stable_diffusion/ldm/modules/diffusionmodules/util.py

# adopted from
# https://github.com/openai/improved-diffusion/blob/main/improved_diffusion/gaussian_diffusion.py
# and
# https://github.com/lucidrains/denoising-diffusion-pytorch/blob/7706bdfc6f527f58d33f84b7b522e61e6e3164b3/denoising_diffusion_pytorch/denoising_diffusion_pytorch.py
# and
# https://github.com/openai/guided-diffusion/blob/0ba878e517b276c45d1195eb29f6f5f72659a05b/guided_diffusion/nn.py
#
# thanks!


import os
import math
import torch
import torch.nn as nn
import numpy as np
from einops import repeat

import sys
sys.path.append('.')

from stable_diffusion.ldm.util import instantiate_from_config


def make_beta_schedule(schedule, n_timestep, linear_start=1e-4, linear_end=2e-2, cosine_s=8e-3):
    if schedule == "linear":
        betas = (
                torch.linspace(linear_start ** 0.5, linear_end ** 0.5, n_timestep, dtype=torch.float64) ** 2
        )

    elif schedule == "cosine":
        timesteps = (
                torch.arange(n_timestep + 1, dtype=torch.float64) / n_timestep + cosine_s
        )
        alphas = timesteps / (1 + cosine_s) * np.pi / 2
        alphas = torch.cos(alphas).pow(2)
        alphas = alphas / alphas[0]
        betas = 1 - alphas[1:] / alphas[:-1]
        betas = np.clip(betas, a_min=0, a_max=0.999)

    elif schedule == "sqrt_linear":
        betas = torch.linspace(linear_start, linear_end, n_timestep, dtype=torch.float64)
    elif schedule == "sqrt":
        betas = torch.linspace(linear_start, linear_end, n_timestep, dtype=torch.float64) ** 0.5
    else:
        raise ValueError(f"schedule '{schedule}' unknown.")
    return betas.numpy()


def make_ddim_timesteps(ddim_discr_method, num_ddim_timesteps, num_ddpm_timesteps, verbose=True):
    if ddim_discr_method == 'uniform':
        c = num_ddpm_timesteps // num_ddim_timesteps
        ddim_timesteps = np.asarray(list(range(0, num_ddpm_timesteps, c)))
    elif ddim_discr_method == 'quad':
        ddim_timesteps = ((np.linspace(0, np.sqrt(num_ddpm_timesteps * .8), num_ddim_timesteps)) ** 2).astype(int)
    else:
        raise NotImplementedError(f'There is no ddim discretization method called "{ddim_discr_method}"')

    # assert ddim_timesteps.shape[0] == num_ddim_timesteps
    # add one to get the final alpha values right (the ones from first scale to data during sampling)
    steps_out = ddim_timesteps + 1
    if verbose:
        print(f'Selected timesteps for ddim sampler: {steps_out}')
    return steps_out


def make_ddim_sampling_parameters(alphacums, ddim_timesteps, eta, verbose=True):
    # select alphas for computing the variance schedule
    alphas = alphacums[ddim_timesteps]
    alphas_prev = np.asarray([alphacums[0]] + alphacums[ddim_timesteps[:-1]].tolist())

    # according the the formula provided in https://arxiv.org/abs/2010.02502
    sigmas = eta * np.sqrt((1 - alphas_prev) / (1 - alphas) * (1 - alphas / alphas_prev))
    if verbose:
        print(f'Selected alphas for ddim sampler: a_t: {alphas}; a_(t-1): {alphas_prev}')
        print(f'For the chosen value of eta, which is {eta}, '
              f'this results in the following sigma_t schedule for ddim sampler {sigmas}')
    return sigmas, alphas, alphas_prev


