
|
width=width,
|
|
num_frames=args.video_sample_n_frames,
|
|
generator=generator,
|
|
num_inference_steps=25,
|
|
guidance_scale=4.5,
|
|
boundary=args.boundary_ratio,
|
|
)
|
|
|
|
sample = output.videos
|
|
audio = output.audio
|
|
os.makedirs(os.path.join(args.output_dir, "sample"), exist_ok=True)
|
|
save_videos_with_audio_grid(
|
|
sample,
|
|
audio,
|
|
os.path.join(
|
|
args.output_dir,
|
|
f"sample/sample-{global_step}-rank{accelerator.process_index}-image-{i}.mp4"
|
|
),
|
|
fps=24,
|
|
audio_sample_rate=getattr(audio_vae.config, 'sample_rate', 24000),
|
|
)
|
|
|
|
del pipeline
|
|
gc.collect()
|
|
torch.cuda.empty_cache()
|
|
torch.cuda.ipc_collect()
|
|
|
|
# Clean up temporarily loaded transformers for validation
|
|
if temp_transformer is not None:
|
|
print("Cleaning up temp_transformer (low noise) after validation...")
|
|
temp_transformer.to('cpu')
|
|
del temp_transformer
|
|
temp_transformer = None
|
|
mova_model.transformer = None
|
|
if temp_transformer_2 is not None:
|
|
print("Cleaning up temp_transformer_2 (high noise) after validation...")
|
|
temp_transformer_2.to('cpu')
|
|
del temp_transformer_2
|
|
temp_transformer_2 = None
|
|
mova_model.transformer_2 = None
|
|
|
|
gc.collect()
|
|
torch.cuda.empty_cache()
|
|
vae.to(accelerator.device if not args.low_vram else "cpu", dtype=weight_dtype)
|
|
audio_vae.to(accelerator.device if not args.low_vram else "cpu")
|
|
if not args.enable_text_encoder_in_dataloader:
|
|
text_encoder.to(accelerator.device if not args.low_vram else "cpu", dtype=weight_dtype)
|
|
|
|
except Exception as e:
|
|
gc.collect()
|
|
torch.cuda.empty_cache()
|
|
torch.cuda.ipc_collect()
|
|
print(f"Eval error on rank {accelerator.process_index} with info {e}")
|
|
|
|
# Clean up temporarily loaded transformers on error
|
|
try:
|
|
if 'temp_transformer' in locals() and temp_transformer is not None:
|
|
if hasattr(temp_transformer, 'device') and temp_transformer.device.type == 'cuda':
|
|
temp_transformer.to('cpu')
|
|
del temp_transformer
|
|
temp_transformer = None
|
|
mova_model.transformer = None
|
|
if 'temp_transformer_2' in locals() and temp_transformer_2 is not None:
|
|
if hasattr(temp_transformer_2, 'device') and temp_transformer_2.device.type == 'cuda':
|
|
temp_transformer_2.to('cpu')
|
|
del temp_transformer_2
|
|
temp_transformer_2 = None
|
|
mova_model.transformer_2 = None
|
|
except:
|
|
pass
|
|
|
|
gc.collect()
|
|
torch.cuda.empty_cache()
|
|
vae.to(accelerator.device if not args.low_vram else "cpu", dtype=weight_dtype)
|
|
audio_vae.to(accelerator.device if not args.low_vram else "cpu")
|
|
if not args.enable_text_encoder_in_dataloader:
|
|
text_encoder.to(accelerator.device if not args.low_vram else "cpu", dtype=weight_dtype)
|
|
|
|
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(
|
|
"--validation_paths",
|
|
type=str,
|
|
default=None,
|
|
nargs="+",
|
|
help=("A set of control videos 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("--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(
|
|
"--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(
|
|
"--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(
|
|
"--auto_tile_batch_size", action="store_true", help="Whether to auto tile batch size.",
|
|
)
|
|
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(
|
|
"--train_sampling_steps",
|
|
type=int,
|
|
default=1000,
|
|
help="Run train_sampling_steps.",
|
|
)
|
|
parser.add_argument(
|
|
"--keep_all_node_same_token_length",
|
|
action="store_true",
|
|
help="Reference of the length token.",
|
|
)
|
|
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 image.",
|
|
)
|
|
parser.add_argument(
|
|
"--fix_sample_size",
|
|
nargs=2, type=int, default=None,
|
|
help="Fix Sample size [height, width] when using bucket and collate_fn."
|
|
)
|
|
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(
|
|
"--transformer_path",
|
|
type=str,
|
|
default=None,
|
|
help=("If you want to load the weight from other transformers, input its path."),
|
|
)
|
|
parser.add_argument(
|
|
"--transformer_high_path",
|
|
type=str,
|
|
default=None,
|
|
help=("If you want to load the weight from other transformers for transformer_2 (high noise), 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(
|
|
'--trainable_modules',
|
|
nargs='+',
|
|
help='Enter a list of trainable modules'
|
|
)
|
|
parser.add_argument(
|
|
'--boundary_ratio',
|
|
type=float,
|
|
default=0.9,
|
|
help='Boundary ratio for switching between high-noise and low-noise DiT. Timesteps below this ratio use low-noise DiT.'
|
|
)
|
|
parser.add_argument(
|
|
"--boundary_type",
|
|
type=str,
|
|
default="full",
|
|
choices=["low", "high", "full"],
|
|
help=(
|
|
'Which DiT to train. "low" = only low-noise DiT, '
|
|
'"high" = only high-noise DiT, "full" = both DiTs.'
|
|
),
|
|
)
|
|
parser.add_argument(
|
|
"--train_components",
|
|
type=str,
|
|
default="all",
|
|
help=(
|
|
'Which components to train. Comma-separated list of: '
|
|
'"transformer", "transformer_2", "transformer_audio", "dual_tower_bridge", or "all". '
|
|
'Example: "transformer,dual_tower_bridge" to train only transformer and bridge.'
|
|
),
|
|
)
|
|
parser.add_argument(
|
|
'--trainable_modules_low_learning_rate',
|
|
nargs='+',
|
|
default=[],
|
|
help='Enter a list of trainable modules with lower learning rate'
|
|
)
|
|
parser.add_argument(
|
|
'--tokenizer_max_length',
|
|
type=int,
|
|
default=512,
|
|
help='Max length of tokenizer'
|
|
)
|
|
parser.add_argument(
|
|
"--use_deepspeed", action="store_true", help="Whether or not to use deepspeed."
|
|
)
|
|
parser.add_argument(
|
|
"--use_fsdp", action="store_true", help="Whether or not to use fsdp."
|
|
)
|
|
parser.add_argument(
|
|
"--low_vram", action="store_true", help="Whether enable low_vram mode."
|
|
)
|
|
parser.add_argument(
|
|
"--i2v_ratio",
|
|
type=float,
|
|
default=0.5,
|
|
help=(
|
|
'Ratio of I2V samples in training. 0.0 = pure T2V, 1.0 = pure I2V, '
|
|
'0.5 = 50%% T2V + 50%% I2V (default).'
