stacks ConvDropoutNormLReLU layers. initial_stride will only be applied to first layer in the stack. The other parameters affect all layers

basically more flexible than v1, architecture is the same Does this look complicated? Nah bro. Functionality > usability Th

Create a wrapper class for the forward SDE (VP type). *** Update: We support discrete-time diffusion models by implementing a picewise

Construct a DPM-Solver. We support both DPM-Solver (`algorithm_type="dpmsolver"`) and DPM-Solver++ (`algorithm_type="dpmsolver++"`).

directory is expected to contain some folder structure: if some subfolder contains only files, all of these

Compute the noised input xt = alpha_t * x + sigma_t * noise. Args: x: A `torch.Tensor` with shape `(batch_size, *shape)`

Compute the entire variational lower-bound, measured in bits-per-dim, as well as other related quantities. :param model: the

This only applies for num_conv_per_stage and convolutional_upsampling=True not real vram consumption. just a constant term to which t

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

Convert primitive modules to float16.

Convert primitive modules to float32, undoing convert_module_to_f16().

Convert the torso of the model to float16.

Convert the torso of the model to float16.

Convert the torso of the model to float32.

Convert the torso of the model to float32.

Convert the torso of the model to float32.

Sample x_{t+1} from the model using DDIM reverse ODE.

Generate samples from the model using DDIM. Same usage as p_sample_loop().

Get the device to use for torch.distributed.

Write all of the diagnostics from the current iteration

The dynamic thresholding method.

Apply the module to `x` given `emb` timestep embeddings.

Apply the block to a Tensor, conditioned on a timestep embedding. :param x: an [N x C x ...] Tensor of features. :param emb:

Apply QKV attention. :param qkv: an [N x (H * 3 * C) x T] tensor of Qs, Ks, and Vs. :return: an [N x (H * C) x T] tensor aft

Apply QKV attention. :param qkv: an [N x (3 * H * C) x T] tensor of Qs, Ks, and Vs. :return: an [N x (H * C) x T] tensor aft

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.

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.

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.

Get directory that log files are being written to. will be None if there is no output directory (i.e., if you didn't call start)

Get a pre-defined beta schedule for the given name. The beta schedule library consists of beta schedules which remain similar in the limi

Inverse the sample `x` from time `t_start` to `t_end` by DPM-Solver. For discrete-time DPMs, we use `t_start=1/N`, where `N` is the t

Load a PyTorch file without redundant fetches across MPI ranks.

The same as logkv(), but if called many times, values averaged.