IDDPM论文阅读与复现

IDDPM介绍

Improved Denoising Diffusion Probabilistic Models(简写为IDDPM)是对原始的DDPM进行的改进模型。通过一系列改进措施，该模型可以获得更好的对数似然，并且可以提升采样速度。在后续的许多新的扩散模型中都采用了IDDPM的方法。

论文：Improved Denoising Diffusion Probabilistic Models Alex Nichol Prafulla Dhariwal

DDPM回顾

定义

给定初始数据分布，我们定义了一个前向的扩散过程，前向过程通过逐步添加噪声生成从到的隐变量，噪声方差。前向过程的联合概率分布如下所示，是一个马尔可夫过程： 当时间充分大时，可以认为服从标准正态分布。

反向过程是从噪声中恢复原始分布的过程。我们用神经网络去拟合这一过程： 和的组合可以看成是一个变分自编码器，我们可以把变分下界写成下面的形式： 除了之外，中的每一项都是两个高斯分布之间的KL散度，因此可以用封闭的形式来计算。其中和都不依赖于，且后者接近于0。

定义以及我们可以写出前向扩散过程任意时刻的分布： 其中是一个标准正态分布的随机噪声。

利用贝叶斯公式，我们可以写出反向过程的均值、方差以及后验概率分布的表达式：

训练

用神经网络去预测反向过程中每一步噪声的均值（也可以说是预测噪声，两者是等价的） 经过推导可以得到损失函数的形式如下：

有了损失函数，就可以利用随机梯度下降法或者其他优化方法进行训练。

IDDPM

IDDPM主要在以下几个方面对DDPM进行了改进：参数化方差、修改噪声策略、减小梯度噪声、加快采样速度、增加条件信息。

学习

DDPM直接将反向过程的方差固定为或，并且发现这两种选项对生成的样本质量影响不大，这是因为除了在最初时刻和差异比较大之外，大部分时刻两者差异是非常小的，尤其是随着扩散步数的增加，这两者几乎是相等的，如下图所示

虽然方差的选择对最终生成的样本的质量影响很小，但是对对数似然值会有影响。在扩散的前几步对变分下界（这里其实是负对数似然函数的上界）的影响比较大，因此选用合适的方差可以更好的提升对数似然值，因此，我们可以学习得到这个参数。

在IDDPM中，方差不再是一个固定的值，而是由神经网络去学习拟合。由于它的值太小，不太容易直接学习，因此我们转化为学习一个系数，具体来说，是将前向过程的方差和反向过程的方差进行加权，如下所示： 这里的就是我们要学习的量。

原先的与没有关系，为了使其与有关，我们将损失函数写为一种混合的形式： 论文中将设置为0.0001。

修改噪声随时间的变化规律

DDPM添加噪声的方法是随时间变化线性增长的，但我们会发现这样在扩散过程的最后的时间段噪声程度已经很大，这一段对提高样本质量没有什么帮助，因此IDDPM提出一种非线性的cosine方法，核心思想为让在接近以及处变化较小:

减小损失函数梯度的噪声

IDDPM更希望直接优化而不是优化，但是实际上它非常难以优化，下图是分别只优化和优化得到的损失函数曲线：

可以看到，两条曲线都非常不光滑，但是优化显然可以获得更小的损失。我们考虑到这可能是由于训练时时间均匀采样的结果。如果我们采用下面的方法对进行重要性采样： 我们得到的曲线就如上图绿色曲线一样，可以更好的优化对数似然，也减小了梯度噪声。重要性采样直观来说就是，在大的地方多采一些点，在小的地方少采一些点。但是重要性采样的方法仅对优化起作用，对于优化没有作用。

加快采样速度

Diffusion模型的最大问题为采样速度较慢。因此作者提出一种简单有效的加速采样方法，其核心思想为以均匀的间隔时间步进行采样，即相隔个时间步进行采样，其中为采样的时间步数。

加入条件信息

采样结果

IDDPM代码理解

补充一些Python语法知识

• Argument Parser：命令行参数解析器

• Python字典的upgrade方法：对于字典不存在键值，使用update会添加键值对到字典中；对于字典已经存在键值，使用update会更新键的值。

account = "optics"
decrypt_passwd = "12345abc"

accountDict = {}
accountDict.update({account:decrypt_passwd})
print(accountDict)

输出如下所示： {‘optics’: ‘12345abc’}

account = "optics"
decrypt_passwd = "abcdef"

accountDict = {'optics': '12345abc'}
accountDict.update({account:decrypt_passwd})
print(accountDict)

输出如下所示： {‘optics’: ‘abcdef’}

---

下面我们将会对IDDPM的代码进行逐模块剖析。

命令行传参技巧

'''从字典中自动生成命令行的argument parser，非常方便'''
def create_argparser():
    defaults = dict(
        clip_denoised=True,
        num_samples=10000,
        batch_size=16,
        use_ddim=False,
        model_path="",
    )
    defaults.update(model_and_diffusion_defaults())
    parser = argparse.ArgumentParser()
    add_dict_to_argparser(parser, defaults)
    return parser

def model_and_diffusion_defaults():
    """
    Defaults for image training.
    """
    return dict(
        image_size=64,
        num_channels=128,
        num_res_blocks=2,
        num_heads=4,
        num_heads_upsample=-1,
        attention_resolutions="16,8",
        dropout=0.0,
        learn_sigma=False,
        sigma_small=False,
        class_cond=False,
        diffusion_steps=1000,
        noise_schedule="linear",
        timestep_respacing="",
        use_kl=False,
        predict_xstart=False,
        rescale_timesteps=True,
        rescale_learned_sigmas=True,
        use_checkpoint=False,
        use_scale_shift_norm=True,
    )

# 从命令行中读取参数，参数的键与刚才定义的字典里的一样，不用再一一敲出来
def add_dict_to_argparser(parser, default_dict):
    for k, v in default_dict.items():
        v_type = type(v)
        if v is None:
            v_type = str
        elif isinstance(v, bool):
            v_type = str2bool
        parser.add_argument(f"--{k}", default=v, type=v_type)

生成模型和扩散过程

def create_model_and_diffusion(
    image_size,
    class_cond,
    learn_sigma,
    sigma_small,
    num_channels,
    num_res_blocks,
    num_heads,
    num_heads_upsample,
    attention_resolutions,
    dropout,
    diffusion_steps,
    noise_schedule,
    timestep_respacing,
    use_kl,
    predict_xstart,
    rescale_timesteps,
    rescale_learned_sigmas,
    use_checkpoint,
    use_scale_shift_norm,
):
    model = create_model(
        image_size,
        num_channels,
        num_res_blocks,
        learn_sigma=learn_sigma,
        class_cond=class_cond,
        use_checkpoint=use_checkpoint,
        attention_resolutions=attention_resolutions,
        num_heads=num_heads,
        num_heads_upsample=num_heads_upsample,
        use_scale_shift_norm=use_scale_shift_norm,
        dropout=dropout,
    )
    diffusion = create_gaussian_diffusion(
        steps=diffusion_steps,
        learn_sigma=learn_sigma,
        sigma_small=sigma_small,
        noise_schedule=noise_schedule,
        use_kl=use_kl,
        predict_xstart=predict_xstart,
        rescale_timesteps=rescale_timesteps,
        rescale_learned_sigmas=rescale_learned_sigmas,
        timestep_respacing=timestep_respacing,
    )
    return model, diffusion