|
|
),
|
|
)
|
|
parser.add_argument(
|
|
"--i2v_noise_scale",
|
|
type=float,
|
|
default=0.0,
|
|
help=(
|
|
'Noise scale for I2V first frame conditioning. '
|
|
'0.0 means first frame is kept clean (default). '
|
|
'Higher values add slight noise to the condition frame.'
|
|
),
|
|
)
|
|
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)
|
|
|
|
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,
|
|
)
|
|
|
|
deepspeed_plugin = accelerator.state.deepspeed_plugin if hasattr(accelerator.state, "deepspeed_plugin") else None
|
|
fsdp_plugin = accelerator.state.fsdp_plugin if hasattr(accelerator.state, "fsdp_plugin") else None
|
|
if deepspeed_plugin is not None:
|
|
zero_stage = int(deepspeed_plugin.zero_stage)
|
|
fsdp_stage = 0
|
|
print(f"Using DeepSpeed Zero stage: {zero_stage}")
|
|
|
|
args.use_deepspeed = True
|
|
if zero_stage == 3:
|
|
print(f"Auto set save_state to True because zero_stage == 3")
|
|
args.save_state = True
|
|
elif fsdp_plugin is not None:
|
|
from torch.distributed.fsdp import ShardingStrategy
|
|
zero_stage = 0
|
|
if fsdp_plugin.sharding_strategy is ShardingStrategy.FULL_SHARD:
|
|
fsdp_stage = 3
|
|
elif fsdp_plugin.sharding_strategy is None: # The fsdp_plugin.sharding_strategy is None in FSDP 2.
|
|
fsdp_stage = 3
|
|
elif fsdp_plugin.sharding_strategy is ShardingStrategy.SHARD_GRAD_OP:
|
|
fsdp_stage = 2
|
|
else:
|
|
fsdp_stage = 0
|
|
print(f"Using FSDP stage: {fsdp_stage}")
|
|
|
|
args.use_fsdp = True
|
|
if fsdp_stage == 3:
|
|
print(f"Auto set save_state to True because fsdp_stage == 3")
|
|
args.save_state = True
|
|
else:
|
|
zero_stage = 0
|
|
fsdp_stage = 0
|
|
print("DeepSpeed is not enabled.")
|
|
|
|
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 mova_model) 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.
|
|
noise_scheduler = FlowMatchEulerDiscreteScheduler.from_pretrained(
|
|
args.pretrained_model_name_or_path,
|
|
subfolder="scheduler"
|
|
)
|
|
|
|
# Get Tokenizer
|
|
tokenizer = AutoTokenizer.from_pretrained(
|
|
args.pretrained_model_name_or_path,
|
|
subfolder="tokenizer",
|
|
)
|
|
|
|
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()):
|
|
# Get Text encoder
|
|
text_encoder = UMT5EncoderModel.from_pretrained(
|
|
args.pretrained_model_name_or_path,
|
|
subfolder="text_encoder",
|
|
low_cpu_mem_usage=True,
|
|
torch_dtype=weight_dtype,
|
|
)
|
|
text_encoder = text_encoder.eval()
|
|
# Get Vae
|
|
vae = AutoencoderKLWan.from_pretrained(
|
|
os.path.join(args.pretrained_model_name_or_path, "video_vae/diffusion_pytorch_model.safetensors")
|
|
).to(weight_dtype)
|
|
vae.eval()
|
|
audio_vae = AutoencoderKLMOVAAudio.from_pretrained(
|
|
args.pretrained_model_name_or_path,
|
|
subfolder="audio_vae",
|
|
torch_dtype=torch.float32,
|
|
)
|
|
audio_vae.eval()
|
|
|
|
# Get MOVA Model components
|
|
# Load transformers based on boundary_type (similar to wan2.2 training)
|
|
# Convention: transformer = low-noise (video_dit_2), transformer_2 = high-noise (video_dit)
|
|
if args.boundary_type == "high" or args.boundary_type == "full":
|
|
print("Loading Video DiT 2 (High Noise) with WanTransformer3DModel...")
|
|
transformer_2 = WanTransformer3DModel.from_pretrained(
|
|
args.pretrained_model_name_or_path,
|
|
subfolder="video_dit",
|
|
low_cpu_mem_usage=True,
|
|
torch_dtype=weight_dtype,
|
|
)
|
|
else:
|
|
transformer_2 = None
|
|
|
|
if args.boundary_type == "low" or args.boundary_type == "full":
|
|
print("Loading Video DiT (Low Noise) with WanTransformer3DModel...")
|
|
transformer = WanTransformer3DModel.from_pretrained(
|
|
args.pretrained_model_name_or_path,
|
|
subfolder="video_dit_2",
|
|
low_cpu_mem_usage=True,
|
|
torch_dtype=weight_dtype,
|
|
)
|
|
else:
|
|
transformer = None
|
|
|
|
# Print which models are loaded
|
|
if args.boundary_type == "low":
|
|
print("Training mode: LOW-NOISE only (transformer loaded)")
|
|
elif args.boundary_type == "high":
|
|
print("Training mode: HIGH-NOISE only (transformer_2 loaded)")
|
|
else:
|
|
print("Training mode: FULL (both transformers loaded)")
|
|
|
|
print("Loading Audio DiT with WanAudioTransformer3DModel...")
|
|
transformer_audio = WanAudioTransformer3DModel.from_pretrained(
|
|
args.pretrained_model_name_or_path,
|
|
subfolder="audio_dit",
|
|
low_cpu_mem_usage=True,
|
|
torch_dtype=weight_dtype,
|
|
)
|
|
|
|
print("Loading Dual Tower Bridge...")
|
|
dual_tower_bridge = MOVADualTowerConditionalBridge.from_pretrained(
|
|
args.pretrained_model_name_or_path,
|
|
subfolder="dual_tower_bridge",
|
|
low_cpu_mem_usage=True,
|
|
torch_dtype=weight_dtype,
|
|
)
|
|
|
|
# Build MOVAModel
|
|
print("Building MOVAModel...")