def betas_for_alpha_bar(num_diffusion_timesteps, alpha_bar, max_beta=0.999):
    """
    Create a beta schedule that discretizes the given alpha_t_bar function,
    which defines the cumulative product of (1-beta) over time from t = [0,1].
    :param num_diffusion_timesteps: the number of betas to produce.
    :param alpha_bar: a lambda that takes an argument t from 0 to 1 and
                      produces the cumulative product of (1-beta) up to that
                      part of the diffusion process.
    :param max_beta: the maximum beta to use; use values lower than 1 to
                     prevent singularities.
    """
    betas = []
    for i in range(num_diffusion_timesteps):
        t1 = i / num_diffusion_timesteps
        t2 = (i + 1) / num_diffusion_timesteps
        betas.append(min(1 - alpha_bar(t2) / alpha_bar(t1), max_beta))
    return np.array(betas)


def extract_into_tensor(a, t, x_shape):
    b, *_ = t.shape
    out = a.gather(-1, t)
    return out.reshape(b, *((1,) * (len(x_shape) - 1)))


def checkpoint(func, inputs, params, flag):
    """
    Evaluate a function without caching intermediate activations, allowing for
    reduced memory at the expense of extra compute in the backward pass.
    :param func: the function to evaluate.
    :param inputs: the argument sequence to pass to `func`.
    :param params: a sequence of parameters `func` depends on but does not
                   explicitly take as arguments. This is also the parameters which we calculate the gradient with respect to
    :param flag: if False, disable gradient checkpointing.
    """

    # Filter out parameters that don't require gradients
    params_with_grads = [p for p in params if p.requires_grad]

    if flag:
        args = tuple(inputs) + tuple(params_with_grads)
        return CheckpointFunction.apply(func, len(inputs), *args)
    else:
        return func(*inputs)


class CheckpointFunction(torch.autograd.Function):
    """
    This is a method to save memory when training the model.
    """
    @staticmethod
    def forward(ctx, run_function, length, *args):  # args = inputs + params
        """
        The forward calls the function but without calculating the gradients.
        As indicated by the with torch.no_grad(). The parameter args contains
        the inputs and also the parameters w.r.t. which we want to calcuate
        the gradient.
        """
        ctx.run_function = run_function
        ctx.input_tensors = list(args[:length])  # inputs
        ctx.input_params = list(args[length:])  # params

        with torch.no_grad():
            output_tensors = ctx.run_function(*ctx.input_tensors)
        return output_tensors

    @staticmethod
    def backward(ctx, *output_grads):
        """
        The backward method takes in the gradients of the outputs
        (from the forward method), and computes the gradients of
        the inputs and parameters. This is done by re-running the
        run_function with gradients enabled, and then using
        torch.autograd.grad to calculate the gradients.
        These gradients are then returned.

        This technique of re-running the forward pass during the
        backward pass to re-compute the values required for gradient
        computation is known as gradient checkpointing. It can
        significantly reduce memory usage during training at the cost
        of increased computation time, as certain intermediate values
        from the forward pass don't need to be stored for the backward pass.
        """
        ctx.input_tensors = [x.detach().requires_grad_(True) for x in ctx.input_tensors]
        with torch.enable_grad():
            # Fixes a bug where the first op in run_function modifies the
            # Tensor storage in place, which is not allowed for detach()'d
            # Tensors.
            shallow_copies = [x.view_as(x) for x in ctx.input_tensors]
            output_tensors = ctx.run_function(*shallow_copies)
        input_grads = torch.autograd.grad(
            output_tensors,
            ctx.input_tensors + ctx.input_params,
            output_grads,
            allow_unused=True,
        )
        del ctx.input_tensors
        del ctx.input_params
        del output_tensors
        return (None, None) + input_grads


def timestep_embedding(timesteps, dim, max_period=10000, repeat_only=False):
    """
    Create sinusoidal timestep embeddings.
    :param timesteps: a 1-D Tensor of N indices, one per batch element.
                      These may be fractional.
    :param dim: the dimension of the output.
    :param max_period: controls the minimum frequency of the embeddings.
    :return: an [N x dim] Tensor of positional embeddings.
    """
    if not repeat_only:
        half = dim // 2
        freqs = torch.exp(
            -math.log(max_period) * torch.arange(start=0, end=half, dtype=torch.float32) / half
        ).to(device=timesteps.device)
        args = timesteps[:, None].float() * freqs[None]
        embedding = torch.cat([torch.cos(args), torch.sin(args)], dim=-1)
        if dim % 2:
            embedding = torch.cat([embedding, torch.zeros_like(embedding[:, :1])], dim=-1)
    else:
        embedding = repeat(timesteps, 'b -> b d', d=dim)
    return embedding