生成高斯扩散过程

def create_gaussian_diffusion(
    *,
    steps=1000,
    learn_sigma=False,
    sigma_small=False,
    noise_schedule="linear",
    use_kl=False,
    predict_xstart=False,
    rescale_timesteps=False,
    rescale_learned_sigmas=False,
    timestep_respacing="",
):
    # 生成一个扩散过程的框架
    # 得到一个噪声方案
    betas = gd.get_named_beta_schedule(noise_schedule, steps)
    # 选择损失函数
    if use_kl:
        loss_type = gd.LossType.RESCALED_KL
    elif rescale_learned_sigmas:
        loss_type = gd.LossType.RESCALED_MSE
    else:
        loss_type = gd.LossType.MSE
    if not timestep_respacing:
        timestep_respacing = [steps]
    return SpacedDiffusion(
        use_timesteps=space_timesteps(steps, timestep_respacing),
        betas=betas,
        model_mean_type=(
            gd.ModelMeanType.EPSILON if not predict_xstart else gd.ModelMeanType.START_X
        ),
        model_var_type=(
            (
                gd.ModelVarType.FIXED_LARGE
                if not sigma_small
                else gd.ModelVarType.FIXED_SMALL
            )
            if not learn_sigma
            else gd.ModelVarType.LEARNED_RANGE
        ),
        loss_type=loss_type,
        rescale_timesteps=rescale_timesteps,
    )
    
# 选择加噪方案
def get_named_beta_schedule(schedule_name, num_diffusion_timesteps):
    """
    Get a pre-defined beta schedule for the given name.

    The beta schedule library consists of beta schedules which remain similar
    in the limit of num_diffusion_timesteps.
    Beta schedules may be added, but should not be removed or changed once
    they are committed to maintain backwards compatibility.
    """
    if schedule_name == "linear":
        # 线性加噪方案
        # Linear schedule from Ho et al, extended to work for any number of
        # diffusion steps.
        scale = 1000 / num_diffusion_timesteps
        beta_start = scale * 0.0001
        beta_end = scale * 0.02
        return np.linspace(
            beta_start, beta_end, num_diffusion_timesteps, dtype=np.float64
        )
        # 余弦加噪方案
    elif schedule_name == "cosine":
        return betas_for_alpha_bar(
            num_diffusion_timesteps,
            lambda t: math.cos((t + 0.008) / 1.008 * math.pi / 2) ** 2,
        )
    else:
        raise NotImplementedError(f"unknown beta schedule: {schedule_name}")
        

# 间隔选取timestep
class SpacedDiffusion(GaussianDiffusion):
    """
    A diffusion process which can skip steps in a base diffusion process.

    :param use_timesteps: a collection (sequence or set) of timesteps from the
                          original diffusion process to retain.
    :param kwargs: the kwargs to create the base diffusion process.
    """

    def __init__(self, use_timesteps, **kwargs):
        self.use_timesteps = set(use_timesteps)
        self.timestep_map = []
        self.original_num_steps = len(kwargs["betas"])

        base_diffusion = GaussianDiffusion(**kwargs)  # pylint: disable=missing-kwoa
        last_alpha_cumprod = 1.0
        new_betas = []
        for i, alpha_cumprod in enumerate(base_diffusion.alphas_cumprod):
            if i in self.use_timesteps:
                new_betas.append(1 - alpha_cumprod / last_alpha_cumprod)
                last_alpha_cumprod = alpha_cumprod
                self.timestep_map.append(i)
        kwargs["betas"] = np.array(new_betas)
        super().__init__(**kwargs)

    def p_mean_variance(
        self, model, *args, **kwargs
    ):  # pylint: disable=signature-differs
        return super().p_mean_variance(self._wrap_model(model), *args, **kwargs)

    def training_losses(
        self, model, *args, **kwargs
    ):  # pylint: disable=signature-differs
        return super().training_losses(self._wrap_model(model), *args, **kwargs)

    def _wrap_model(self, model):
        if isinstance(model, _WrappedModel):
            return model
        return _WrappedModel(
            model, self.timestep_map, self.rescale_timesteps, self.original_num_steps
        )

    def _scale_timesteps(self, t):
        # Scaling is done by the wrapped model.
        return t
    
# 高斯扩散类
class GaussianDiffusion:
    """
    Utilities for training and sampling diffusion models.

    Ported directly from here, and then adapted over time to further experimentation.
    https://github.com/hojonathanho/diffusion/blob/1e0dceb3b3495bbe19116a5e1b3596cd0706c543/diffusion_tf/diffusion_utils_2.py#L42

    :param betas: a 1-D numpy array of betas for each diffusion timestep,
                  starting at T and going to 1.
    :param model_mean_type: a ModelMeanType determining what the model outputs.
    :param model_var_type: a ModelVarType determining how variance is output.
    :param loss_type: a LossType determining the loss function to use.
    :param rescale_timesteps: if True, pass floating point timesteps into the
                              model so that they are always scaled like in the
                              original paper (0 to 1000).
    """

    def __init__(
        self,
        *,
        betas,
        model_mean_type,
        model_var_type,
        loss_type,
        rescale_timesteps=False,
    ):
        self.model_mean_type = model_mean_type
        self.model_var_type = model_var_type
        self.loss_type = loss_type
        self.rescale_timesteps = rescale_timesteps

        # Use float64 for accuracy.
        betas = np.array(betas, dtype=np.float64)
        self.betas = betas
        assert len(betas.shape) == 1, "betas must be 1-D"
        assert (betas > 0).all() and (betas <= 1).all()

        self.num_timesteps = int(betas.shape[0])

        alphas = 1.0 - betas
        self.alphas_cumprod = np.cumprod(alphas, axis=0)
        self.alphas_cumprod_prev = np.append(1.0, self.alphas_cumprod[:-1])
        self.alphas_cumprod_next = np.append(self.alphas_cumprod[1:], 0.0)
        assert self.alphas_cumprod_prev.shape == (self.num_timesteps,)

        # calculations for diffusion q(x_t | x_{t-1}) and others
        self.sqrt_alphas_cumprod = np.sqrt(self.alphas_cumprod)
        self.sqrt_one_minus_alphas_cumprod = np.sqrt(1.0 - self.alphas_cumprod)
        self.log_one_minus_alphas_cumprod = np.log(1.0 - self.alphas_cumprod)
        self.sqrt_recip_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod)
        self.sqrt_recipm1_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod - 1)

        # calculations for posterior q(x_{t-1} | x_t, x_0)
        self.posterior_variance = (
            betas * (1.0 - self.alphas_cumprod_prev) / (1.0 - self.alphas_cumprod)
        )
        # log calculation clipped because the posterior variance is 0 at the
        # beginning of the diffusion chain.
        self.posterior_log_variance_clipped = np.log(
            np.append(self.posterior_variance[1], self.posterior_variance[1:])
        )
        self.posterior_mean_coef1 = (
            betas * np.sqrt(self.alphas_cumprod_prev) / (1.0 - self.alphas_cumprod)
        )
        self.posterior_mean_coef2 = (
            (1.0 - self.alphas_cumprod_prev)
            * np.sqrt(alphas)
            / (1.0 - self.alphas_cumprod)
        )

    def q_mean_variance(self, x_start, t):
        """
        Get the distribution q(x_t | x_0).

