Training a Diffuser

Learn how to train a diffuser

Setup basic parameters

params = Parameters().from_json('../config.json')
_ = params.from_json('../tokens.json')

Since the model checkpoints are quite large, install Git-LFS to version these large files:

!sudo apt -qq install git-lfs
!git config --global credential.helper store

git-lfs is already the newest version (2.9.2-1).
0 upgraded, 0 newly installed, 0 to remove and 20 not upgraded.

Set training hyperparameters

params.config = Parameters(
    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, #10
    save_model_epochs = 30, #30
    mixed_precision = "fp16",  # `no` for float32, `fp16` for automatic mixed precision
    output_dir = f"{params.path.output}/ddpm-butterflies-128",  # the model name locally and on the HF Hub
    push_to_hub = False,  # whether to upload the saved model to the HF Hub
    hub_private_repo = False,
    overwrite_output_dir = True,  # overwrite the old model when re-running the notebook
    seed = 0)

Load the dataset

Load the Smithsonian Butterflies dataset with the 🤗 Datasets library:

params.config.dataset_name = "huggan/smithsonian_butterflies_subset"
dataset = load_dataset(params.config.dataset_name, split="train")

Let’s explore the dataset and look at some images

print(f'The dataset has {len(dataset)} images')
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()

The dataset has 1000 images

We want to create a transform that would resize the image to the same size, do random flip and create a tensor

preprocess = transforms.Compose(
    [
        transforms.Resize((params.config.image_size, params.config.image_size)),
        transforms.RandomHorizontalFlip(),
        transforms.ToTensor(),
        transforms.Normalize([0.5], [0.5]),
    ]
)

def transform(examples):
    images = [preprocess(image.convert("RGB")) for image in examples["image"]]
    return {"images": images}


dataset.set_transform(transform)

Define dataloader

train_dataloader = torch.utils.data.DataLoader(dataset, batch_size=params.config.train_batch_size, shuffle=True)

Create a UNet2DModel

model = UNet2DModel(
    sample_size=params.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
    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",
    ),
)

Checking the model sample output’s shape matches the input’s shape

sample_image = dataset[0]["images"].unsqueeze(0)
print("Input shape:", sample_image.shape)

print("Output shape:", model(sample_image, timestep=0).sample.shape)

Input shape: torch.Size([1, 3, 128, 128])
Output shape: torch.Size([1, 3, 128, 128])

Create a scheduler

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])

Define loss and prediction

noise_pred = model(noisy_image, timesteps).sample
loss = F.mse_loss(noise_pred, noise)

Train the model

optimizer and a learning rate scheduler

optimizer = torch.optim.AdamW(model.parameters(), lr=params.config.learning_rate)
lr_scheduler = get_cosine_schedule_with_warmup(
    optimizer=optimizer,
    num_warmup_steps=params.config.lr_warmup_steps,
    num_training_steps=(len(train_dataloader) * params.config.num_epochs),
)

Evaluation functions

def make_grid(images, rows, cols):
    w, h = images[0].size
    grid = Image.new("RGB", size=(cols * w, rows * h))
    for i, image in enumerate(images):
        grid.paste(image, box=(i % cols * w, i // cols * h))
    return grid


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=params.config.eval_batch_size,
        generator=torch.manual_seed(config.seed),
    ).images

    # Make a grid out of the images
    image_grid = make_grid(images, rows=4, cols=4)

    # Save the images
    test_dir = os.path.join(config.output_dir, f"samples")
    os.makedirs(test_dir, exist_ok=True)
    image_grid.save(f"{test_dir}/{epoch:04d}_{config.seed}.png")

Training Loop

def get_full_repo_name(model_id: str, organization: str = None, token: str = None):
    if token is None:
        token = HfFolder.get_token()
    if organization is None:
        username = whoami(token)["name"]
        return f"{username}/{model_id}"
    else:
        return f"{organization}/{model_id}"


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",
        logging_dir=os.path.join(config.output_dir, "logs"),
    )
    if accelerator.is_main_process:
        if config.push_to_hub:
            repo_name = get_full_repo_name(Path(config.output_dir).name)
            repo = Repository(config.output_dir, clone_from=repo_name)
        elif config.output_dir is not None:
            os.makedirs(config.output_dir, exist_ok=True)
        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).to(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
            ).long()

            # 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:
                    repo.push_to_hub(commit_message=f"Epoch {epoch}", blocking=True)
                else:
                    pipeline.save_pretrained(config.output_dir)

Let’s train

args = (params.config, model, noise_scheduler, optimizer, train_dataloader, lr_scheduler)

notebook_launcher(train_loop, args, num_processes=1)

Launching training on one GPU.

View the output

sample_images = sorted(glob.glob(f"{params.config.output_dir}/samples/*.png"))
Image.open(sample_images[-1])

sample_images = sorted(glob.glob(f"{params.config.output_dir}/samples/*.png"))
Image.open(sample_images[-2])

Image.open(sample_images[-3])