三、训练扩散模型
训练的基础
不能直接训练pipeline,得单独训练组件。
通常,最好的结果是通过在特定数据集上微调预训练模型来获得的。可以在Hub上找到许多这样的检查点. 本教程将教如何在Smithsonian Butterflies数据集的子集上从头开始训练UNet2DModel,以生成自己的 🦋 蝴蝶 🦋。
训练前设置
在开始之前,请确保安装数据集来加载和预处理图像数据集,并安装了加速,以简化在任意数量的 GPU 上的训练。以下命令还将安装**TensorBoard以可视化训练指标(您还可以使用权重和偏差**来跟踪您的训练)。
pip install differs[training]
上传模型:从笔记本登录并在出现提示时输入您的令牌。确保您的令牌具有写入角色。
from huggingface_hub import notebook_login
notebook_login()
# 或者从终端登录:
huggingface-cli login
由于模型检查点非常大,因此安装**Git-LFS**来对这些大文件进行版本控制:
!sudo apt -qq install git-lfs
!git config --global credential.helper store
训练配置
为了方便起见,创建一个TrainingConfig
包含训练超参数的类(随意调整它们):
from dataclasses import dataclass
@dataclass
class TrainingConfig:
image_size = 128 # the generated image resolution
train_batch_size = 16
eval_batch_size = 16 # how many images to sample during evaluation
num_epochs = 50
gradient_accumulation_steps = 1
learning_rate = 1e-4
lr_warmup_steps = 500
save_image_epochs = 10
save_model_epochs = 30
mixed_precision = "fp16" # `no` for float32, `fp16` for automatic mixed precision
output_dir = "ddpm-butterflies-128" # the model name locally and on the HF Hub
push_to_hub = True # whether to upload the saved model to the HF Hub
hub_model_id = "<your-username>/<my-awesome-model>" # the name of the repository to create on the HF Hub
hub_private_repo = False
overwrite_output_dir = True # overwrite the old model when re-running the notebook
seed = 0
config = TrainingConfig()
加载数据集
使用 🤗 数据集库轻松加载**史密森尼蝴蝶数据集**
from datasets import load_dataset
config.dataset_name = "huggan/smithsonian_butterflies_subset"
dataset = load_dataset(config.dataset_name, split="train")
**可以从HugGan 社区活动中找到其他数据集,也可以通过[ImageFolder](https://huggingface.co/docs/datasets/image_dataset#imagefolder)
**创建本地config.dataset_name
如果数据集来自 HugGan 社区活动,或者imagefolder
您使用自己的图像,则设置为数据集的存储库 ID。
快速可视化
Datasets 使用**Image功能自动解码图像数据并将其加载为[PIL.Image](https://pillow.readthedocs.io/en/stable/reference/Image.html)
**我们可以可视化的图像:
import matplotlib.pyplot as plt
fig, axs = plt.subplots(1, 4, figsize=(16, 4))
for i, image in enumerate(dataset[:4]["image"]):
axs[i].imshow(image)
axs[i].set_axis_off()
fig.show()
预处理
不过,这些图像的大小不同,因此您需要先对它们进行预处理:
Resizeconfig.image_size
将图像大小更改为 中定义的大小。RandomHorizontalFlip
通过随机镜像图像来增强数据集。Normalize
将像素值重新缩放到 [-1, 1] 范围内非常重要,这正是模型所期望的。
from torchvision import transforms
preprocess = transforms.Compose(
[ transforms.Resize((config.image_size, config.image_size)),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize([0.5], [0.5]),
])
数据集的**set_transform**preprocess
方法在训练期间动态应用该函数:
def transform(examples):
images = [preprocess(image.convert("RGB")) for image in examples["image"]]
return {"images": images}
dataset.set_transform(transform)
训练模型
构造数据加载器
import torch
train_dataloader = torch.utils.data.DataLoader(dataset, batch_size=config.train_batch_size, shuffle=True)
Diffusers 中的预训练模型可以使用您想要的参数轻松地从其模型类创建。例如,创建**UNet2DModel**:
from diffusers import UNet2DModel
model = UNet2DModel(
sample_size=config.image_size, # the target image resolution
in_channels=3, # the number of input channels, 3 for RGB images
out_channels=3, # the number of output channels
layers_per_block=2, # how many ResNet layers to use per UNet block