|
|
mova_model = MOVAModel(
|
|
transformer=transformer,
|
|
transformer_2=transformer_2,
|
|
transformer_audio=transformer_audio,
|
|
dual_tower_bridge=dual_tower_bridge,
|
|
)
|
|
mova_model = mova_model.to(weight_dtype)
|
|
|
|
# Freeze vae and text_encoder and set models to trainable
|
|
vae.requires_grad_(False)
|
|
audio_vae.requires_grad_(False)
|
|
text_encoder.requires_grad_(False)
|
|
|
|
# Freeze all MOVAModel components first
|
|
if transformer is not None:
|
|
transformer.requires_grad_(False)
|
|
if transformer_2 is not None:
|
|
transformer_2.requires_grad_(False)
|
|
transformer_audio.requires_grad_(False)
|
|
dual_tower_bridge.requires_grad_(False)
|
|
|
|
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
|
|
|
|
if transformer is not None:
|
|
m, u = transformer.load_state_dict(state_dict, strict=False)
|
|
|
|
if args.transformer_high_path is not None:
|
|
print(f"Loading transformer_2 (high noise) from checkpoint: {args.transformer_high_path}")
|
|
if args.transformer_high_path.endswith("safetensors"):
|
|
from safetensors.torch import load_file, safe_open
|
|
state_dict = load_file(args.transformer_high_path)
|
|
else:
|
|
state_dict = torch.load(args.transformer_high_path, map_location="cpu")
|
|
state_dict = state_dict["state_dict"] if "state_dict" in state_dict else state_dict
|
|
|
|
if transformer_2 is not None:
|
|
m, u = transformer_2.load_state_dict(state_dict, strict=False)
|
|
print(f"transformer_2 (high noise) - 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)}")
|
|
assert len(u) == 0
|
|
|
|
# A good trainable modules is showed below now.
|
|
# For 3D Patch: trainable_modules = ['ff.net', 'pos_embed', 'attn2', 'proj_out', 'timepositionalencoding', 'h_position', 'w_position']
|
|
# For 2D Patch: trainable_modules = ['ff.net', 'attn2', 'timepositionalencoding', 'h_position', 'w_position']
|
|
if transformer is not None:
|
|
transformer.train()
|
|
if transformer_2 is not None:
|
|
transformer_2.train()
|
|
transformer_audio.train()
|
|
dual_tower_bridge.train()
|
|
|
|
if accelerator.is_main_process:
|
|
accelerator.print(
|
|
f"Trainable modules '{args.trainable_modules}'."
|
|
)
|
|
|
|
# Parse train_components
|
|
if args.train_components == "all":
|
|
components_to_train = ["transformer", "transformer_2", "transformer_audio", "dual_tower_bridge"]
|
|
else:
|
|
components_to_train = [c.strip() for c in args.train_components.split(",")]
|
|
|
|
if accelerator.is_main_process:
|
|
accelerator.print(f"Training components: {components_to_train}")
|
|
|
|
# Set trainable parameters for each component based on train_components
|
|
trainable_components = []
|
|
if transformer is not None and "transformer" in components_to_train:
|
|
trainable_components.append(("transformer", transformer))
|
|
if transformer_2 is not None and "transformer_2" in components_to_train:
|
|
trainable_components.append(("transformer_2", transformer_2))
|
|
if "transformer_audio" in components_to_train:
|
|
trainable_components.append(("transformer_audio", transformer_audio))
|
|
if "dual_tower_bridge" in components_to_train:
|
|
trainable_components.append(("dual_tower_bridge", dual_tower_bridge))
|
|
|
|
for component_name, component in trainable_components:
|
|
matched_params = 0
|
|
for name, param in component.named_parameters():
|
|
for trainable_module_name in args.trainable_modules + args.trainable_modules_low_learning_rate:
|
|
if trainable_module_name in name:
|
|
param.requires_grad = True
|
|
matched_params += 1
|
|
break
|
|
if accelerator.is_main_process:
|
|
print(f"[Trainable] {component_name}: {matched_params} parameters matched {args.trainable_modules}")
|
|
|
|
# Create EMA for the components.
|
|
if args.use_ema:
|
|
if zero_stage == 3:
|
|
raise NotImplementedError("FSDP does not support EMA.")
|
|
|
|
# Create EMA models based on boundary_type and train_components
|
|
ema_transformer = None
|
|
ema_transformer_2 = None
|
|
|
|
if args.boundary_type == "high" or args.boundary_type == "full":
|
|
if "transformer_2" in components_to_train:
|
|
ema_transformer_2 = WanTransformer3DModel.from_pretrained(
|
|
args.pretrained_model_name_or_path, subfolder="video_dit"
|
|
).to(weight_dtype)
|
|
|
|
if args.boundary_type == "low" or args.boundary_type == "full":
|
|
if "transformer" in components_to_train:
|
|
ema_transformer = WanTransformer3DModel.from_pretrained(
|
|
args.pretrained_model_name_or_path, subfolder="video_dit_2"
|
|
).to(weight_dtype)
|
|
|
|
if "transformer_audio" in components_to_train:
|
|
ema_transformer_audio = WanAudioTransformer3DModel.from_pretrained(
|
|
args.pretrained_model_name_or_path, subfolder="audio_dit"
|
|
).to(weight_dtype)
|
|
|
|
if "dual_tower_bridge" in components_to_train:
|
|
ema_dual_tower_bridge = MOVADualTowerConditionalBridge.from_pretrained(
|
|
args.pretrained_model_name_or_path, subfolder="dual_tower_bridge"
|
|
).to(weight_dtype)
|
|
|
|
# Collect parameters for EMA
|
|
ema_params = []
|
|
if ema_transformer is not None:
|
|
ema_params.extend(ema_transformer.parameters())
|
|
if ema_transformer_2 is not None:
|
|
ema_params.extend(ema_transformer_2.parameters())
|
|
if "transformer_audio" in components_to_train:
|
|
ema_params.extend(ema_transformer_audio.parameters())
|
|
if "dual_tower_bridge" in components_to_train:
|
|
ema_params.extend(ema_dual_tower_bridge.parameters())
|
|
|
|
ema_mova_model = EMAModel(ema_params)
|
|
ema_mova_model = ema_mova_model.to(weight_dtype)
|
|
|
|
# `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
|
|
if fsdp_stage != 0 or zero_stage == 3:
|
|
def save_model_hook(models, weights, output_dir):
|
|
from safetensors.torch import save_file
|
|
|
|
# Save models based on components_to_train (split saving)
|
|
# NOTE: accelerator.get_state_dict must be called by ALL processes, not just main
|
|
# Otherwise it will hang due to collective communication
|
|
if transformer is not None and "transformer" in components_to_train:
|
|
accelerate_state_dict = accelerator.get_state_dict(transformer, unwrap=True)
|
|
if accelerator.is_main_process:
|
|
accelerate_state_dict = {k: v.to(dtype=weight_dtype) for k, v in accelerate_state_dict.items()}
|
|
safetensor_save_path = os.path.join(output_dir, "transformer", "diffusion_pytorch_model.safetensors")
|
|
os.makedirs(os.path.dirname(safetensor_save_path), exist_ok=True)
|
|
save_file(accelerate_state_dict, safetensor_save_path, metadata={"format": "pt"})
|
|
|
|
if transformer_2 is not None and "transformer_2" in components_to_train:
|
|
accelerate_state_dict = accelerator.get_state_dict(transformer_2, unwrap=True)
|
|
if accelerator.is_main_process:
|
|
accelerate_state_dict = {k: v.to(dtype=weight_dtype) for k, v in accelerate_state_dict.items()}
|
|
safetensor_save_path = os.path.join(output_dir, "transformer_2", "diffusion_pytorch_model.safetensors")
|
|
os.makedirs(os.path.dirname(safetensor_save_path), exist_ok=True)
|
|
save_file(accelerate_state_dict, safetensor_save_path, metadata={"format": "pt"})
|
|
|
|
if "transformer_audio" in components_to_train:
|
|
accelerate_state_dict = accelerator.get_state_dict(transformer_audio, unwrap=True)
|
|
if accelerator.is_main_process:
|
|
accelerate_state_dict = {k: v.to(dtype=weight_dtype) for k, v in accelerate_state_dict.items()}
|
|
safetensor_save_path = os.path.join(output_dir, "transformer_audio", "diffusion_pytorch_model.safetensors")
|
|
os.makedirs(os.path.dirname(safetensor_save_path), exist_ok=True)
|
|
save_file(accelerate_state_dict, safetensor_save_path, metadata={"format": "pt"})
|
|
|
|
if "dual_tower_bridge" in components_to_train:
|
|
accelerate_state_dict = accelerator.get_state_dict(dual_tower_bridge, unwrap=True)
|
|
if accelerator.is_main_process:
|
|
accelerate_state_dict = {k: v.to(dtype=weight_dtype) for k, v in accelerate_state_dict.items()}
|
|
safetensor_save_path = os.path.join(output_dir, "dual_tower_bridge", "diffusion_pytorch_model.safetensors")
|
|
os.makedirs(os.path.dirname(safetensor_save_path), exist_ok=True)
|
|
save_file(accelerate_state_dict, safetensor_save_path, metadata={"format": "pt"})
|
|
|
|
if accelerator.is_main_process:
|
|
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}.")
