Return a `VideoSeqPos` view for a single frame at absolute index `t_idx`. This is useful for streaming / KV-cache inference where the model i

Generate samples from the batch. Based on given batch, it will automatically determine whether to generate image or video samples.

Generate samples from the batch. Based on given batch, it will automatically determine whether to generate image or video samples.

Read bytes from a given ``filepath`` with 'rb' mode in range [offset, offset + size). Args: filepath (str): Path to read data.

embed_dim: output dimension for each position pos: a list of positions to be encoded: size (M,) out: (M, D)

embed_dim: output dimension for each position pos: a list of positions to be encoded: size (M,) out: (M, D)

Returns data type choices for backward pass according to arch tag (attention.utils.get_arch_tag). Parameters: arch_tag (int): Arch t

Returns CPU time measurement in string format.

Returns CUDA time measurement in string format.

Get the data corresponding to the modality for a trajectory by a base index. This method will call the corresponding helper method based on th

Returns the compute capability of a given torch device if it's a CUDA device, otherwise returns 0. Args: device: torch device.

r"""This method map time from 0,1 to discrete steps

Returns data type choices according to arch tag (attention.utils.get_arch_tag). Parameters: arch_tag (int): Arch tag for the current

Get the chunk index for an episode index.

Get frames from a video at specified timestamps. Args: video_path (str): Path to the video file. timestamps (list[int] | np.ndarra

Returns data type choices for forward pass according to arch tag (attention.utils.get_arch_tag). Parameters: arch_tag (int): Arch ta

Returns a singleton instance of DeploymentEnv. The instance is created on first call and cached for subsequent calls. This e

r""" This method computes interpolation `X_t` and their time derivatives `dotX_t` at the specified time points `t`. Note that `x_0` is

Gets the maximum difference and its location in the tensor. Args: differences: Tensor containing differences. flatten: Whether to

Get the results of previously submitted checkpoints and pass them to callbacks if checkpoint succeeded

Get the rank (GPU device) of the worker. Returns: rank (int): The rank of the worker.

Get the rank (GPU device) of the worker. Returns: rank (int): The rank of the worker.

Get the dimension of a state or action key from the dataset metadata.

Get the RAW data for a single step in a trajectory. No transforms are applied. Args: trajectory_id (int): The name of the traject

Read data from a given ``filepath`` with 'r' mode. Args: filepath (str or Path): Path to read data. encoding (str): T

Get the data for a trajectory.

Generates a callable function `velocity_fn` based on the provided data batch and guidance factor. This function first processes the

Get a list of filepaths for all videos in the input directory.

Infer a suitable file client based on the URI and arguments. Args: file_client_args (dict, optional): Arguments to instantiate a

Run action-conditioned video generation inference using resolved setup and per-run arguments.

Creates the optimizer and scheduler for the model. Args: config_model (ModelConfig): The config object for the model. Re

Initialize and validate S3 path.

Inverse of quaternion(s). Equivalent to conjugate of quaternion. For unit quaternions, the inverse equals the conjugate. For non-unit quatern

Input validation function a specified backend. Runs the common and backend-specific checks. Returns False if any checks fail, otherwise True.

Input validation function a specified multi-dimensional backend. Runs the common and backend-specific checks. Returns False if any checks fai

Check if the video frames are safe.

Check if verbose mode is enabled via VERBOSE environment variable.

Check whether a file path is a directory. Args: filepath (str or Path): Path to be checked whether it is a direct

Check whether a file path is a file. Args: filepath (str or Path): Path to be checked whether it is a file. Returns:

Convert sRGB values to physically linear ones. The transformation is uniform in RGB, so *srgb* can be of any shape. *srgb* values should rang

Scan a directory to find the interested directories or files in arbitrary order. Note: Most object stores have no concept

For loading pretrained teacher net weights

Load experiment configuration from YAML file and merge with default Python configuration. Args: experiment_name: Name of the exp

Locate and return an object ``x`` using an input string ``{x.__module__}.{x.__qualname__}``, such as "module.submodule.class_name". Rais

Log a video tensor in shape [B, C, T, H, W] with values in [-1, 1].

Wrap attention with KV cache support and initialize caches. Each block's self-attention op is wrapped by `AttentionOpWithKVCache`, which adds

This is trivial in the base class, but may be used by derived classes in a more interesting way

Mean normalize a tensor by subtracting the mean and dividing by the standard deviation. Args: tensor (torch.tensor): The ten