""" This code started out as a PyTorch port of Ho et al's diffusion models: https://github.com/hojonathanho/diffusion/blob/1e0dceb3b3495bbe19116a5e1b3596cd0706c543/diffusion_tf/diffusion_utils_2.py Docstrings have been added, as well as DDIM sampling and a new collection of beta schedules. """ import enum import math import torch import numpy as np from .nn import mean_flat from .losses import normal_kl, discretized_gaussian_log_likelihood 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": # 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 ) 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, ) elif schedule_name == "sqrt": return betas_for_alpha_bar( num_diffusion_timesteps, lambda t: 1 - np.sqrt(t + 0.0001), ) elif schedule_name == "trunc_cos": return betas_for_alpha_bar2( num_diffusion_timesteps, lambda t: np.cos((t + 0.1) / 1.1 * np.pi / 2) ** 2, ) elif schedule_name == "trunc_lin": scale = 1000 / num_diffusion_timesteps beta_start = scale * 0.0001 + 0.01 beta_end = scale * 0.02 + 0.01 return np.linspace( beta_start, beta_end, num_diffusion_timesteps, dtype=np.float64 ) elif schedule_name == "pw_lin": scale = 1000 / num_diffusion_timesteps beta_start = scale * 0.0001 + 0.01 beta_mid = scale * 0.0001 # scale * 0.02 beta_end = scale * 0.02 first_part = np.linspace(beta_start, beta_mid, 10, dtype=np.float64) second_part = np.linspace( beta_mid, beta_end, num_diffusion_timesteps - 10, dtype=np.float64 ) return np.concatenate([first_part, second_part]) else: raise NotImplementedError(f"unknown beta schedule: {schedule_name}") def betas_for_alpha_bar2(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 = [] betas.append(min(1 - alpha_bar(0), max_beta)) for i in range(num_diffusion_timesteps - 1): 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 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) class ModelMeanType(enum.Enum): """ Which type of output the model predicts. """ PREVIOUS_X = enum.auto() # the model predicts x_{t-1} START_X = enum.auto() # the model predicts x_0 EPSILON = enum.auto() # the model predicts epsilon class ModelVarType(enum.Enum): """ What is used as the model's output variance. The LEARNED_RANGE option has been added to allow the model to predict values between FIXED_SMALL and FIXED_LARGE, making its job easier. """ LEARNED = enum.auto() FIXED_SMALL = enum.auto() FIXED_LARGE = enum.auto() LEARNED_RANGE = enum.auto() class LossType(enum.Enum): MSE = enum.auto() # use raw MSE loss (and KL when learning variances) RESCALED_MSE = ( enum.auto() ) # use raw MSE loss (with RESCALED_KL when learning variances) KL = enum.auto() # use the variational lower-bound RESCALED_KL = enum.auto() # like KL, but rescale to estimate the full VLB E2E_KL = enum.auto() E2E_MSE = enum.auto() E2E_Simple_MSE = enum.auto() E2E_Simple_KL = enum.auto() def is_vb(self): return self == LossType.KL or self == LossType.RESCALED_KL 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, model_arch=None, training_mode="emb", ): self.model_mean_type = model_mean_type self.model_var_type = model_var_type self.loss_type = loss_type self.rescale_timesteps = rescale_timesteps self.model_arch = model_arch # 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) ) self.training_mode = training_mode self.mapping_func = None # # if training_mode == 'e2e': # self.training_losses = self.training_losses_e2e # else: # self.training_losses = self.training_losses_emb self.maxt = -1 def training_losses(self, model, *args, **kwargs): return self.training_losses_e2e(model, *args, **kwargs) # if self.training_mode == "e2e": # return self.training_losses_e2e(model, *args, **kwargs) # elif self.training_mode == "e2e-simple": # return self.training_losses_e2e_simple(model, *args, **kwargs) # else: # return self.training_losses_emb(model, *args, **kwargs) def calc_bpd_loop(self, model, *args, **kwargs): if self.training_mode == "e2e": return self.calc_bpd_loop_e2e(model, *args, **kwargs) else: return self.calc_bpd_loop_emb(model, *args, **kwargs) 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 = torch.