|
|
else:
|
|
# 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:
|
|
# Save EMA models (only for trained components)
|
|
if args.use_ema:
|
|
if ema_transformer is not None and "transformer" in components_to_train:
|
|
ema_transformer.save_pretrained(os.path.join(output_dir, "ema_transformer"))
|
|
if ema_transformer_2 is not None and "transformer_2" in components_to_train:
|
|
ema_transformer_2.save_pretrained(os.path.join(output_dir, "ema_transformer_2"))
|
|
if "transformer_audio" in components_to_train:
|
|
ema_transformer_audio.save_pretrained(os.path.join(output_dir, "ema_transformer_audio"))
|
|
if "dual_tower_bridge" in components_to_train:
|
|
ema_dual_tower_bridge.save_pretrained(os.path.join(output_dir, "ema_dual_tower_bridge"))
|
|
|
|
# Save main models (only for trained components)
|
|
if transformer is not None and "transformer" in components_to_train:
|
|
transformer.save_pretrained(os.path.join(output_dir, "transformer"))
|
|
if transformer_2 is not None and "transformer_2" in components_to_train:
|
|
transformer_2.save_pretrained(os.path.join(output_dir, "transformer_2"))
|
|
if "transformer_audio" in components_to_train:
|
|
transformer_audio.save_pretrained(os.path.join(output_dir, "transformer_audio"))
|
|
if "dual_tower_bridge" in components_to_train:
|
|
dual_tower_bridge.save_pretrained(os.path.join(output_dir, "dual_tower_bridge"))
|
|
|
|
if not args.use_deepspeed:
|
|
weights.pop()
|
|
|
|
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):
|
|
if args.use_ema:
|
|
# Load EMA models (only for trained components)
|
|
if ema_transformer is not None and "transformer" in components_to_train and os.path.exists(os.path.join(input_dir, "ema_transformer")):
|
|
ema_transformer.from_pretrained(os.path.join(input_dir, "ema_transformer")).to(accelerator.device)
|
|
if ema_transformer_2 is not None and "transformer_2" in components_to_train and os.path.exists(os.path.join(input_dir, "ema_transformer_2")):
|
|
ema_transformer_2.from_pretrained(os.path.join(input_dir, "ema_transformer_2")).to(accelerator.device)
|
|
if "transformer_audio" in components_to_train and os.path.exists(os.path.join(input_dir, "ema_transformer_audio")):
|
|
ema_transformer_audio.from_pretrained(os.path.join(input_dir, "ema_transformer_audio")).to(accelerator.device)
|
|
if "dual_tower_bridge" in components_to_train and os.path.exists(os.path.join(input_dir, "ema_dual_tower_bridge")):
|
|
ema_dual_tower_bridge.from_pretrained(os.path.join(input_dir, "ema_dual_tower_bridge")).to(accelerator.device)
|
|
|
|
for i in range(len(models)):
|
|
models.pop()
|
|
|
|
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:
|
|
mova_model.enable_gradient_checkpointing()
|
|
|
|
if args.low_vram:
|
|
mova_model.enable_model_offload()
|
|
|
|
# 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 Exception:
|
|
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
|
|
|
|
# Collect trainable parameters from all components
|
|
trainable_params = []
|
|
if transformer is not None:
|
|
trainable_params.extend(filter(lambda p: p.requires_grad, transformer.parameters()))
|
|
if transformer_2 is not None:
|
|
trainable_params.extend(filter(lambda p: p.requires_grad, transformer_2.parameters()))
|
|
trainable_params.extend(filter(lambda p: p.requires_grad, transformer_audio.parameters()))
|
|
trainable_params.extend(filter(lambda p: p.requires_grad, dual_tower_bridge.parameters()))
|
|
|
|
trainable_params_optim = [
|
|
{'params': [], 'lr': args.learning_rate},
|
|
{'params': [], 'lr': args.learning_rate / 2},
|
|
]
|
|
in_already = []
|
|
|
|
# Collect from all components
|
|
components_to_optimize = []
|
|
if transformer is not None:
|
|
components_to_optimize.append(("transformer", transformer))
|
|
if transformer_2 is not None:
|
|
components_to_optimize.append(("transformer_2", transformer_2))
|
|
components_to_optimize.append(("transformer_audio", transformer_audio))
|
|
components_to_optimize.append(("dual_tower_bridge", dual_tower_bridge))
|
|
|
|
for component_name, component in components_to_optimize:
|
|
for name, param in component.named_parameters():
|
|
high_lr_flag = False
|
|
full_name = f"{component_name}.{name}"
|
|
if full_name in in_already:
|
|
continue
|
|
for trainable_module_name in args.trainable_modules:
|
|
if trainable_module_name in name:
|
|
in_already.append(full_name)
|
|
high_lr_flag = True
|
|
trainable_params_optim[0]['params'].append(param)
|
|
if accelerator.is_main_process:
|
|
print(f"Set {full_name} to lr : {args.learning_rate}")
|
|
break
|
|
if high_lr_flag:
|
|
continue
|
|
for trainable_module_name in args.trainable_modules_low_learning_rate:
|
|
if trainable_module_name in name:
|
|
in_already.append(full_name)
|
|
trainable_params_optim[1]['params'].append(param)
|
|
if accelerator.is_main_process:
|
|
print(f"Set {full_name} to lr : {args.learning_rate / 2}")
|
|
break
|
|
|
|
if args.use_came:
|
|
optimizer = optimizer_cls(
|
|
trainable_params_optim,
|
|
lr=args.learning_rate,
|
|
# weight_decay=args.adam_weight_decay,
|
|
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.config.temporal_compression_ratio
|
|
|
|
if args.fix_sample_size is not None and args.enable_bucket:
|
|
args.video_sample_size = max(max(args.fix_sample_size), args.video_sample_size)
|
|
args.image_sample_size = max(max(args.fix_sample_size), args.image_sample_size)
|
|
args.training_with_video_token_length = False
|
|
args.random_hw_adapt = False
|
|
|
|
# Get the dataset
|
|
train_dataset = VideoSpeechDataset(
|
|
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,
|
|
enable_bucket=args.enable_bucket, enable_inpaint=True, audio_sr=getattr(audio_vae.config, 'sample_rate', 24000),