        :param x_start: the [N x C x ...] tensor of noiseless inputs.
        :param t: the number of diffusion steps (minus 1). Here, 0 means one step.
        :return: A tuple (mean, variance, log_variance), all of x_start's shape.
        """
        mean = (
            _extract_into_tensor(self.sqrt_alphas_cumprod, t, x_start.shape) * x_start
        )
        variance = _extract_into_tensor(1.0 - self.alphas_cumprod, t, x_start.shape)
        log_variance = _extract_into_tensor(
            self.log_one_minus_alphas_cumprod, t, x_start.shape
        )
        return mean, variance, log_variance

    def q_sample(self, x_start, t, noise=None):
        """
        Diffuse the data for a given number of diffusion steps.

        In other words, sample from q(x_t | x_0).

        :param x_start: the initial data batch.
        :param t: the number of diffusion steps (minus 1). Here, 0 means one step.
        :param noise: if specified, the split-out normal noise.
        :return: A noisy version of x_start.
        """
        if noise is None:
            noise = th.randn_like(x_start)
        assert noise.shape == x_start.shape
        return (
            _extract_into_tensor(self.sqrt_alphas_cumprod, t, x_start.shape) * x_start
            + _extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x_start.shape)
            * noise
        )

    def q_posterior_mean_variance(self, x_start, x_t, t):
        """
        Compute the mean and variance of the diffusion posterior:

            q(x_{t-1} | x_t, x_0)

        """
        assert x_start.shape == x_t.shape
        posterior_mean = (
            _extract_into_tensor(self.posterior_mean_coef1, t, x_t.shape) * x_start
            + _extract_into_tensor(self.posterior_mean_coef2, t, x_t.shape) * x_t
        )
        posterior_variance = _extract_into_tensor(self.posterior_variance, t, x_t.shape)
        posterior_log_variance_clipped = _extract_into_tensor(
            self.posterior_log_variance_clipped, t, x_t.shape
        )
        assert (
            posterior_mean.shape[0]
            == posterior_variance.shape[0]
            == posterior_log_variance_clipped.shape[0]
            == x_start.shape[0]
        )
        return posterior_mean, posterior_variance, posterior_log_variance_clipped

    def p_mean_variance(
        self, model, x, t, clip_denoised=True, denoised_fn=None, model_kwargs=None
    ):
        """
        Apply the model to get p(x_{t-1} | x_t), as well as a prediction of
        the initial x, x_0.

        :param model: the model, which takes a signal and a batch of timesteps
                      as input.
        :param x: the [N x C x ...] tensor at time t.
        :param t: a 1-D Tensor of timesteps.
        :param clip_denoised: if True, clip the denoised signal into [-1, 1].
        :param denoised_fn: if not None, a function which applies to the
            x_start prediction before it is used to sample. Applies before
            clip_denoised.
        :param model_kwargs: if not None, a dict of extra keyword arguments to
            pass to the model. This can be used for conditioning.
        :return: a dict with the following keys:
                 - 'mean': the model mean output.
                 - 'variance': the model variance output.
                 - 'log_variance': the log of 'variance'.
                 - 'pred_xstart': the prediction for x_0.
        """
        if model_kwargs is None:
            model_kwargs = {}

        B, C = x.shape[:2]
        assert t.shape == (B,)
        model_output = model(x, self._scale_timesteps(t), **model_kwargs)

        if self.model_var_type in [ModelVarType.LEARNED, ModelVarType.LEARNED_RANGE]:
            assert model_output.shape == (B, C * 2, *x.shape[2:])
            model_output, model_var_values = th.split(model_output, C, dim=1)
            if self.model_var_type == ModelVarType.LEARNED:
                model_log_variance = model_var_values
                model_variance = th.exp(model_log_variance)
            else:
                min_log = _extract_into_tensor(
                    self.posterior_log_variance_clipped, t, x.shape
                )
                max_log = _extract_into_tensor(np.log(self.betas), t, x.shape)
                # The model_var_values is [-1, 1] for [min_var, max_var].
                frac = (model_var_values + 1) / 2
                model_log_variance = frac * max_log + (1 - frac) * min_log
                model_variance = th.exp(model_log_variance)
        else:
            model_variance, model_log_variance = {
                # for fixedlarge, we set the initial (log-)variance like so
                # to get a better decoder log likelihood.
                ModelVarType.FIXED_LARGE: (
                    np.append(self.posterior_variance[1], self.betas[1:]),
                    np.log(np.append(self.posterior_variance[1], self.betas[1:])),
                ),
                ModelVarType.FIXED_SMALL: (
                    self.posterior_variance,
                    self.posterior_log_variance_clipped,
                ),
            }[self.model_var_type]
            model_variance = _extract_into_tensor(model_variance, t, x.shape)
            model_log_variance = _extract_into_tensor(model_log_variance, t, x.shape)

        def process_xstart(x):
            if denoised_fn is not None:
                x = denoised_fn(x)
            if clip_denoised:
                return x.clamp(-1, 1)
            return x

        if self.model_mean_type == ModelMeanType.PREVIOUS_X:
            pred_xstart = process_xstart(
                self._predict_xstart_from_xprev(x_t=x, t=t, xprev=model_output)
            )
            model_mean = model_output
        elif self.model_mean_type in [ModelMeanType.START_X, ModelMeanType.EPSILON]:
            if self.model_mean_type == ModelMeanType.START_X:
                pred_xstart = process_xstart(model_output)
            else:
                pred_xstart = process_xstart(
                    self._predict_xstart_from_eps(x_t=x, t=t, eps=model_output)
                )
            model_mean, _, _ = self.q_posterior_mean_variance(
                x_start=pred_xstart, x_t=x, t=t
            )
        else:
            raise NotImplementedError(self.model_mean_type)

        assert (
            model_mean.shape == model_log_variance.shape == pred_xstart.shape == x.shape
        )
        return {
            "mean": model_mean,
            "variance": model_variance,
            "log_variance": model_log_variance,
            "pred_xstart": pred_xstart,
        }

    def _predict_xstart_from_eps(self, x_t, t, eps):
        assert x_t.shape == eps.shape
        return (
            _extract_into_tensor(self.sqrt_recip_alphas_cumprod, t, x_t.shape) * x_t
            - _extract_into_tensor(self.sqrt_recipm1_alphas_cumprod, t, x_t.shape) * eps
        )

    def _predict_xstart_from_xprev(self, x_t, t, xprev):
        assert x_t.shape == xprev.shape
        return (  # (xprev - coef2*x_t) / coef1
            _extract_into_tensor(1.0 / self.posterior_mean_coef1, t, x_t.shape) * xprev
            - _extract_into_tensor(
                self.posterior_mean_coef2 / self.posterior_mean_coef1, t, x_t.shape
            )
            * x_t
        )

    def _predict_eps_from_xstart(self, x_t, t, pred_xstart):
        return (
            _extract_into_tensor(self.sqrt_recip_alphas_cumprod, t, x_t.shape) * x_t
            - pred_xstart
        ) / _extract_into_tensor(self.sqrt_recipm1_alphas_cumprod, t, x_t.shape)

    def _scale_timesteps(self, t):
        if self.rescale_timesteps:
            return t.float() * (1000.0 / self.num_timesteps)
        return t

    def p_sample(
        self, model, x, t, clip_denoised=True, denoised_fn=None, model_kwargs=None
    ):
        """
        Sample x_{t-1} from the model at the given timestep.