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, caption=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. """ caption_state, caption_mask = caption[0], caption[1] if model_kwargs is None: model_kwargs = {} if self.model_arch == "conv-unet" or self.model_arch == "1d-unet": B, C = x.shape[:2] else: B, C = x.size(0), x.size(-1) assert t.shape == (B,) # print(x.shape) model_output = model( x, self._scale_timesteps(t), caption_state, caption_mask, **model_kwargs ) if self.model_var_type in [ModelVarType.LEARNED, ModelVarType.LEARNED_RANGE]: if self.model_arch == "conv-unet": assert model_output.shape == (B, C * 2, *x.shape[2:]) model_output, model_var_values = torch.split(model_output, C, dim=1) # print('conv-unet') elif self.model_arch == "1d-unet": assert model_output.shape == (B, C * 2, *x.shape[2:]) model_output, model_var_values = torch.split(model_output, C, dim=1) else: assert model_output.shape == (B, x.size(1), C * 2) model_output, model_var_values = torch.split(model_output, C, dim=-1) if self.model_var_type == ModelVarType.LEARNED: model_log_variance = model_var_values model_variance = torch.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 = torch.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: # print(denoised_fn) x = denoised_fn(x, t) 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, top_p=None, caption=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, caption=caption, ) if top_p is not None and top_p > 0: # print('top_p sampling') noise = torch.randn_like(x) replace_mask = torch.abs(noise) > top_p while replace_mask.any(): noise[replace_mask] = torch.randn_like(noise[replace_mask]) replace_mask = torch.abs(noise) > top_p assert (torch.abs(noise) <= top_p).all() else: noise = torch.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 * torch.exp(0.5 * out["log_variance"]) * noise ) return { "sample": sample, "pred_xstart": out["pred_xstart"], "greedy_mean": out["mean"], "out": out, } def p_debug_loop( self, model, shape, noise=None, clip_denoised=True, denoised_fn=None, model_kwargs=None, device=None, progress=False, ): final = None for sample in self.p_debug_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_debug_loop_progressive( self, model, shape, noise=None, clip_denoised=True, denoised_fn=None, model_kwargs=None, device=None, progress=False, custom_t_start=100, ): """ 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 = torch.randn(*shape, device=device) indices = list(range(custom_t_start))[::-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 = torch.tensor([i] * shape[0], device=device) with torch.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 p_sample_loop( self, model, shape, noise=None, clip_denoised=True, denoised_fn=None, model_kwargs=None, device=None, progress=False, top_p=None, caption=None, ): """ 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, top_p=top_p, caption=caption, ): 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, top_p=None, caption=None, ): """ 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.to(device) else: img = torch.randn(*shape, device=device) indices = list(range(self.num_timesteps))[::-1] # print(indices[-10:]) # indices = indices[:-1]+[1,1,1,1,1,1,1]*60+[0] # print(indices[-10:]) if progress: # Lazy import so that we don't depend on tqdm. from tqdm.auto import tqdm indices = tqdm(indices) if caption