# 每个 UNet 块的输出通道数
block_out_channels=(128, 128, 256, 256, 512, 512), # the number of output channels for each UNet block
down_block_types=(
"DownBlock2D", # a regular ResNet downsampling block
"DownBlock2D",
"DownBlock2D",
"DownBlock2D",
"AttnDownBlock2D", # a ResNet downsampling block with spatial self-attention
"DownBlock2D",
),
up_block_types=(
"UpBlock2D", # a regular ResNet upsampling block
"AttnUpBlock2D", # a ResNet upsampling block with spatial self-attention
"UpBlock2D",
"UpBlock2D",
"UpBlock2D",
"UpBlock2D",
),
)
创建调度程序
根据使用模型进行训练还是推理,调度程序的行为会有所不同。在推理过程中,调度器根据噪声生成图像。在训练期间,调度程序从扩散过程中的特定点获取模型输出(或样本),并根据噪声调度和更新规则将噪声应用于图像。
让我们看一下**DDPMScheduler**并使用该add_noise
方法向之前添加一些随机噪声sample_image
:
import torch
from PIL import Image
from diffusers import DDPMScheduler
noise_scheduler = DDPMScheduler(num_train_timesteps=1000)
noise = torch.randn(sample_image.shape)
timesteps = torch.LongTensor([50])
noisy_image = noise_scheduler.add_noise(sample_image, noise, timesteps)
Image.fromarray(((noisy_image.permute(0, 2, 3, 1) + 1.0) * 127.5).type(torch.uint8).numpy()[0])
模型的训练目标是预测添加到图像中的噪声。这一步的损失可以通过下式计算:
import torch.nn.functional as F
noise_pred = model(noisy_image, timesteps).sample
loss = F.mse_loss(noise_pred, noise)
训练模型
到目前为止,您已经掌握了开始训练模型的大部分内容,剩下的就是将所有内容组合在一起。
首先,需要一个优化器和一个学习率调度器:
from diffusers.optimization import get_cosine_schedule_with_warmup
optimizer = torch.optim.AdamW(model.parameters(), lr=config.learning_rate)
lr_scheduler = get_cosine_schedule_with_warmup(
optimizer=optimizer,
num_warmup_steps=config.lr_warmup_steps,
num_training_steps=(len(train_dataloader) * config.num_epochs),
)
然后,您需要一种评估模型的方法。为了进行评估,您可以使用**DDPMPipeline**生成一批样本图像并将其保存为网格:
from diffusers import DDPMPipeline
from diffusers.utils import make_image_grid
import os
def evaluate(config, epoch, pipeline):
# Sample some images from random noise (this is the backward diffusion process).
# The default pipeline output type is `List[PIL.Image]`
images = pipeline(
batch_size=config.eval_batch_size,
generator=torch.manual_seed(config.seed),
).images
# Make a grid out of the images
image_grid = make_image_grid(images, rows=4, cols=4)
# Save the images
test_dir = os.path.join(config.output_dir, "samples")
os.makedirs(test_dir, exist_ok=True)
image_grid.save(f"{test_dir}/{epoch:04d}.png")
构建训练循环
现在,您可以使用 Accelerate 将所有这些组件包装在一个训练循环中,以轻松进行 TensorBoard 日志记录、梯度累积和混合精度训练。要将模型上传到 Hub,请编写一个函数来获取存储库名称和信息,然后将其推送到 Hub。
from accelerate import Accelerator
from huggingface_hub import create_repo, upload_folder
from tqdm.auto import tqdm
from pathlib import Path
import os
def train_loop(config, model, noise_scheduler, optimizer, train_dataloader, lr_scheduler):
# Initialize accelerator and tensorboard logging
accelerator = Accelerator(
mixed_precision=config.mixed_precision,
gradient_accumulation_steps=config.gradient_accumulation_steps,
log_with="tensorboard",
project_dir=os.path.join(config.output_dir, "logs"),
)
if accelerator.is_main_process:
if config.output_dir is not None:
os.makedirs(config.output_dir, exist_ok=True)
if config.push_to_hub:
repo_id = create_repo(
repo_id=config.hub_model_id or Path(config.output_dir).name, exist_ok=True
).repo_id
accelerator.init_trackers("train_example")