        :param model: the model to sample from.
        :param x: the current tensor at x_{t-1}.
        :param t: the value of t, starting at 0 for the first diffusion step.
        :param clip_denoised: if True, clip the x_start prediction to [-1, 1].
        :param denoised_fn: if not None, a function which applies to the
            x_start prediction before it is used to sample.
        :param model_kwargs: if not None, a dict of extra keyword arguments to
            pass to the model. This can be used for conditioning.
        :return: a dict containing the following keys:
                 - 'sample': a random sample from the model.
                 - 'pred_xstart': a prediction of x_0.
        """
        out = self.p_mean_variance(
            model,
            x,
            t,
            clip_denoised=clip_denoised,
            denoised_fn=denoised_fn,
            model_kwargs=model_kwargs,
        )
        noise = th.randn_like(x)
        nonzero_mask = (
            (t != 0).float().view(-1, *([1] * (len(x.shape) - 1)))
        )  # no noise when t == 0
        sample = out["mean"] + nonzero_mask * th.exp(0.5 * out["log_variance"]) * noise
        return {"sample": sample, "pred_xstart": out["pred_xstart"]}

    def p_sample_loop(
        self,
        model,
        shape,
        noise=None,
        clip_denoised=True,
        denoised_fn=None,
        model_kwargs=None,
        device=None,
        progress=False,
    ):
        """
        Generate samples from the model.

        :param model: the model module.
        :param shape: the shape of the samples, (N, C, H, W).
        :param noise: if specified, the noise from the encoder to sample.
                      Should be of the same shape as `shape`.
        :param clip_denoised: if True, clip x_start predictions to [-1, 1].
        :param denoised_fn: if not None, a function which applies to the
            x_start prediction before it is used to sample.
        :param model_kwargs: if not None, a dict of extra keyword arguments to
            pass to the model. This can be used for conditioning.
        :param device: if specified, the device to create the samples on.
                       If not specified, use a model parameter's device.
        :param progress: if True, show a tqdm progress bar.
        :return: a non-differentiable batch of samples.
        """
        final = None
        for sample in self.p_sample_loop_progressive(
            model,
            shape,
            noise=noise,
            clip_denoised=clip_denoised,
            denoised_fn=denoised_fn,
            model_kwargs=model_kwargs,
            device=device,
            progress=progress,
        ):
            final = sample
        return final["sample"]

    def p_sample_loop_progressive(
        self,
        model,
        shape,
        noise=None,
        clip_denoised=True,
        denoised_fn=None,
        model_kwargs=None,
        device=None,
        progress=False,
    ):
        """
        Generate samples from the model and yield intermediate samples from
        each timestep of diffusion.

        Arguments are the same as p_sample_loop().
        Returns a generator over dicts, where each dict is the return value of
        p_sample().
        """
        if device is None:
            device = next(model.parameters()).device
        assert isinstance(shape, (tuple, list))
        if noise is not None:
            img = noise
        else:
            img = th.randn(*shape, device=device)
        indices = list(range(self.num_timesteps))[::-1]

        if progress:
            # Lazy import so that we don't depend on tqdm.
            from tqdm.auto import tqdm

            indices = tqdm(indices)

        for i in indices:
            t = th.tensor([i] * shape[0], device=device)
            with th.no_grad():
                out = self.p_sample(
                    model,
                    img,
                    t,
                    clip_denoised=clip_denoised,
                    denoised_fn=denoised_fn,
                    model_kwargs=model_kwargs,
                )
                yield out
                img = out["sample"]

    def ddim_sample(
        self,
        model,
        x,
        t,
        clip_denoised=True,
        denoised_fn=None,
        model_kwargs=None,
        eta=0.0,
    ):
        """
        Sample x_{t-1} from the model using DDIM.

        Same usage as p_sample().
        """
        out = self.p_mean_variance(
            model,
            x,
            t,
            clip_denoised=clip_denoised,
            denoised_fn=denoised_fn,
            model_kwargs=model_kwargs,
        )
        # Usually our model outputs epsilon, but we re-derive it
        # in case we used x_start or x_prev prediction.
        eps = self._predict_eps_from_xstart(x, t, out["pred_xstart"])
        alpha_bar = _extract_into_tensor(self.alphas_cumprod, t, x.shape)
        alpha_bar_prev = _extract_into_tensor(self.alphas_cumprod_prev, t, x.shape)
        sigma = (
            eta
            * th.sqrt((1 - alpha_bar_prev) / (1 - alpha_bar))
            * th.sqrt(1 - alpha_bar / alpha_bar_prev)
        )
        # Equation 12.
        noise = th.randn_like(x)
        mean_pred = (
            out["pred_xstart"] * th.sqrt(alpha_bar_prev)
            + th.sqrt(1 - alpha_bar_prev - sigma ** 2) * eps
        )
        nonzero_mask = (
            (t != 0).float().view(-1, *([1] * (len(x.shape) - 1)))
        )  # no noise when t == 0
        sample = mean_pred + nonzero_mask * sigma * noise
        return {"sample": sample, "pred_xstart": out["pred_xstart"]}

    def ddim_reverse_sample(
        self,
        model,
        x,
        t,
        clip_denoised=True,
        denoised_fn=None,
        model_kwargs=None,
        eta=0.0,
    ):
        """
        Sample x_{t+1} from the model using DDIM reverse ODE.
        """
        assert eta == 0.0, "Reverse ODE only for deterministic path"
        out = self.p_mean_variance(
            model,
            x,
            t,
            clip_denoised=clip_denoised,
            denoised_fn=denoised_fn,
            model_kwargs=model_kwargs,
        )
        # Usually our model outputs epsilon, but we re-derive it
        # in case we used x_start or x_prev prediction.
        eps = (
            _extract_into_tensor(self.sqrt_recip_alphas_cumprod, t, x.shape) * x
            - out["pred_xstart"]
        ) / _extract_into_tensor(self.sqrt_recipm1_alphas_cumprod, t, x.shape)
        alpha_bar_next = _extract_into_tensor(self.alphas_cumprod_next, t, x.shape)

        # Equation 12. reversed
        mean_pred = (
            out["pred_xstart"] * th.sqrt(alpha_bar_next)
            + th.sqrt(1 - alpha_bar_next) * eps
        )

        return {"sample": mean_pred, "pred_xstart": out["pred_xstart"]}

    def ddim_sample_loop(
        self,
        model,
        shape,
        noise=None,
        clip_denoised=True,
        denoised_fn=None,
        model_kwargs=None,
        device=None,
        progress=False,
        eta=0.0,
    ):
        """
        Generate samples from the model using DDIM.

        Same usage as p_sample_loop().
        """
        final = None
        for sample in self.ddim_sample_loop_progressive(
            model,
            shape,
            noise=noise,
            clip_denoised=clip_denoised,
            denoised_fn=denoised_fn,
            model_kwargs=model_kwargs,
            device=device,
            progress=progress,
            eta=eta,
        ):
            final = sample
        return final["sample"]

    def ddim_sample_loop_progressive(
        self,
        model,
        shape,
        noise=None,
        clip_denoised=True,
        denoised_fn=None,
        model_kwargs=None,
        device=None,
        progress=False,
        eta=0.0,
    ):
        """
        Use DDIM to sample from the model and yield intermediate samples from
        each timestep of DDIM.