is not None: print("Text Guiding Generation ......") caption = ( caption[0].to(img.device), caption[1].to(img.device), ) # (caption_state, caption_mask) for i in indices: t = torch.tensor([i] * shape[0], device=device) with torch.no_grad(): out = self.p_sample( model, img, t, clip_denoised=clip_denoised, denoised_fn=denoised_fn, model_kwargs=model_kwargs, top_p=top_p, caption=caption, ) yield out img = out["sample"] def p_sample_loop_langevin_progressive( self, model, shape, noise=None, clip_denoised=True, denoised_fn=None, model_kwargs=None, device=None, progress=False, langevin_func=None, top_p=None, ): """ 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 = torch.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 = torch.tensor([i] * shape[0], device=device) with torch.no_grad(): out = self.p_sample( model, img, t, clip_denoised=clip_denoised, denoised_fn=denoised_fn, model_kwargs=model_kwargs, top_p=top_p, ) if langevin_func is not None: out["t"] = t out["img"] = img out = langevin_func(out) yield out img = out["sample"] def p_sample_loop_progressive_infill( self, model, shape, partial_enc, partial_mask, noise=None, clip_denoised=True, denoised_fn=None, model_kwargs=None, device=None, progress=False, greedy=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 # img = img[partial_mask] + partial_enc_with_noise[~partial_mask] else: t_batch = torch.tensor([self.num_timesteps - 1] * shape[0], device=device) partial_enc_with_noise = self.q_sample(partial_enc, t_batch) img = torch.randn(*shape, device=device) # print(img.shape, partial_enc_with_noise.shape, partial_mask.shape) # img = img[partial_mask] + partial_enc_with_noise[~partial_mask] img[~partial_mask] = partial_enc_with_noise[~partial_mask] 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 = torch.tensor([i] * shape[0], device=device) with torch.no_grad(): out = self.p_sample( model, img, t, clip_denoised=clip_denoised, denoised_fn=denoised_fn, model_kwargs=model_kwargs, ) if i > 0: partial_enc_with_noise = self.q_sample(partial_enc, t - 1) else: partial_enc_with_noise = partial_enc if greedy: img = out["greedy_mean"] img[~partial_mask] = partial_enc[~partial_mask] out["sample"] = img else: img = out["sample"] img[~partial_mask] = partial_enc[~partial_mask] # img[~partial_mask] = partial_enc_with_noise[~partial_mask] out["sample"] = img yield out def p_sample_loop_progressive_merge( self, model, shape, partial_enc, partial_mask, noise=None, clip_denoised=True, denoised_fn=None, model_kwargs=None, device=None, progress=False, greedy=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 # img = img[partial_mask] + partial_enc_with_noise[~partial_mask] else: t_batch = torch.tensor([self.num_timesteps - 1] * shape[0], device=device) partial_enc_with_noise = self.q_sample(partial_enc, t_batch) img = torch.randn(*shape, device=device) # print(img.shape, partial_enc_with_noise.shape, partial_mask.shape) # img = img[partial_mask] + partial_enc_with_noise[~partial_mask] img[~partial_mask] = partial_enc_with_noise[~partial_mask] 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 = torch.tensor([i] * shape[0], device=device) with torch.no_grad(): out = self.p_sample( model, img, t, clip_denoised=clip_denoised, denoised_fn=denoised_fn, model_kwargs=model_kwargs, ) if i > 0: partial_enc_with_noise = self.q_sample(partial_enc, t - 1) else: partial_enc_with_noise = partial_enc if greedy: img = out["greedy_mean"] img[~partial_mask] = partial_enc[~partial_mask] out["sample"] = img else: img = out["sample"] img[~partial_mask] = partial_enc[~partial_mask] # img[~partial_mask] = partial_enc_with_noise[~partial_mask] out["sample"] = img yield out def ddim_sample( self, model, x, t, clip_denoised=True, denoised_fn=None, model_kwargs=None, eta=0.0, langevin_fn=None, caption=None, ): """ 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, caption=caption, ) # 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 * torch.sqrt((1 - alpha_bar_prev) / (1 - alpha_bar)) * torch.sqrt(1 - alpha_bar / alpha_bar_prev) ) # Equation 12. noise = torch.randn_like(x) mean_pred = ( out["pred_xstart"] * torch.sqrt(alpha_bar_prev) + torch.