|
|
)
|
|
|
|
def worker_init_fn(_seed):
|
|
_seed = _seed * 256
|
|
def _worker_init_fn(worker_id):
|
|
print(f"worker_init_fn with {_seed + worker_id}")
|
|
np.random.seed(_seed + worker_id)
|
|
random.seed(_seed + worker_id)
|
|
return _worker_init_fn
|
|
|
|
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,
|
|
)
|
|
|
|
def collate_fn(examples):
|
|
def get_length_to_frame_num(token_length):
|
|
if args.video_sample_size > 256:
|
|
sample_sizes = list(range(256, args.video_sample_size + 1, 128))
|
|
|
|
if sample_sizes[-1] != args.video_sample_size:
|
|
sample_sizes.append(args.video_sample_size)
|
|
else:
|
|
sample_sizes = [args.video_sample_size]
|
|
|
|
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
|
|
}
|
|
|
|
return length_to_frame_num
|
|
|
|
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 = 0.90
|
|
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)
|
|
|
|
# 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"] = []
|
|
new_examples["audio"] = []
|
|
new_examples["fps"] = []
|
|
|
|
# Used in Inpaint mode
|
|
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"]
|
|
f, h, w, c = np.shape(pixel_value)
|
|
|
|
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())
|
|
local_video_sample_size = np.random.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)
|
|
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()}
|
|
|
|
if args.fix_sample_size is not None:
|
|
fix_sample_size = [int(x / 16) * 16 for x in args.fix_sample_size]
|
|
elif 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 / 64) * 64 for x in random_sample_size]
|
|
else:
|
|
closest_size, closest_ratio = get_closest_ratio(h, w, ratios=aspect_ratio_sample_size)
|
|
closest_size = [int(x / 64) * 64 for x in closest_size]
|
|
|
|
min_example_length = min(
|
|
[example["pixel_values"].shape[0] for example in examples]
|
|
)
|
|
batch_video_length = int(min(batch_video_length, min_example_length))
|
|
|
|
# Magvae needs the number of frames to be 4n + 1.
|
|
batch_video_length = (batch_video_length - 1) // sample_n_frames_bucket_interval * sample_n_frames_bucket_interval + 1
|
|
|
|
if batch_video_length <= 0:
|
|
batch_video_length = 1
|
|
|
|
for example in examples:
|
|
if args.fix_sample_size is not None:
|
|
# 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
|
|
fix_sample_size = list(map(lambda x: int(x), fix_sample_size))
|
|
transform = transforms.Compose([
|
|
transforms.Resize(fix_sample_size, interpolation=transforms.InterpolationMode.BILINEAR), # Image.BICUBIC
|
|
transforms.CenterCrop(fix_sample_size),
|
|
transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5], inplace=True),
|
|
])
|
|
elif 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)[:batch_video_length])
|
|
new_examples["text"].append(example["text"])
|
|
|
|
audio_length = np.shape(example["audio"])[0]
|
|
batch_audio_length = int(audio_length / pixel_values.size()[0] * batch_video_length)
|
|
new_examples["audio"].append(example["audio"][:batch_audio_length])
|
|
new_examples["fps"].append(example.get("fps", 24))
|
|
|
|
mask = get_random_mask(new_examples["pixel_values"][-1].size(), image_start_only=True)
|
|
mask_pixel_values = new_examples["pixel_values"][-1] * (1 - mask)
|
|
# Wan 2.1 use 0 for masked pixels
|
|
# + torch.ones_like(new_examples["pixel_values"][-1]) * -1 * mask
|
|
new_examples["mask_pixel_values"].append(mask_pixel_values)
|
|
new_examples["mask"].append(mask)
|
|
|
|
clip_pixel_values = new_examples["pixel_values"][-1][0].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 for example in new_examples["pixel_values"]])
|
|
new_examples["mask_pixel_values"] = torch.stack([example for example in new_examples["mask_pixel_values"]])
|
|
new_examples["mask"] = torch.stack([example for example in new_examples["mask"]])
|
|
new_examples["clip_pixel_values"] = torch.stack([example for example in new_examples["clip_pixel_values"]])
|
|
|
|
# Pad audio to same length and stack
|
|
new_examples["audio"] = torch.stack([example for example in new_examples["audio"]])
|
|
new_examples["fps"] = new_examples["fps"]
|
|
|
|
# Encode prompts when enable_text_encoder_in_dataloader=True
|
|
if args.enable_text_encoder_in_dataloader:
|
|
# Gemma expects left padding for chat-style prompts
|
|
tokenizer.padding_side = "left"
|
|
if tokenizer.pad_token is None:
|
|
tokenizer.pad_token = tokenizer.eos_token
|
|
|
|
prompt_ids = tokenizer(
|
|
new_examples['text'],
|
|
max_length=args.tokenizer_max_length,
|
|
padding="max_length",
|
|
add_special_tokens=True,
|
|
truncation=True,
|
|
return_tensors="pt"
|
|
)
|
|
text_encoder_outputs = text_encoder(
|
|
input_ids=prompt_ids.input_ids,
|
|
attention_mask=prompt_ids.attention_mask,
|
|
output_hidden_states=True
|
|
)
|
|
text_encoder_hidden_states = text_encoder_outputs.hidden_states
|
|
text_encoder_hidden_states = torch.stack(text_encoder_hidden_states, dim=-1)
|
|
|
|
# Pack text embeddings (normalized and flattened)
|
|
sequence_lengths = prompt_ids.attention_mask.sum(dim=-1)
|
|
prompt_embeds = _pack_text_embeds(
|
|
text_encoder_hidden_states,
|
|
sequence_lengths,
|
|
device=text_encoder_hidden_states.device,
|
|
padding_side=tokenizer.padding_side,
|
|
scale_factor=8,
|
|
)
|
|
new_examples['encoder_attention_mask'] = prompt_ids.attention_mask
|
|
new_examples['encoder_hidden_states'] = prompt_embeds
|
|
|
|
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,
|
|
worker_init_fn=worker_init_fn(args.seed + accelerator.process_index)
|
|
)
|
|
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,
|
|
worker_init_fn=worker_init_fn(args.seed + accelerator.process_index)
|
|
)