        Same usage as p_sample_loop_progressive().
        """
        if device is None:
            device = next(model.parameters()).device
        assert isinstance(shape, (tuple, list))
        if noise is not None:
            img = noise
        else:
            img = th.randn(*shape, device=device)
        indices = list(range(self.num_timesteps))[::-1]

        if progress:
            # Lazy import so that we don't depend on tqdm.
            from tqdm.auto import tqdm

            indices = tqdm(indices)

        for i in indices:
            t = th.tensor([i] * shape[0], device=device)
            with th.no_grad():
                out = self.ddim_sample(
                    model,
                    img,
                    t,
                    clip_denoised=clip_denoised,
                    denoised_fn=denoised_fn,
                    model_kwargs=model_kwargs,
                    eta=eta,
                )
                yield out
                img = out["sample"]

    def _vb_terms_bpd(
        self, model, x_start, x_t, t, clip_denoised=True, model_kwargs=None
    ):
        """
        Get a term for the variational lower-bound.

        The resulting units are bits (rather than nats, as one might expect).
        This allows for comparison to other papers.

        :return: a dict with the following keys:
                 - 'output': a shape [N] tensor of NLLs or KLs.
                 - 'pred_xstart': the x_0 predictions.
        """
        true_mean, _, true_log_variance_clipped = self.q_posterior_mean_variance(
            x_start=x_start, x_t=x_t, t=t
        )
        out = self.p_mean_variance(
            model, x_t, t, clip_denoised=clip_denoised, model_kwargs=model_kwargs
        )
        kl = normal_kl(
            true_mean, true_log_variance_clipped, out["mean"], out["log_variance"]
        )
        kl = mean_flat(kl) / np.log(2.0)

        decoder_nll = -discretized_gaussian_log_likelihood(
            x_start, means=out["mean"], log_scales=0.5 * out["log_variance"]
        )
        assert decoder_nll.shape == x_start.shape
        decoder_nll = mean_flat(decoder_nll) / np.log(2.0)

        # At the first timestep return the decoder NLL,
        # otherwise return KL(q(x_{t-1}|x_t,x_0) || p(x_{t-1}|x_t))
        output = th.where((t == 0), decoder_nll, kl)
        return {"output": output, "pred_xstart": out["pred_xstart"]}

    def training_losses(self, model, x_start, t, model_kwargs=None, noise=None):
        """
        Compute training losses for a single timestep.

        :param model: the model to evaluate loss on.
        :param x_start: the [N x C x ...] tensor of inputs.
        :param t: a batch of timestep indices.
        :param model_kwargs: if not None, a dict of extra keyword arguments to
            pass to the model. This can be used for conditioning.
        :param noise: if specified, the specific Gaussian noise to try to remove.
        :return: a dict with the key "loss" containing a tensor of shape [N].
                 Some mean or variance settings may also have other keys.
        """
        if model_kwargs is None:
            model_kwargs = {}
        if noise is None:
            noise = th.randn_like(x_start)
        x_t = self.q_sample(x_start, t, noise=noise)

        terms = {}

        if self.loss_type == LossType.KL or self.loss_type == LossType.RESCALED_KL:
            terms["loss"] = self._vb_terms_bpd(
                model=model,
                x_start=x_start,
                x_t=x_t,
                t=t,
                clip_denoised=False,
                model_kwargs=model_kwargs,
            )["output"]
            if self.loss_type == LossType.RESCALED_KL:
                terms["loss"] *= self.num_timesteps
        elif self.loss_type == LossType.MSE or self.loss_type == LossType.RESCALED_MSE:
            model_output = model(x_t, self._scale_timesteps(t), **model_kwargs)

            if self.model_var_type in [
                ModelVarType.LEARNED,
                ModelVarType.LEARNED_RANGE,
            ]:
                B, C = x_t.shape[:2]
                assert model_output.shape == (B, C * 2, *x_t.shape[2:])
                model_output, model_var_values = th.split(model_output, C, dim=1)
                # Learn the variance using the variational bound, but don't let
                # it affect our mean prediction.
                frozen_out = th.cat([model_output.detach(), model_var_values], dim=1)
                terms["vb"] = self._vb_terms_bpd(
                    model=lambda *args, r=frozen_out: r,
                    x_start=x_start,
                    x_t=x_t,
                    t=t,
                    clip_denoised=False,
                )["output"]
                if self.loss_type == LossType.RESCALED_MSE:
                    # Divide by 1000 for equivalence with initial implementation.
                    # Without a factor of 1/1000, the VB term hurts the MSE term.
                    terms["vb"] *= self.num_timesteps / 1000.0

            target = {
                ModelMeanType.PREVIOUS_X: self.q_posterior_mean_variance(
                    x_start=x_start, x_t=x_t, t=t
                )[0],
                ModelMeanType.START_X: x_start,
                ModelMeanType.EPSILON: noise,
            }[self.model_mean_type]
            assert model_output.shape == target.shape == x_start.shape
            terms["mse"] = mean_flat((target - model_output) ** 2)
            if "vb" in terms:
                terms["loss"] = terms["mse"] + terms["vb"]
            else:
                terms["loss"] = terms["mse"]
        else:
            raise NotImplementedError(self.loss_type)

        return terms

    def _prior_bpd(self, x_start):
        """
        Get the prior KL term for the variational lower-bound, measured in
        bits-per-dim.

        This term can't be optimized, as it only depends on the encoder.

        :param x_start: the [N x C x ...] tensor of inputs.
        :return: a batch of [N] KL values (in bits), one per batch element.
        """
        batch_size = x_start.shape[0]
        t = th.tensor([self.num_timesteps - 1] * batch_size, device=x_start.device)
        qt_mean, _, qt_log_variance = self.q_mean_variance(x_start, t)
        kl_prior = normal_kl(
            mean1=qt_mean, logvar1=qt_log_variance, mean2=0.0, logvar2=0.0
        )
        return mean_flat(kl_prior) / np.log(2.0)

    def calc_bpd_loop(self, model, x_start, clip_denoised=True, model_kwargs=None):
        """
        Compute the entire variational lower-bound, measured in bits-per-dim,
        as well as other related quantities.

        :param model: the model to evaluate loss on.
        :param x_start: the [N x C x ...] tensor of inputs.
        :param clip_denoised: if True, clip denoised samples.
        :param model_kwargs: if not None, a dict of extra keyword arguments to
            pass to the model. This can be used for conditioning.

        :return: a dict containing the following keys:
                 - total_bpd: the total variational lower-bound, per batch element.
                 - prior_bpd: the prior term in the lower-bound.
                 - vb: an [N x T] tensor of terms in the lower-bound.
                 - xstart_mse: an [N x T] tensor of x_0 MSEs for each timestep.
                 - mse: an [N x T] tensor of epsilon MSEs for each timestep.
        """
        device = x_start.device
        batch_size = x_start.shape[0]

        vb = []
        xstart_mse = []
        mse = []
        for t in list(range(self.num_timesteps))[::-1]:
            t_batch = th.tensor([t] * batch_size, device=device)
            noise = th.randn_like(x_start)
            x_t = self.q_sample(x_start=x_start, t=t_batch, noise=noise)
            # Calculate VLB term at the current timestep
            with th.no_grad():
                out = self._vb_terms_bpd(
                    model,
                    x_start=x_start,
                    x_t=x_t,
                    t=t_batch,
                    clip_denoised=clip_denoised,
                    model_kwargs=model_kwargs,
                )
            vb.append(out["output"])
            xstart_mse.append(mean_flat((out["pred_xstart"] - x_start) ** 2))
            eps = self._predict_eps_from_xstart(x_t, t_batch, out["pred_xstart"])
            mse.append(mean_flat((eps - noise) ** 2))

        vb = th.stack(vb, dim=1)
        xstart_mse = th.stack(xstart_mse, dim=1)
        mse = th.stack(mse, dim=1)

        prior_bpd = self._prior_bpd(x_start)
        total_bpd = vb.sum(dim=1) + prior_bpd
        return {
            "total_bpd": total_bpd,
            "prior_bpd": prior_bpd,
            "vb": vb,
            "xstart_mse": xstart_mse,
            "mse": mse,
        }


def _extract_into_tensor(arr, timesteps, broadcast_shape):
    """
    Extract values from a 1-D numpy array for a batch of indices.