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 # print(sigma.mean()) sample = mean_pred + nonzero_mask * sigma * noise if langevin_fn: print(t.shape) sample = langevin_fn( sample, mean_pred, sigma, self.alphas_cumprod_prev[t[0]], t, x ) 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"] * torch.sqrt(alpha_bar_next) + torch.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, top_p=-1.0, langevin_fn=None, caption=None, ): """ 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, langevin_fn=langevin_fn, caption=caption, ): 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, langevin_fn=None, caption=None, ): """ 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 = torch.randn(*shape, device=device) indices = list(range(self.num_timesteps))[::-1] if caption is not None: print("Text Guiding Generation ......") caption = ( caption[0].to(img.device), caption[1].to(img.device), ) # (caption_state, caption_mask) 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 = torch.tensor([i] * shape[0], device=device) with torch.no_grad(): out = self.ddim_sample( model, img, t, clip_denoised=clip_denoised, denoised_fn=denoised_fn, model_kwargs=model_kwargs, eta=eta, langevin_fn=langevin_fn, caption=caption, ) yield out img = out["sample"] def _vb_terms_bpd( self, model, x_start, x_t, t, clip_denoised=True, model_kwargs=None, noise=None, denoised_fn=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. """ # lambda *args, r=frozen_out: r, true_mean, _, true_log_variance_clipped = self.q_posterior_mean_variance( x_start=x_start, x_t=x_t, t=t ) if model_kwargs is not None and "input_ids" in model_kwargs: input_ids = model_kwargs.pop("input_ids") mapping_func = model_kwargs.pop("mapping_func", self.mapping_func) else: input_ids = None # noise=None out = self.p_mean_variance( model, x_t, t, clip_denoised=clip_denoised, model_kwargs=model_kwargs, denoised_fn=denoised_fn, ) kl = normal_kl( true_mean, true_log_variance_clipped, out["mean"], out["log_variance"] ) kl = mean_flat(kl) / np.log(2.0) if input_ids is not None: # print('input_ids is not None') # from torch.distributions import Normal # normal_dist = Normal(out["mean"], (0.5 * out["log_variance"]).exp()) # decoder_nll = -normal_dist.log_prob(x_start) assert mapping_func is not None if mapping_func is not None and torch.any(t == 0): decoder_nll = mapping_func(out["mean"], input_ids) / out["mean"].size( -1 ) else: decoder_nll = torch.zeros_like(x_start) model_kwargs["input_ids"] = input_ids model_kwargs["mapping_func"] = mapping_func # 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] # # print(out['mean'].shape, x_start.shape, self.model_mean_type, noise) # assert out["mean"].shape == target.shape == x_start.shape # decoder_nll = (target - out["mean"]) ** 2 else: 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 = torch.where((t == 0), decoder_nll, kl) return {"output": output, "pred_xstart": out["pred_xstart"]} def _vb_terms_bpd_e2e( self, model, x_start, x_t, t, input_ids, get_logits, x_start_mean, x_start_log_var, clip_denoised=True, model_kwargs=None, noise=None, denoised_fn=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. """ # lambda *args, r=frozen_out: r, true_mean, _, true_log_variance_clipped = self.q_posterior_mean_variance( x_start=x_start, x_t=x_t, t=t ) assert input_ids is not None mapping_func = model_kwargs.pop("mapping_func", self.mapping_func) # assert 'input_ids' in model_kwargs # input_ids = model_kwargs.pop('input_ids') out = self.p_mean_variance( model, x_t, t, clip_denoised=clip_denoised, model_kwargs=model_kwargs, denoised_fn=denoised_fn, ) # print(true_log_variance_clipped[0], out["log_variance"][0], 'line1259') kl = normal_kl( true_mean, true_log_variance_clipped, out["mean"], out["log_variance"] ) kl = mean_flat(kl) / np.log(2.0) decoder_nll = self.token_discrete_loss(x_start, get_logits, input_ids) # t=-1 decoder_nll = decoder_nll / out["mean"].size(-1) decoder_nll = decoder_nll / np.log(2.0) mask_1 = t == 0 if mask_1.any(): kl_T = normal_kl( x_start_mean, x_start_log_var, out["mean"], out["log_variance"] ) kl_T = mean_flat(kl_T) / np.log(2.0) kl = torch.where(mask_1, kl_T, kl) out_mean, out_variance, out_log_variance_clipped = self.q_mean_variance( x_start, torch.LongTensor([self.num_timesteps - 1]).to(x_start.device) ) kl_T = normal_kl(out_mean, out_log_variance_clipped, 0, 0) kl_T = mean_flat(kl_T) / np.log(2.0) # print(decoder_nll, ) # print() # 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 =torch.where((t == 0), decoder_nll, kl) output = kl + decoder_nll + kl_T return { "output": output, "pred_xstart": out["pred_xstart"], "kl": kl, "decoder_nll": decoder_nll, "kl_T": kl_T, } def get_x_start(self, x_start_mean, std): """ Using the interpolating policy OR using the convolution policy... :param x_start_mean: :return: """ noise = torch.randn_like(x_start_mean) # print(std.shape, noise.shape, x_start_mean.shape) assert noise.shape == x_start_mean.shape # print(x_start_mean.device, noise.device) return x_start_mean + std * noise def token_discrete_loss(self, x_t, get_logits, input_ids): if self.model_arch == "conv-unet" or self.model_arch == "1d-unet": reshaped_x_t = x_t.view(x_t.size(0), x_t.size(1), -1).permute(0, 2, 1) else: # print(x_t.shape) reshaped_x_t = x_t # logits = get_logits(reshaped_x_t) # bsz, seqlen, vocab logits = get_logits(reshaped_x_t) loss_fct = torch.nn.CrossEntropyLoss(reduction="none") decoder_nll = loss_fct( logits.view(-1, logits.size(-1)), input_ids.view(-1) ).view(input_ids.shape) decoder_nll = decoder_nll.mean(dim=-1) return decoder_nll def x0_helper(self, model_output, x, t): if self.model_mean_type == ModelMeanType.PREVIOUS_X: pred_xstart = self._predict_xstart_from_xprev( x_t=x, t=t, xprev=model_output ) pred_prev = model_output elif self.model_mean_type in [ModelMeanType.START_X, ModelMeanType.EPSILON]: if self.model_mean_type == ModelMeanType.START_X: pred_xstart = model_output else: pred_xstart = self._predict_xstart_from_eps( x_t=x, t=t, eps=model_output ) pred_prev, _, _ = self.q_posterior_mean_variance( x_start=pred_xstart, x_t=x, t=t ) else: raise NotImplementedError(self.model_mean_type) return {"pred_xprev": pred_prev, "pred_xstart": pred_xstart} def training_losses_e2e(self, model, micro, t, noise=None): """ The function `training_losses_e2e` calculates various loss terms for an end-to-end training process in a machine learning model. :param model: The `model` parameter in the `training_losses_e2e` function seems to be an instance of a model used for training. It is likely a neural network model that is being trained for a specific task, such as sequence generation or prediction. The model is used within the function to make predictions :param micro: The `micro` parameter in the `training_losses_e2e` function seems to be a tuple containing the following elements: :param t: The `t` parameter in the `training_losses_e2e` function seems to represent the time step or timestep index. It is used to determine certain conditions within the function, such