Utilities for training and sampling diffusion models. Ported directly from here, and then adapted over time to further experimentation.

Compose a data pipeline with a sequence of transforms. Args: transforms (list[dict | callable]): Either config dicts of trans

A wrapper of concatenated dataset. Same as :obj:`torch.utils.data.dataset.ConcatDataset`, but add `get_cat_ids` function. Args: da

A DistributedDataParallel wrapper for models in 3D mesh estimation task. In 3D mesh estimation task, there is a need to wrap different modules i

A wrapper of repeated dataset. The length of repeated dataset will be `times` larger than the original dataset. This is useful when the data l

A diffusion process which can skip steps in a base diffusion process. :param use_timesteps: a collection (sequence or set) of timesteps from

Base class for mogen architecture.

Base motion dataset. Args: data_prefix (str): the prefix of data path. pipeline (list): a list of dict, where each element represe

Collect data from the loader relevant to the specific task. This is usually the last stage of the data loader pipeline. Args: keys (

r"""Crop motion sequences. Args: crop_size (int): The size of the cropped motion sequence.

State of file existence.

Define GAN loss. Args: gan_type (str): Support 'vanilla', 'lsgan', 'wgan', 'hinge'. real_label_val (float): The value for real la

MSELoss. Args: reduction (str, optional): The method that reduces the loss to a scalar. Options are "none", "mean" and "sum".

Which type of output the model predicts.

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_

Normalize motion sequences. Args: mean_path (str): Path of mean file. std_path (str): Path of std file.

r"""Process siamese motion sequences. The code is borrowed from https://github.com/tr3e/InterGen/blob/master/utils/utils.py

r"""Random crop motion sequences. Each sequence will be padded with zeros to the maximum length. Args: min_size (int or None): The mi

A distribution over timesteps in the diffusion process, intended to reduce variance of the objective. By default, samplers perform unbias

r"""Swap motion sequences. Args: prob (float): The probability of swapping siamese motions

TextMotion dataset. Args: text_dir (str): Path to the directory containing the text files.

Wrap fields of the data dictionary into lists for evaluation. This class can be used as a last step of a test or validation pipeline for sing