    :param arr: the 1-D numpy array.
    :param timesteps: a tensor of indices into the array to extract.
    :param broadcast_shape: a larger shape of K dimensions with the batch
                            dimension equal to the length of timesteps.
    :return: a tensor of shape [batch_size, 1, ...] where the shape has K dims.
    """
    res = th.from_numpy(arr).to(device=timesteps.device)[timesteps].float()
    while len(res.shape) < len(broadcast_shape):
        res = res[..., None]
    return res.expand(broadcast_shape)

确定采样方法

def create_named_schedule_sampler(name, diffusion):
    """
    Create a ScheduleSampler from a library of pre-defined samplers.

    :param name: the name of the sampler.
    :param diffusion: the diffusion object to sample for.
    """
    # 均匀采样
    if name == "uniform":
        return UniformSampler(diffusion)
    # 基于Loss重要性采样
    elif name == "loss-second-moment":
        return LossSecondMomentResampler(diffusion)
    else:
        raise NotImplementedError(f"unknown schedule sampler: {name}")

加载图片和图片预处理

def load_data(
    *, data_dir, batch_size, image_size, class_cond=False, deterministic=False
):
    """
    For a dataset, create a generator over (images, kwargs) pairs.

    Each images is an NCHW float tensor, and the kwargs dict contains zero or
    more keys, each of which map to a batched Tensor of their own.
    The kwargs dict can be used for class labels, in which case the key is "y"
    and the values are integer tensors of class labels.

    :param data_dir: a dataset directory.
    :param batch_size: the batch size of each returned pair.
    :param image_size: the size to which images are resized.
    :param class_cond: if True, include a "y" key in returned dicts for class
                       label. If classes are not available and this is true, an
                       exception will be raised.
    :param deterministic: if True, yield results in a deterministic order.
    """
    if not data_dir:
        raise ValueError("unspecified data directory")
    all_files = _list_image_files_recursively(data_dir)
    classes = None
    if class_cond:
        # Assume classes are the first part of the filename,
        # before an underscore.
        class_names = [bf.basename(path).split("_")[0] for path in all_files]
        sorted_classes = {x: i for i, x in enumerate(sorted(set(class_names)))}
        classes = [sorted_classes[x] for x in class_names]
    dataset = ImageDataset(
        image_size,
        all_files,
        classes=classes,
        shard=MPI.COMM_WORLD.Get_rank(),
        num_shards=MPI.COMM_WORLD.Get_size(),
    )
    if deterministic:
        loader = DataLoader(
            dataset, batch_size=batch_size, shuffle=False, num_workers=1, drop_last=True
        )
    else:
        loader = DataLoader(
            dataset, batch_size=batch_size, shuffle=True, num_workers=1, drop_last=True
        )
    while True:
        yield from loader


def _list_image_files_recursively(data_dir):
    results = []
    for entry in sorted(bf.listdir(data_dir)):
        full_path = bf.join(data_dir, entry)
        ext = entry.split(".")[-1]
        if "." in entry and ext.lower() in ["jpg", "jpeg", "png", "gif"]:
            results.append(full_path)
        elif bf.isdir(full_path):
            results.extend(_list_image_files_recursively(full_path))
    return results


class ImageDataset(Dataset):
    def __init__(self, resolution, image_paths, classes=None, shard=0, num_shards=1):
        super().__init__()
        self.resolution = resolution
        self.local_images = image_paths[shard:][::num_shards]
        self.local_classes = None if classes is None else classes[shard:][::num_shards]

    def __len__(self):
        return len(self.local_images)

    def __getitem__(self, idx):
        path = self.local_images[idx]
        with bf.BlobFile(path, "rb") as f:
            pil_image = Image.open(f)
            pil_image.load()

        # We are not on a new enough PIL to support the `reducing_gap`
        # argument, which uses BOX downsampling at powers of two first.
        # Thus, we do it by hand to improve downsample quality.
        while min(*pil_image.size) >= 2 * self.resolution:
            pil_image = pil_image.resize(
                tuple(x // 2 for x in pil_image.size), resample=Image.BOX
            )

        scale = self.resolution / min(*pil_image.size)
        pil_image = pil_image.resize(
            tuple(round(x * scale) for x in pil_image.size), resample=Image.BICUBIC
        )

        arr = np.array(pil_image.convert("RGB"))
        crop_y = (arr.shape[0] - self.resolution) // 2
        crop_x = (arr.shape[1] - self.resolution) // 2
        arr = arr[crop_y : crop_y + self.resolution, crop_x : crop_x + self.resolution]
        arr = arr.astype(np.float32) / 127.5 - 1

        out_dict = {}
        if self.local_classes is not None:
            out_dict["y"] = np.array(self.local_classes[idx], dtype=np.int64)
        return np.transpose(arr, [2, 0, 1]), out_dict

Attention Based U-Net

class TimestepEmbedSequential(nn.Sequential, TimestepBlock):
    """
    A sequential module that passes timestep embeddings to the children that
    support it as an extra input.
    """

    def forward(self, x, emb):
        for layer in self:
            # 只有是TimestepBlock的子类才传入emb
            if isinstance(layer, TimestepBlock):
                x = layer(x, emb)
            # 否则不输入emb
            else:
                x = layer(x)
        return x

class Upsample(nn.Module):
    """
    An upsampling layer with an optional convolution.

    :param channels: channels in the inputs and outputs.
    :param use_conv: a bool determining if a convolution is applied.
    :param dims: determines if the signal is 1D, 2D, or 3D. If 3D, then
                 upsampling occurs in the inner-two dimensions.
    """

    def __init__(self, channels, use_conv, dims=2):
        super().__init__()
        self.channels = channels
        self.use_conv = use_conv
        self.dims = dims
        if use_conv:
            self.conv = conv_nd(dims, channels, channels, 3, padding=1)

    def forward(self, x):
        assert x.shape[1] == self.channels
        if self.dims == 3:
            x = F.interpolate(
                x, (x.shape[2], x.shape[3] * 2, x.shape[4] * 2), mode="nearest"
            )
        else:
            x = F.interpolate(x, scale_factor=2, mode="nearest")
        if self.use_conv:
            x = self.conv(x)
        return x

class Downsample(nn.Module):
    """
    A downsampling layer with an optional convolution.

