Assumes only one parameters object in the Spatial Transform

sets params to be self.params + grad_data

Assumes only one parameters object in the Spatial Transform

Builds some adversarial examples given real inputs and labels ARGS: inputs : torch.Tensor (NxCxHxW) - tensor with examples needed

Create a 1D, 2D, or 3D average pooling module.

Create a beta schedule that discretizes the given alpha_t_bar function, which defines the cumulative product of (1-beta) over time from t = [

Defaults for classifier models.

Clips the parameters to be between 0 and 1 and also within the color space's gamut.

Clips the parameters to be between -1 and 1 as required for grid_sample

Returns a list of [adversarials, originals] of the SUCCESSFUL attacks only, according to the given classifier_net, normalizer

Compute the mean for the previous step, given a function cond_fn that computes the gradient of a conditional log probability with res

Compute what the p_mean_variance output would have been, should the model's score function be conditioned by cond_fn. See co

tensordot(x, y, 1).

Fused `tf.nn.conv2d()` followed by `downsample_2d()`. Padding is performed only once at the beginning, not between the operations. The fused

Returns the number of data points in this logger instance

Sample x_{t-1} from the model using DDIM. Same usage as p_sample().

Use DDIM to sample from the model and yield intermediate samples from each timestep of DDIM. Same usage as p_sample_loop_pro

Discretize the SDE in the form: x_{i+1} = x_i + f_i(x_i) + G_i z_i. Useful for reverse diffusion sampling and probabiliy flow sampling. Defau

Discretizes an image tensor into a tensor filled with ints ranging between 0 and 255 ARGS: img_tensor : floatTensor (NxCxHxW) - t

Compute the log-likelihood of a Gaussian distribution discretizing to a given image. :param x: the target images. It is assumed that thi

Get the drift function of the reverse-time SDE.

Outputs the accuracy of the adversarial examples NOTE: notice the difference between N and N' in the argument ARGS:

Normalizes var by subtracting the mean of each channel and then dividing each channel by standard dev ARGS: self - st

Form a torch tensor with the given `shape` from a flattened numpy array `x`.

Legacy code to reproduce previous results on DDPM. Not recommended for new work.

Create a function to give the output of the score-based model. Args: model: The score model. train: `True` for training and `False` f

Probability flow ODE sampler with the black-box ODE solver. Args: sde: An `sde_lib.SDE` object that represents the forward SDE. shape: A se

Create a Predictor-Corrector (PC) sampler. Args: sde: An `sde_lib.SDE` object representing the forward SDE. shape: A sequence of integers.

Create a loss function for training with arbirary SDEs. Args: sde: An `sde_lib.SDE` object that represents the forward SDE. train: `True` f

Legacy code to reproduce previous results on SMLD(NCSN). Not recommended for new work.

This matches the implementation in Denoising Diffusion Probabilistic Models: From Fairseq. Build sinusoidal embeddings. This matches

Sample from p(x_{t-1} | x_t)

You can create this dataloader using: train_data = imagenet_lmdb_dataset(traindir, transform=train_transform) valid_data = imagenet_lmdb_

Skips the whole parameter argument thing and just loads the whole state dict from a filename. ARGS: filename : string - filename

Logs the full data point ARGS: name: string - name of existing series in self.series data_tuple : tuple of form ((epo

Log a value of some diagnostic Call this once for each diagnostic quantity, each iteration If called many times, last value will be used.

Copy the master parameter data back into the model parameters.

Takes in an other instance of this same class with the same shape of parameters (NxSHAPE) and a self_mask bytetensor of length

Takes in an other instance of this same class with the same shape of parameters (NxSHAPE) and a self_mask bytetensor of length

Copy the gradients from the model parameters into the master parameters from make_master_params().

Copied from: https://github.com/openai/baselines/blob/ea25b9e8b234e6ee1bca43083f8f3cf974143998/baselines/common/mpi_util.py#L110 Perform a we

Create the drift and diffusion functions for the reverse SDE

Optimizes with warmup and gradient clipping (disabled if negative).

Generate samples from the model and yield intermediate samples from each timestep of diffusion. Arguments are the same as p_

Parse filenames of the form path/to/modelNNNNNN.pt, where NNNNNN is the checkpoint's number of steps.