def zero_module(module):
    """
    Zero out the parameters of a module and return it.
    """
    for p in module.parameters():
        p.detach().zero_()
    return module


def scale_module(module, scale):
    """
    Scale the parameters of a module and return it.
    """
    for p in module.parameters():
        p.detach().mul_(scale)
    return module


def mean_flat(tensor):
    """
    Take the mean over all non-batch dimensions.
    """
    return tensor.mean(dim=list(range(1, len(tensor.shape))))


def normalization(channels):
    """
    Make a standard normalization layer.
    :param channels: number of input channels.
    :return: an nn.Module for normalization.
    """
    return GroupNorm32(32, channels)


# PyTorch 1.7 has SiLU, but we support PyTorch 1.5.
class SiLU(nn.Module):
    def forward(self, x):
        return x * torch.sigmoid(x)


class GroupNorm32(nn.GroupNorm):
    """
    Group Normalization is a method that divides the channels of your
    inputs into groups and normalizes these groups separately. This is
    in contrast to Batch Normalization, which normalizes each channel
    across the batch dimension.

    torch.nn.GroupNorm used to apply Group Normalization to a 4D or 5D
    input (a mini-batch of input features).

    The difference in the GroupNorm32 class is that it changes the data
    type of the input tensor to float before passing it to the original
    GroupNorm layer, and then changes it back to the original data type
    after the GroupNorm layer has processed it.

    This can be useful in situations where the input tensor is not of
    type float. PyTorch's normalization layers expect the input to be
    a float tensor, and may behave unexpectedly if given an integer
    tensor or a tensor of a different floating-point precision. By
    explicitly changing the tensor to float before the normalization,
    this class ensures that the normalization layer always receives a
    tensor of the correct type.
    """
    def forward(self, x):
        return super().forward(x.float()).type(x.dtype)


def conv_nd(dims, *args, **kwargs):
    """
    Create a 1D, 2D, or 3D convolution module.
    """
    if dims == 1:
        return nn.Conv1d(*args, **kwargs)
    elif dims == 2:
        return nn.Conv2d(*args, **kwargs)
    elif dims == 3:
        return nn.Conv3d(*args, **kwargs)
    raise ValueError(f"unsupported dimensions: {dims}")


def linear(*args, **kwargs):
    """
    Create a linear module.
    """
    return nn.Linear(*args, **kwargs)


def avg_pool_nd(dims, *args, **kwargs):
    """
    Create a 1D, 2D, or 3D average pooling module.
    """
    if dims == 1:
        return nn.AvgPool1d(*args, **kwargs)
    elif dims == 2:
        return nn.AvgPool2d(*args, **kwargs)
    elif dims == 3:
        return nn.AvgPool3d(*args, **kwargs)
    raise ValueError(f"unsupported dimensions: {dims}")


class HybridConditioner(nn.Module):

    def __init__(self, c_concat_config, c_crossattn_config):
        super().__init__()
        self.concat_conditioner = instantiate_from_config(c_concat_config)
        self.crossattn_conditioner = instantiate_from_config(c_crossattn_config)

    def forward(self, c_concat, c_crossattn):
        c_concat = self.concat_conditioner(c_concat)
        c_crossattn = self.crossattn_conditioner(c_crossattn)
        return {'c_concat': [c_concat], 'c_crossattn': [c_crossattn]}


def noise_like(shape, device, repeat=False):
    repeat_noise = lambda: torch.randn((1, *shape[1:]), device=device).repeat(shape[0], *((1,) * (len(shape) - 1)))
    noise = lambda: torch.randn(shape, device=device)
    return repeat_noise() if repeat else noise()