|
|
|
|
# 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`.
|
|
mova_model, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
|
|
mova_model, optimizer, train_dataloader, lr_scheduler
|
|
)
|
|
|
|
if fsdp_stage != 0 or zero_stage != 0:
|
|
from functools import partial
|
|
|
|
from videox_fun.dist import set_multi_gpus_devices, shard_model
|
|
shard_fn = partial(shard_model, device_id=accelerator.device, param_dtype=weight_dtype, module_to_wrapper=text_encoder.encoder.block)
|
|
text_encoder = shard_fn(text_encoder)
|
|
|
|
if args.use_ema:
|
|
ema_mova_model.to(accelerator.device)
|
|
|
|
# Move text_encode and vae to gpu and cast to weight_dtype
|
|
vae.to(accelerator.device if not args.low_vram else "cpu", dtype=weight_dtype)
|
|
audio_vae.to(accelerator.device if not args.low_vram else "cpu")
|
|
if not args.enable_text_encoder_in_dataloader:
|
|
text_encoder.to(accelerator.device if not args.low_vram else "cpu", dtype=weight_dtype)
|
|
|
|
# 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))
|
|
keys_to_pop = [k for k, v in tracker_config.items() if isinstance(v, list)]
|
|
for k in keys_to_pop:
|
|
tracker_config.pop(k)
|
|
print(f"Removed tracker_config['{k}']")
|
|
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}.")
|
|
|
|
accelerator.print(f"Resuming from checkpoint {path}")
|
|
accelerator.load_state(os.path.join(args.output_dir, path))
|
|
else:
|
|
initial_global_step = 0
|
|
|
|
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:
|
|
# create extra cuda streams to speedup vae computation
|
|
vae_stream_1 = torch.cuda.Stream()
|
|
else:
|
|
vae_stream_1 = None
|
|
|
|
# Calculate sampling range based on boundary_type (similar to wan2.2)
|
|
boundary = args.boundary_ratio
|
|
split_timesteps = args.train_sampling_steps * boundary
|
|
differences = torch.abs(noise_scheduler.timesteps - split_timesteps)
|
|
closest_index = torch.argmin(differences).item()
|
|
|
|
if args.boundary_type == "high":
|
|
# High noise model: sample from [0, boundary]
|
|
start_num_idx = 0
|
|
train_sampling_steps = closest_index
|
|
print(f"Training HIGH-NOISE only (transformer_2): boundary={boundary}, sampling timesteps [0, {closest_index}]")
|
|
elif args.boundary_type == "low":
|
|
# Low noise model: sample from [boundary, max]
|
|
start_num_idx = closest_index
|
|
train_sampling_steps = args.train_sampling_steps - closest_index
|
|
print(f"Training LOW-NOISE only (transformer): boundary={boundary}, sampling timesteps [{closest_index}, {args.train_sampling_steps}]")
|
|
else:
|
|
# Full: sample from all timesteps
|
|
start_num_idx = 0
|
|
train_sampling_steps = args.train_sampling_steps
|
|
print(f"Training FULL: sampling all timesteps [0, {args.train_sampling_steps}]")
|
|
|
|
idx_sampling = DiscreteSampling(train_sampling_steps, start_num_idx=start_num_idx, 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")
|
|
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]}.mp4", rescale=True)
|
|
|
|
with accelerator.accumulate(mova_model):
|
|
# Convert images to latent space
|
|
pixel_values = batch["pixel_values"].to(weight_dtype)
|
|
audio = batch["audio"]
|
|
fps = batch["fps"][0] if batch["fps"] else 24 # Use fps from dataset
|
|
|
|
# Increase the batch size when the length of the latent sequence of the current sample is small
|
|
if args.auto_tile_batch_size and args.training_with_video_token_length and zero_stage != 3:
|
|
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['encoder_hidden_states'] = torch.tile(batch['encoder_hidden_states'], (4, 1, 1))
|
|
batch['encoder_attention_mask'] = torch.tile(batch['encoder_attention_mask'], (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['encoder_hidden_states'] = torch.tile(batch['encoder_hidden_states'], (2, 1, 1))
|
|
batch['encoder_attention_mask'] = torch.tile(batch['encoder_attention_mask'], (2, 1))
|
|
else:
|
|
batch['text'] = batch['text'] * 2
|
|
|
|
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
|
|
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))]
|
|
select_frames_prob = np.array(_create_special_list(len(select_frames)))
|
|
|
|
if len(select_frames) != 0:
|
|
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)
|
|
else:
|
|
temp_n_frames = 1
|
|
|
|
# Magvae needs the number of frames to be 4n + 1.
|
|
temp_n_frames = (temp_n_frames - 1) // sample_n_frames_bucket_interval + 1
|
|
|
|
pixel_values = pixel_values[:, :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)
|
|
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
|
|
actual_video_length = int(max(actual_video_length, 1))
|
|
|
|
# Magvae needs the number of frames to be 4n + 1.
|
|
actual_video_length = (actual_video_length - 1) // sample_n_frames_bucket_interval + 1
|
|
|
|
pixel_values = pixel_values[:, :actual_video_length, :, :]
|
|
|
|
if args.low_vram:
|
|
torch.cuda.empty_cache()
|
|
vae.to(accelerator.device)
|
|
audio_vae.to(accelerator.device)
|
|
if not args.enable_text_encoder_in_dataloader:
|
|
text_encoder.to("cpu")
|
|
|
|
# 50% probability: first frame only or first+last frame conditioning
|
|
use_last_frame = random.random() < 0.5
|
|
|
|
with torch.no_grad():
|
|
# This way is quicker when batch grows up
|
|
def _batch_encode_vae(pixel_values):
|
|
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)
|
|
return torch.cat(new_pixel_values, dim = 0)
|
|
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)
|
|
|
|
# Encode condition video to latent space using same batch encoding as latents
|
|
if use_last_frame:
|
|
# First + last frame conditioning
|
|
video_condition_pixel = torch.cat([
|
|
pixel_values[:, 0:1, :, :, :], # First frame
|
|
torch.zeros_like(pixel_values[:, 1:-1, :, :, :]), # Zeros for middle frames
|
|
pixel_values[:, -1:, :, :, :] # Last frame
|
|
], dim=1)
|
|
else:
|
|
# First frame only conditioning (last_image = None case)
|
|
video_condition_pixel = torch.cat([
|
|
pixel_values[:, 0:1, :, :, :], # First frame
|
|
torch.zeros_like(pixel_values[:, 1:, :, :, :]) # Zeros for remaining frames
|
|
], dim=1)
|
|
|
|
latent_condition = _batch_encode_vae(video_condition_pixel)
|
|
|
|
# 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)
|