    :param channels: channels in the inputs and outputs.
    :param use_conv: a bool determining if a convolution is applied.
    :param dims: determines if the signal is 1D, 2D, or 3D. If 3D, then
                 downsampling occurs in the inner-two dimensions.
    """

    def __init__(self, channels, use_conv, dims=2):
        super().__init__()
        self.channels = channels
        self.use_conv = use_conv
        self.dims = dims
        stride = 2 if dims != 3 else (1, 2, 2)
        # 使用卷积或池化进行下采样，缩小空间维度
        if use_conv:
            self.op = conv_nd(dims, channels, channels, 3, stride=stride, padding=1)
        else:
            self.op = avg_pool_nd(stride)

    def forward(self, x):
        assert x.shape[1] == self.channels
        return self.op(x)

class AttentionBlock(nn.Module):
    """
    An attention block that allows spatial positions to attend to each other.

    Originally ported from here, but adapted to the N-d case.
    https://github.com/hojonathanho/diffusion/blob/1e0dceb3b3495bbe19116a5e1b3596cd0706c543/diffusion_tf/models/unet.py#L66.
    """

    def __init__(self, channels, num_heads=1, use_checkpoint=False):
        super().__init__()
        self.channels = channels
        self.num_heads = num_heads
        self.use_checkpoint = use_checkpoint

        self.norm = normalization(channels)
        self.qkv = conv_nd(1, channels, channels * 3, 1)
        self.attention = QKVAttention()
        self.proj_out = zero_module(conv_nd(1, channels, channels, 1))

    def forward(self, x):
        return checkpoint(self._forward, (x,), self.parameters(), self.use_checkpoint)

    def _forward(self, x):
        b, c, *spatial = x.shape
        x = x.reshape(b, c, -1)
        qkv = self.qkv(self.norm(x))
        qkv = qkv.reshape(b * self.num_heads, -1, qkv.shape[2])
        h = self.attention(qkv)
        h = h.reshape(b, -1, h.shape[-1])
        h = self.proj_out(h)
        return (x + h).reshape(b, c, *spatial)  # 带残差

class QKVAttention(nn.Module):
    """
    A module which performs QKV attention.
    """

    def forward(self, qkv):
        """
        Apply QKV attention.

        :param qkv: an [N x (C * 3) x T] tensor of Qs, Ks, and Vs.
        :return: an [N x C x T] tensor after attention.
        """
        ch = qkv.shape[1] // 3
        q, k, v = th.split(qkv, ch, dim=1)
        scale = 1 / math.sqrt(math.sqrt(ch))
        weight = th.einsum(
            "bct,bcs->bts", q * scale, k * scale
        )  # More stable with f16 than dividing afterwards
        weight = th.softmax(weight.float(), dim=-1).type(weight.dtype)
        return th.einsum("bts,bcs->bct", weight, v)

    @staticmethod
    def count_flops(model, _x, y):
        """
        A counter for the `thop` package to count the operations in an
        attention operation.

        Meant to be used like:

            macs, params = thop.profile(
                model,
                inputs=(inputs, timestamps),
                custom_ops={QKVAttention: QKVAttention.count_flops},
            )

        """
        b, c, *spatial = y[0].shape
        num_spatial = int(np.prod(spatial))
        # We perform two matmuls with the same number of ops.
        # The first computes the weight matrix, the second computes
        # the combination of the value vectors.
        matmul_ops = 2 * b * (num_spatial ** 2) * c
        model.total_ops += th.DoubleTensor([matmul_ops])

class ResBlock(TimestepBlock):
    """
    A residual block that can optionally change the number of channels.

    :param channels: the number of input channels.
    :param emb_channels: the number of timestep embedding channels.
    :param dropout: the rate of dropout.
    :param out_channels: if specified, the number of out channels.
    :param use_conv: if True and out_channels is specified, use a spatial
        convolution instead of a smaller 1x1 convolution to change the
        channels in the skip connection.
    :param dims: determines if the signal is 1D, 2D, or 3D.
    :param use_checkpoint: if True, use gradient checkpointing on this module.
    """

    def __init__(
        self,
        channels,
        emb_channels,
        dropout,
        out_channels=None,
        use_conv=False,
        use_scale_shift_norm=False,
        dims=2,
        use_checkpoint=False,
    ):
        super().__init__()
        self.channels = channels
        self.emb_channels = emb_channels
        self.dropout = dropout
        self.out_channels = out_channels or channels
        self.use_conv = use_conv
        self.use_checkpoint = use_checkpoint
        self.use_scale_shift_norm = use_scale_shift_norm

        self.in_layers = nn.Sequential(
            normalization(channels),
            SiLU(),
            conv_nd(dims, channels, self.out_channels, 3, padding=1),
        )
        self.emb_layers = nn.Sequential(
            SiLU(),
            linear(
                emb_channels,
                2 * self.out_channels if use_scale_shift_norm else self.out_channels,
            ),
        )
        self.out_layers = nn.Sequential(
            normalization(self.out_channels),
            SiLU(),
            nn.Dropout(p=dropout),
            zero_module(
                conv_nd(dims, self.out_channels, self.out_channels, 3, padding=1)
            ),
        )
		
        # 判断通道数是否相等
        if self.out_channels == channels:
            self.skip_connection = nn.Identity()
        elif use_conv:  # 如果不相等，通过一个卷积层使得输入输出通道数相等
            self.skip_connection = conv_nd(
                dims, channels, self.out_channels, 3, padding=1
            )
        else:
            self.skip_connection = conv_nd(dims, channels, self.out_channels, 1)

    def forward(self, x, emb):
        """
        Apply the block to a Tensor, conditioned on a timestep embedding.

        :param x: an [N x C x ...] Tensor of features.
        :param emb: an [N x emb_channels] Tensor of timestep embeddings.
        :return: an [N x C x ...] Tensor of outputs.
        """
        return checkpoint(
            self._forward, (x, emb), self.parameters(), self.use_checkpoint
        )

    def _forward(self, x, emb):
        h = self.in_layers(x)
        emb_out = self.emb_layers(emb).type(h.dtype)
        while len(emb_out.shape) < len(h.shape):
            emb_out = emb_out[..., None]
        if self.use_scale_shift_norm:
            out_norm, out_rest = self.out_layers[0], self.out_layers[1:]
            scale, shift = th.chunk(emb_out, 2, dim=1)
            h = out_norm(h) * (1 + scale) + shift
            h = out_rest(h)
        else:
            h = h + emb_out
            h = self.out_layers(h)
        return self.skip_connection(x) + h

class UNetModel(nn.Module):
    """
    The full UNet model with attention and timestep embedding.
	包含了注意力机制已经时间嵌入的U-Net模型
    :param in_channels: channels in the input Tensor.
    :param model_channels: base channel count for the model.
    :param out_channels: channels in the output Tensor.
    :param num_res_blocks: number of residual blocks per downsample.
    :param attention_resolutions: a collection of downsample rates at which
        attention will take place. May be a set, list, or tuple.
        For example, if this contains 4, then at 4x downsampling, attention
        will be used.
    :param dropout: the dropout probability.
    :param channel_mult: channel multiplier for each level of the UNet.
    :param conv_resample: if True, use learned convolutions for upsampling and
        downsampling.
    :param dims: determines if the signal is 1D, 2D, or 3D.
    :param num_classes: if specified (as an int), then this model will be
        class-conditional with `num_classes` classes.
    :param use_checkpoint: use gradient checkpointing to reduce memory usage.
    :param num_heads: the number of attention heads in each attention layer.
    """

    def __init__(
        self,
        in_channels,
        model_channels,
        out_channels,
        num_res_blocks,
        attention_resolutions,
        dropout=0,
        channel_mult=(1, 2, 4, 8),
        conv_resample=True,
        dims=2,
        num_classes=None,
        use_checkpoint=False,
        num_heads=1,
        num_heads_upsample=-1,
        use_scale_shift_norm=False,
    ):
        super().__init__()