# Prepare everything
# There is no specific order to remember, you just need to unpack the
# objects in the same order you gave them to the prepare method.
model, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
model, optimizer, train_dataloader, lr_scheduler
)
global_step = 0
# Now you train the model
for epoch in range(config.num_epochs):
progress_bar = tqdm(total=len(train_dataloader), disable=not accelerator.is_local_main_process)
progress_bar.set_description(f"Epoch {epoch}")
for step, batch in enumerate(train_dataloader):
clean_images = batch["images"]
# Sample noise to add to the images
noise = torch.randn(clean_images.shape, device=clean_images.device)
bs = clean_images.shape[0]
# Sample a random timestep for each image
timesteps = torch.randint(
0, noise_scheduler.config.num_train_timesteps, (bs,), device=clean_images.device,
dtype=torch.int64
)
# Add noise to the clean images according to the noise magnitude at each timestep
# (this is the forward diffusion process)
noisy_images = noise_scheduler.add_noise(clean_images, noise, timesteps)
with accelerator.accumulate(model):
# Predict the noise residual
noise_pred = model(noisy_images, timesteps, return_dict=False)[0]
loss = F.mse_loss(noise_pred, noise)
accelerator.backward(loss)
accelerator.clip_grad_norm_(model.parameters(), 1.0)
optimizer.step()
lr_scheduler.step()
optimizer.zero_grad()
progress_bar.update(1)
logs = {"loss": loss.detach().item(), "lr": lr_scheduler.get_last_lr()[0], "step": global_step}
progress_bar.set_postfix(**logs)
accelerator.log(logs, step=global_step)
global_step += 1
# After each epoch you optionally sample some demo images with evaluate() and save the model
if accelerator.is_main_process:
pipeline = DDPMPipeline(unet=accelerator.unwrap_model(model), scheduler=noise_scheduler)
if (epoch + 1) % config.save_image_epochs == 0 or epoch == config.num_epochs - 1:
evaluate(config, epoch, pipeline)
if (epoch + 1) % config.save_model_epochs == 0 or epoch == config.num_epochs - 1:
if config.push_to_hub:
upload_folder(
repo_id=repo_id,
folder_path=config.output_dir,
commit_message=f"Epoch {epoch}",
ignore_patterns=["step_*", "epoch_*"],
)
else:
pipeline.save_pretrained(config.output_dir)
启动训练
**Accelerate 的notebook_launcher**函数启动训练了。向函数传递训练循环、所有训练参数和进程数(您可以将此值更改为可用的 GPU 数量)以用于训练:
from accelerate import notebook_launcher
args = (config, model, noise_scheduler, optimizer, train_dataloader, lr_scheduler)
notebook_launcher(train_loop, args, num_processes=1)
import glob
sample_images = sorted(glob.glob(f"{config.output_dir}/samples/*.png"))
Image.open(sample_images[-1])
增强
无条件图像生成是可训练任务的一个示例。您可以访问** 扩散器培训示例**页面来探索其他任务和培训技术。以下是您可以学到的一些示例:
- 文本反转,一种向模型传授特定视觉概念并将其集成到生成图像中的算法。
- DreamBooth,一种在给定主题的多个输入图像的情况下生成主题的个性化图像的技术。
- 在您自己的数据集上微调稳定扩散模型的**指南。**
- LoRA 使用**指南,这是一种内存高效技术,可以更快地微调大型模型。**