|
|
|
# Get latent dimensions from VAE output for later use
|
|
bsz, _, latent_num_frames, latent_height, latent_width = latents.size()
|
|
|
|
# Encode audio to latents
|
|
with torch.no_grad():
|
|
audio_batch = audio.to(device=accelerator.device, dtype=torch.float32)
|
|
# audio_batch shape: [batch, channels, samples] or [batch, samples]
|
|
if audio_batch.ndim == 2:
|
|
audio_batch = audio_batch.unsqueeze(1) # [batch, 1, samples]
|
|
|
|
# Preprocess audio (padding to match hop_length) following official MOVA
|
|
audio_batch = audio_vae.preprocess(audio_batch, sample_rate=getattr(audio_vae.config, 'sample_rate', 24000))
|
|
|
|
# Encode audio using audio_vae
|
|
# audio_latents_raw shape: [batch, latent_channels, latent_time]
|
|
audio_latents_raw = audio_vae.encode(audio_batch)[0].mode()
|
|
|
|
if args.low_vram:
|
|
vae.to('cpu')
|
|
audio_vae.to('cpu')
|
|
torch.cuda.empty_cache()
|
|
if not args.enable_text_encoder_in_dataloader:
|
|
text_encoder.to(accelerator.device)
|
|
|
|
if args.enable_text_encoder_in_dataloader:
|
|
prompt_embeds = batch['encoder_hidden_states'].to(device=latents.device, dtype=weight_dtype)
|
|
prompt_attention_mask = batch['encoder_attention_mask'].to(device=latents.device)
|
|
else:
|
|
with torch.no_grad():
|
|
# UMT5 tokenizer (T5-style)
|
|
tokenizer.padding_side = "right"
|
|
if tokenizer.pad_token is None:
|
|
tokenizer.pad_token = tokenizer.eos_token
|
|
|
|
# Clean prompts following pipeline convention
|
|
cleaned_texts = [whitespace_clean(basic_clean(text)) for text in batch['text']]
|
|
|
|
prompt_ids = tokenizer(
|
|
cleaned_texts,
|
|
padding="max_length",
|
|
max_length=args.tokenizer_max_length,
|
|
truncation=True,
|
|
add_special_tokens=True,
|
|
return_tensors="pt"
|
|
)
|
|
text_input_ids = prompt_ids.input_ids.to(latents.device)
|
|
prompt_attention_mask = prompt_ids.attention_mask.to(latents.device)
|
|
|
|
# Get sequence lengths for truncation
|
|
seq_lens = prompt_attention_mask.gt(0).sum(dim=1).long()
|
|
|
|
# Get text encoder hidden states (UMT5 returns last_hidden_state directly)
|
|
prompt_embeds = text_encoder(
|
|
input_ids=text_input_ids,
|
|
attention_mask=prompt_attention_mask
|
|
).last_hidden_state
|
|
prompt_embeds = prompt_embeds.to(dtype=weight_dtype)
|
|
|
|
# Truncate to actual sequence length and re-pad (following pipeline convention)
|
|
# This removes padding token embeddings which could affect training
|
|
prompt_embeds_list = [embed[:seq_len] for embed, seq_len in zip(prompt_embeds, seq_lens)]
|
|
prompt_embeds = torch.stack(
|
|
[torch.cat([embed, embed.new_zeros(args.tokenizer_max_length - embed.size(0), embed.size(1))])
|
|
for embed in prompt_embeds_list], dim=0
|
|
)
|
|
|
|
if args.low_vram and not args.enable_text_encoder_in_dataloader:
|
|
text_encoder.to('cpu')
|
|
torch.cuda.empty_cache()
|
|
|
|
noise = torch.randn(latents.size(), device=latents.device, generator=torch_rng, dtype=weight_dtype)
|
|
audio_noise = torch.randn(audio_latents_raw.size(), device=latents.device, generator=torch_rng, dtype=weight_dtype)
|
|
|
|
if not args.uniform_sampling:
|
|
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()
|
|
else:
|
|
# Sample a random timestep for each image
|
|
indices = idx_sampling(bsz, generator=torch_rng, device=latents.device)
|
|
indices = indices.long().cpu()
|
|
timesteps = noise_scheduler.timesteps[indices].to(device=latents.device)
|
|
|
|
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
|
|
|
|
# ------------------ I2V Conditioning ------------------
|
|
# Create mask following pipeline logic (prepare_latents method)
|
|
# Supports both first-frame and last-frame conditioning
|
|
temporal_compression_ratio = vae.config.temporal_compression_ratio
|
|
num_pixel_frames = pixel_values.shape[1]
|
|
|
|
# Build mask: frame dim is pixel, spatial dims are latent (same as pipeline)
|
|
mask_pixel = torch.ones(
|
|
bsz, 1, num_pixel_frames, latent_height, latent_width,
|
|
device=latents.device, dtype=latents.dtype
|
|
)
|
|
|
|
if use_last_frame:
|
|
# First and last frame are condition (middle frames are 0)
|
|
mask_pixel[:, :, 1:-1, :, :] = 0
|
|
else:
|
|
# Only first frame is condition
|
|
mask_pixel[:, :, 1:, :, :] = 0
|
|
|
|
# Extract first frame mask and repeat (same as pipeline line 281-282)
|
|
first_frame_mask = mask_pixel[:, :, 0:1, :, :] # [B, 1, 1, H, W]
|
|
first_frame_mask = torch.repeat_interleave(
|
|
first_frame_mask, dim=2, repeats=temporal_compression_ratio
|
|
) # [B, 1, 4, H, W]
|
|
|
|
# Concatenate and reshape to latent temporal dim (same as pipeline line 283-285)
|
|
mask_pixel = torch.cat([first_frame_mask, mask_pixel[:, :, 1:, :]], dim=2)
|
|
# View: [B, 1, F_pixel+3, H, W] -> [B, num_latent_frames, 4, H_latent, W_latent]
|
|
mask_lat_size = mask_pixel.view(
|
|
bsz, -1, temporal_compression_ratio, latent_height, latent_width
|
|
)
|
|
mask_lat_size = mask_lat_size.transpose(1, 2) # [B, 4, F_latent, H, W]
|
|
mask_lat_size = mask_lat_size.to(latent_condition.device)
|
|
|
|
# Concatenate mask and condition latents
|
|
# Result: [B, temporal_compression_ratio + C, F, H, W]
|
|
conditioning_latents = torch.concat([mask_lat_size, latent_condition], dim=1)
|
|
|
|
# ------------------ Video Latents ------------------
|
|
# Add noise according to flow matching
|
|
# zt = (1 - sigma) * x + sigma * noise
|
|
sigmas = get_sigmas(timesteps, n_dim=latents.ndim, dtype=latents.dtype)
|
|
noisy_latents = (1.0 - sigmas) * latents + sigmas * noise
|
|
target = noise - latents
|
|
|
|
# Pack: [noisy_latents + condition]
|
|
# This matches inference: latents (noisy) + condition [mask, first_frame_latent]
|
|
latent_model_input = torch.cat([noisy_latents, conditioning_latents], dim=1)
|
|
|
|
# ------------------ Audio Latents ------------------
|
|
# Add noise to audio latents for training (flow matching)
|
|
audio_sigmas = get_sigmas(timesteps, n_dim=audio_latents_raw.ndim, dtype=audio_latents_raw.dtype)
|
|
noisy_audio_latents = (1.0 - audio_sigmas) * audio_latents_raw + audio_sigmas * audio_noise
|
|
audio_target = audio_noise - audio_latents_raw
|
|
|
|
# -------- Forward --------
|
|