as comparing it to a threshold value of 400 and scaling timesteps. The function performs various calculations and computations based :param noise: The `noise` parameter in the `training_losses_e2e` function is used to pass a tensor representing random noise. If the `noise` parameter is not provided when calling the function, it generates random noise using `torch.randn_like(mix_start)`. This noise is then used in the :return: The function `training_losses_e2e` returns a dictionary `terms` containing different loss terms based on the specified loss type. The specific terms included in the dictionary depend on the conditions and calculations performed within the function for the given loss type. The function calculates and populates the `terms` dictionary with relevant loss values such as mean squared error (mse), variational bound (vb), decoder negative """ selfies_ids = micro[0] caption_state = micro[1] caption_mask = micro[2] corrupted_selfies_ids = micro[3] assert corrupted_selfies_ids.shape == selfies_ids.shape ######################################### mix_ids = torch.where( t.reshape(-1, 1) < 400, corrupted_selfies_ids, selfies_ids ) if t.max() > self.maxt: self.maxt = t.max() # print("Recieving max t:{}".format(self.maxt)) ########################################## # print(f"Model dir: {dir(model)}") try: x_start_mean = model.model.get_embeds(selfies_ids) mix_start_mean = model.model.get_embeds(mix_ids) except: x_start_mean = model.model.module.get_embeds(selfies_ids) mix_start_mean = model.model.module.get_embeds(mix_ids) std = _extract_into_tensor( self.sqrt_one_minus_alphas_cumprod, torch.tensor([0]).to(x_start_mean.device), x_start_mean.shape, ) x_start = self.get_x_start(x_start_mean, std) mix_start = self.get_x_start(mix_start_mean, std) if noise is None: noise = torch.randn_like(mix_start) x_t = self.q_sample(mix_start, t, noise=noise) # reparametrization trick. try: get_logits = model.model.get_logits except: get_logits = model.model.module.get_logits terms = {} if self.loss_type == LossType.E2E_KL: pass elif ( self.loss_type == LossType.E2E_MSE or self.loss_type == LossType.E2E_RESCALED_MSE ): model_output = model( x_t, self._scale_timesteps(t), caption_state, caption_mask ) if self.model_var_type in [ ModelVarType.LEARNED, ModelVarType.LEARNED_RANGE, ]: pass 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 ] # this is exactly x_start # print(model_output.shape ,target.shape , x_start.shape) assert model_output.shape == target.shape == x_start.shape terms["mse"] = mean_flat((target - model_output) ** 2) # print( terms["mse"]) model_out_x_start = self.x0_helper(model_output, x_t, t)[ "pred_xstart" ] # this is exactly model_output t0_mask = t == 0 t0_loss = mean_flat((x_start_mean - model_out_x_start) ** 2) # print(terms["mse"].shape, ) terms["mse"] = torch.where(t0_mask, t0_loss, terms["mse"]) # tT_mask = (t == self.num_timesteps - 1) out_mean, _, _ = self.q_mean_variance( x_start, torch.LongTensor([self.num_timesteps - 1]).to(x_start.device) ) tT_loss = mean_flat(out_mean**2) decoder_nll = self.token_discrete_loss(x_start, get_logits, selfies_ids) if "vb" in terms: terms["loss"] = terms["mse"] + terms["vb"] else: terms["loss"] = terms["mse"] + (decoder_nll + tT_loss) 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 = torch.