        if num_heads_upsample == -1:
            num_heads_upsample = num_heads

        self.in_channels = in_channels
        self.model_channels = model_channels
        self.out_channels = out_channels
        self.num_res_blocks = num_res_blocks
        self.attention_resolutions = attention_resolutions
        self.dropout = dropout
        self.channel_mult = channel_mult
        self.conv_resample = conv_resample
        self.num_classes = num_classes
        self.use_checkpoint = use_checkpoint
        self.num_heads = num_heads
        self.num_heads_upsample = num_heads_upsample

        time_embed_dim = model_channels * 4
        # 时间嵌入层，包含两个线性层和一个非线性激活函数
        self.time_embed = nn.Sequential(
            linear(model_channels, time_embed_dim),
            SiLU(),
            linear(time_embed_dim, time_embed_dim),
        )
		# 如果包含了条件，也将条件嵌入进去
        if self.num_classes is not None:
            self.label_emb = nn.Embedding(num_classes, time_embed_dim)
		
        
        # input_blocks是由许多个ResBlock和Downsample组成的，对应U-Net左边的部分
        self.input_blocks = nn.ModuleList(
            [
                # 在这里并不传入emb，因为conv_nd并不是TimestepBlock的子类
                TimestepEmbedSequential(
                    conv_nd(dims, in_channels, model_channels, 3, padding=1)
                )
            ]
        )
        input_block_chans = [model_channels]
        ch = model_channels
        ds = 1  # 
        
        for level, mult in enumerate(channel_mult):
            for _ in range(num_res_blocks):
                layers = [
                    ResBlock(
                        ch,  # 输入通道数
                        time_embed_dim,
                        dropout,
                        out_channels=mult * model_channels,  # 输出通道数
                        dims=dims,
                        use_checkpoint=use_checkpoint,
                        use_scale_shift_norm=use_scale_shift_norm,
                    )
                ]
                ch = mult * model_channels
                if ds in attention_resolutions:
                    layers.append(
                        AttentionBlock(
                            ch, use_checkpoint=use_checkpoint, num_heads=num_heads
                        )
                    )
                self.input_blocks.append(TimestepEmbedSequential(*layers))
                input_block_chans.append(ch)
            if level != len(channel_mult) - 1:
                self.input_blocks.append(
                    TimestepEmbedSequential(Downsample(ch, conv_resample, dims=dims))
                )
                input_block_chans.append(ch)
                ds *= 2
		
        # U-Net的中间部分，由两个ResBlock和一个AttentionBlock组成的
        self.middle_block = TimestepEmbedSequential(
            ResBlock(
                ch,
                time_embed_dim,
                dropout,
                dims=dims,
                use_checkpoint=use_checkpoint,
                use_scale_shift_norm=use_scale_shift_norm,
            ),
            AttentionBlock(ch, use_checkpoint=use_checkpoint, num_heads=num_heads),
            ResBlock(
                ch,
                time_embed_dim,
                dropout,
                dims=dims,
                use_checkpoint=use_checkpoint,
                use_scale_shift_norm=use_scale_shift_norm,
            ),
        )
		
        # U-Net的右边部分，由多个ResBlock和Umsample组成的
        self.output_blocks = nn.ModuleList([])
        for level, mult in list(enumerate(channel_mult))[::-1]:
            for i in range(num_res_blocks + 1):
                layers = [
                    ResBlock(
                        ch + input_block_chans.pop(),
                        time_embed_dim,
                        dropout,
                        out_channels=model_channels * mult,
                        dims=dims,
                        use_checkpoint=use_checkpoint,
                        use_scale_shift_norm=use_scale_shift_norm,
                    )
                ]
                ch = model_channels * mult
                if ds in attention_resolutions:
                    layers.append(
                        AttentionBlock(
                            ch,
                            use_checkpoint=use_checkpoint,
                            num_heads=num_heads_upsample,
                        )
                    )
                if level and i == num_res_blocks:
                    layers.append(Upsample(ch, conv_resample, dims=dims))
                    ds //= 2 # 通道数在不断减小
                self.output_blocks.append(TimestepEmbedSequential(*layers))

        self.out = nn.Sequential(
            normalization(ch),
            SiLU(),
            zero_module(conv_nd(dims, model_channels, out_channels, 3, padding=1)),
        )

    def convert_to_fp16(self):
        """
        Convert the torso of the model to float16.
        """
        self.input_blocks.apply(convert_module_to_f16)
        self.middle_block.apply(convert_module_to_f16)
        self.output_blocks.apply(convert_module_to_f16)

    def convert_to_fp32(self):
        """
        Convert the torso of the model to float32.
        """
        self.input_blocks.apply(convert_module_to_f32)
        self.middle_block.apply(convert_module_to_f32)
        self.output_blocks.apply(convert_module_to_f32)

    @property
    def inner_dtype(self):
        """
        Get the dtype used by the torso of the model.
        """
        return next(self.input_blocks.parameters()).dtype
    
    # x,timestep,y是输入
    def forward(self, x, timesteps, y=None):
        """
        Apply the model to an input batch.

        :param x: an [N x C x ...] Tensor of inputs.
        :param timesteps: a 1-D batch of timesteps.
        :param y: an [N] Tensor of labels, if class-conditional.
        :return: an [N x C x ...] Tensor of outputs.
        """
        assert (y is not None) == (
            self.num_classes is not None
        ), "must specify y if and only if the model is class-conditional"

        hs = []
        emb = self.time_embed(timestep_embedding(timesteps, self.model_channels))

        if self.num_classes is not None:
            assert y.shape == (x.shape[0],)
            emb = emb + self.label_emb(y)

        h = x.type(self.inner_dtype)
        # 左边
        for module in self.input_blocks:
            h = module(h, emb)
            hs.append(h)
        # 中间
        h = self.middle_block(h, emb)
        # 右边
        for module in self.output_blocks:
            cat_in = th.cat([h, hs.pop()], dim=1)
            h = module(cat_in, emb)
        h = h.type(x.dtype)
        return self.out(h)

    def get_feature_vectors(self, x, timesteps, y=None):
        """
        Apply the model and return all of the intermediate tensors.

        :param x: an [N x C x ...] Tensor of inputs.
        :param timesteps: a 1-D batch of timesteps.
        :param y: an [N] Tensor of labels, if class-conditional.
        :return: a dict with the following keys:
                 - 'down': a list of hidden state tensors from downsampling.
                 - 'middle': the tensor of the output of the lowest-resolution
                             block in the model.
                 - 'up': a list of hidden state tensors from upsampling.
        """
        hs = []
        emb = self.time_embed(timestep_embedding(timesteps, self.model_channels))
        if self.num_classes is not None:
            assert y.shape == (x.shape[0],)
            emb = emb + self.label_emb(y)
        result = dict(down=[], up=[])
        h = x.type(self.inner_dtype)
        for module in self.input_blocks:
            h = module(h, emb)
            hs.append(h)
            result["down"].append(h.type(x.dtype))
        h = self.middle_block(h, emb)
        result["middle"] = h.type(x.dtype)
        for module in self.output_blocks:
            cat_in = th.cat([h, hs.pop()], dim=1)
            h = module(cat_in, emb)
            result["up"].append(h.type(x.dtype))
        return result

采样结果