# Predict the noise residual using MOVA model
|
|
# Wan2.2 convention: transformer_2 = high-noise (large t), transformer = low-noise (small t)
|
|
# For boundary_type == "low" or "high", we always use the loaded model
|
|
# For boundary_type == "full", we switch based on timestep
|
|
if args.boundary_type == "high":
|
|
use_low_noise_dit = False # Always use transformer_2 (high noise)
|
|
elif args.boundary_type == "low":
|
|
use_low_noise_dit = True # Always use transformer (low noise)
|
|
else:
|
|
# Full mode: switch based on timestep
|
|
boundary_timestep = args.boundary_ratio * noise_scheduler.config.num_train_timesteps
|
|
use_low_noise_dit = timesteps[0].item() < boundary_timestep # small t = low noise = transformer
|
|
|
|
with torch.cuda.amp.autocast(dtype=weight_dtype), torch.cuda.device(device=accelerator.device):
|
|
noise_pred_video, noise_pred_audio = mova_model(
|
|
visual_latents=latent_model_input,
|
|
audio_latents=noisy_audio_latents,
|
|
context=prompt_embeds,
|
|
timestep=timesteps,
|
|
audio_timestep=timesteps,
|
|
frame_rate=fps,
|
|
use_low_noise_dit=use_low_noise_dit,
|
|
)
|
|
|
|
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
|
|
|
|
weighting = compute_loss_weighting_for_sd3(weighting_scheme=args.weighting_scheme, sigmas=sigmas)
|
|
|
|
# Video loss
|
|
video_loss = custom_mse_loss(noise_pred_video.float(), target.float(), weighting.float())
|
|
|
|
if args.motion_sub_loss and noise_pred_video.size()[2] > 2:
|
|
gt_sub_noise = noise_pred_video[:, :, 1:].float() - noise_pred_video[:, :, :-1].float()
|
|
pre_sub_noise = target[:, :, 1:].float() - target[:, :, :-1].float()
|
|
sub_loss = F.mse_loss(gt_sub_noise, pre_sub_noise, reduction="mean")
|
|
video_loss = video_loss * (1 - args.motion_sub_loss_ratio) + sub_loss * args.motion_sub_loss_ratio
|
|
|
|
# Audio loss
|
|
# Use same custom_mse_loss with threshold for consistency with video loss
|
|
audio_weighting = compute_loss_weighting_for_sd3(weighting_scheme=args.weighting_scheme, sigmas=audio_sigmas)
|
|
audio_loss = custom_mse_loss(noise_pred_audio.float(), audio_target.float(), audio_weighting.float() if audio_weighting is not None else None)
|
|
|
|
# Combined loss (equal weighting for video and audio)
|
|
loss = video_loss + 0.1 * audio_loss
|
|
|
|
# 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
|
|
# Ensure mova_model components are on the correct device before backward (for gradient checkpointing + offload)
|
|
if args.low_vram:
|
|
if transformer is not None:
|
|
transformer.to(accelerator.device)
|
|
if transformer_2 is not None:
|
|
transformer_2.to(accelerator.device)
|
|
transformer_audio.to(accelerator.device)
|
|
dual_tower_bridge.to(accelerator.device)
|
|
accelerator.backward(loss)
|
|
if accelerator.sync_gradients:
|
|
if not args.use_deepspeed and not args.use_fsdp:
|
|
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 not args.use_fsdp 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 mova_model.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 not args.use_fsdp 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:
|
|
|
|
if args.use_ema:
|
|
ema_mova_model.step(mova_model.parameters())
|
|
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 args.use_fsdp 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)
|
|
|
|
gc.collect()
|
|
torch.cuda.empty_cache()
|
|
torch.cuda.ipc_collect()
|
|
save_path = os.path.join(args.output_dir, f"checkpoint-{global_step}")
|
|
accelerator.save_state(save_path)
|
|
logger.info(f"Saved state to {save_path}")
|
|
|
|
if args.validation_prompts is not None and global_step % args.validation_steps == 0:
|
|
if args.use_ema:
|
|
# Collect all trainable parameters for EMA
|
|
all_params = []
|
|
if transformer is not None:
|
|
all_params.extend(transformer.parameters())
|
|
if transformer_2 is not None:
|
|
all_params.extend(transformer_2.parameters())
|
|
all_params.extend(transformer_audio.parameters())
|
|
all_params.extend(dual_tower_bridge.parameters())
|
|
# Store the parameters temporarily and load the EMA parameters to perform inference.
|
|
ema_mova_model.store(all_params)
|
|
ema_mova_model.copy_to(all_params)
|
|
log_validation(
|
|
vae,
|
|
audio_vae,
|
|
text_encoder,
|
|
tokenizer,
|
|
transformer,
|
|
transformer_2,
|
|
transformer_audio,
|
|
dual_tower_bridge,
|
|
mova_model,
|
|
args,
|
|
accelerator,
|
|
weight_dtype,
|
|
global_step,
|
|
)
|
|
if args.use_ema:
|
|
# Switch back to the original parameters.
|
|
ema_mova_model.restore(all_params)
|
|
|
|
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 args.validation_prompts is not None and epoch % args.validation_epochs == 0:
|
|
if args.use_ema:
|
|
# Collect all trainable parameters for EMA
|
|
all_params = []
|
|
if transformer is not None:
|
|
all_params.extend(transformer.parameters())
|
|
if transformer_2 is not None:
|
|
all_params.extend(transformer_2.parameters())
|
|
all_params.extend(transformer_audio.parameters())
|
|
all_params.extend(dual_tower_bridge.parameters())
|
|
# Store the parameters temporarily and load the EMA parameters to perform inference.
|
|
ema_mova_model.store(all_params)
|
|
ema_mova_model.copy_to(all_params)
|
|
log_validation(
|
|
vae,
|
|
audio_vae,
|
|
text_encoder,
|
|
tokenizer,
|
|
transformer,
|
|
transformer_2,
|
|
transformer_audio,
|
|
dual_tower_bridge,
|
|
mova_model,
|
|
args,
|
|
accelerator,
|
|
weight_dtype,
|
|
global_step,
|
|
)
|
|
if args.use_ema:
|
|
# Switch back to the original parameters.
|
|
ema_mova_model.restore(all_params)
|
|
|
|
# Create the pipeline using the trained modules and save it.
|
|
accelerator.wait_for_everyone()
|
|
if accelerator.is_main_process:
|
|
mova_model = unwrap_model(mova_model)
|
|
if args.use_ema:
|
|
ema_mova_model.copy_to(mova_model.parameters())
|
|
|
|
if args.use_deepspeed or args.use_fsdp or accelerator.is_main_process:
|
|
gc.collect()
|
|
torch.cuda.empty_cache()
|
|
torch.cuda.ipc_collect()
|
|
save_path = os.path.join(args.output_dir, f"checkpoint-{global_step}")
|
|
accelerator.save_state(save_path)
|
|
logger.info(f"Saved state to {save_path}")
|
|
|
|
accelerator.end_training()
|
|
|
|
|
|
if __name__ == "__main__":
|
|
main()
|