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_e2e( self, model, x_start, clip_denoised=True, model_kwargs=None, denoised_fn=None ): device = x_start.device batch_size = x_start.shape[0] input_ids = model_kwargs.pop("input_ids").to(device) x_start_mean = model.get_embeds(input_ids) if self.model_arch == "conv-unet": seqlen = int(np.sqrt(input_ids.size(1))) x_start_mean = x_start_mean.view( x_start_mean.size(0), seqlen, seqlen, x_start_mean.size(-1) ).permute(0, 3, 1, 2) elif self.model_arch == "1d-unet": x_start_mean = x_start_mean.permute(0, 2, 1) std = _extract_into_tensor( self.sqrt_one_minus_alphas_cumprod, torch.tensor([0]).to(x_start_mean.device), x_start_mean.shape, ) x_start_log_var = 2 * torch.log(std) x_start = self.get_x_start(x_start_mean, std) get_logits = model.get_logits vb = [] xstart_mse = [] mse = [] for t in list(range(self.num_timesteps))[::-1]: t_batch = torch.tensor([t] * batch_size, device=device) noise = torch.randn_like(x_start) x_t = self.q_sample(x_start=x_start, t=t_batch, noise=noise) with torch.no_grad(): out = self._vb_terms_bpd_e2e( model, x_start=x_start, x_t=x_t, t=t_batch, input_ids=input_ids, get_logits=get_logits, x_start_mean=x_start_mean, x_start_log_var=x_start_log_var, clip_denoised=clip_denoised, model_kwargs=model_kwargs, noise=noise, denoised_fn=denoised_fn, ) if t == self.num_timesteps - 1: assert len(vb) == 0 vb.append(out["kl_T"]) vb.append(out["kl"]) 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.append(out["decoder_nll"]) vb = torch.stack(vb, dim=1) xstart_mse = torch.stack(xstart_mse, dim=1) mse = torch.stack(mse, dim=1) # prior_bpd = self._prior_bpd(x_start) prior_bpd = out["kl_T"] total_bpd = vb.sum(dim=1) return { "total_bpd": total_bpd, "prior_bpd": prior_bpd, "vb": vb, "xstart_mse": xstart_mse, "mse": mse, } def calc_bpd_loop_emb( self, model, x_start, clip_denoised=True, model_kwargs=None, denoised_fn=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 = torch.tensor([t] * batch_size, device=device) noise = torch.randn_like(x_start) # print(t) x_t = self.q_sample(x_start=x_start, t=t_batch, noise=noise) # Calculate VLB term at the current timestep with torch.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, noise=noise, denoised_fn=denoised_fn, ) 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"]) # # ## DEBUG # def is_very_close(a, b): # return (((a - b) ** 2).mean()) # x_start_cycle = self._predict_xstart_from_eps(x_t=x_t, t=t_batch, eps=noise) # gold_eps_cycle = self._predict_eps_from_xstart(x_t, t_batch, x_start_cycle) # print(((gold_eps_cycle-noise)**2).mean()) # print(is_very_close(out2['pred_xstart'],out["pred_xstart"]), 'first isclose --> check p_mean') # model.eval() # with torch.no_grad(): # direct_pred_eps = model(x_t, self._scale_timesteps(t_batch), **model_kwargs) # print(((direct_pred_eps - noise) ** 2).mean(), 'ans1', self.rescale_timesteps) # x_start_cycle_pred = self._predict_xstart_from_eps(x_t=x_t, t=t_batch, eps=direct_pred_eps) # model_kwargs['debug_x_t'] = x_t # model_kwargs['debug_t_batch'] = t_batch # model_kwargs['debug_direct_pred_eps'] = direct_pred_eps # model_kwargs['debug_x_start_cycle_pred'] = x_start_cycle_pred # out2 = self.p_mean_variance( # model, x_t, t_batch, clip_denoised=clip_denoised, model_kwargs=model_kwargs # ) # # print(((out["pred_xstart"] - x_start_cycle_pred) ** 2).mean(), 'if not align issue with vb_terms') # print(is_very_close(out2['pred_xstart'], x_start_cycle_pred), '2nd isclose --> check our flattened') # gold_eps_cycle_pred = self._predict_eps_from_xstart(x_t, t_batch, x_start_cycle_pred) # print(((eps - noise) ** 2).mean(), 'ans2', self._scale_timesteps) # print() # print(((gold_eps_cycle_pred - direct_pred_eps) ** 2).mean(), 'should be same, exactly same computation..') ## DEBUG mse.append(mean_flat((eps - noise) ** 2)) vb = torch.stack(vb, dim=1) xstart_mse = torch.stack(xstart_mse, dim=1) mse = torch.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 = torch.from_numpy(arr).to(device=timesteps.device)[timesteps].float() while len(res.shape) < len(broadcast_shape): res = res[..., None] return res.expand(broadcast_shape)