Set the CUDA devices for the workers. Args: num_workers (int): Number of workers. devices (str or None): Comma-separated stri

Removes segments. Positive values shave the first segments, negative shave the last segments.

r""" Returns the state of the ExponentialMovingAverage as a dict. This method is used by accelerate during checkpointing to save the e

r""" Combines a latent iamge img_vae of shape (B, C, H, W) and a CLIP-embedded image img_clip of shape (B, 1, clip_img_dim) into a sin

Common preparation logic shared across all look_ahead_explore methods. Returns: tuple: (vis_path, action_ids_list, batch

Accumulate possibly repeated keys into list.

An object that should act as the task root when randomly placing. Sometimes the task can involve a large appliance (e.g. dish washer)

r""" Simply iterates over the state dict and replaces the patterns in `mapping` with the corresponding values. Args: state_dict (`dic

r""" Offloads all models to CPU using accelerate, significantly reducing memory usage. When called, unet, text_encoder, vae and safety

r""" Offloads all models (`unet`, `text_encoder`, `vae`, and `safety checker` state dicts) to CPU using 🤗 Accelerate, significantly re

Return [min_id, max_id] for an event type.

Save a chat transcript in a clear, readable JSON file and return it. Parameters ---------- messages The list passed to ``ope

r""" Fuses the LoRA parameters into the original parameters of the corresponding blocks. <Tip warning={true}> This is an exp

Return final output of a batch, if any.

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

grid_size: int of the grid height and width return: pos_embed: [grid_size*grid_size, embed_dim] or [1+grid_size*grid_size, embed_dim] (w/ or

RoPE for video tokens with 3D structure. Args: embed_dim: (`int`): The embedding dimension size, corresponding to hidden_size_he

Convert a panorama image to perspective views. Args: per_hfov (float): Horizontal field of view for the perspective views. pa

Initialises the episode. Called each time the scene is reset. Here we usually define how the task changes across variations. Based on

Args: Compares the current bitsandbytes version to a given reference with an operation. operation (`str`): A string repre

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

r""" Convert a PIL image or a list of PIL images to NumPy arrays. Args: images (`PIL.Image.Image` or `List[PIL.Image.Imag

Splits 'input_dict' into smaller batches and puts them into input_queue. Assumes all lists in 'input_dict' have the same length and

Compute the mean and variance of the diffusion posterior: q(x_{t-1} | x_t, x_0)

Reads a 4-byte length from sock, then that many bytes, unpickles.

Reads one or more .mp4 videos from a list of file paths and separates them into frames. The function returns a numpy array with shape (B, num

What conditions need to be met for the task to be a success. Note: this replaces any previously registered conditions! :param condit

Updates paths inside resnets to the new naming scheme (local renaming)

Rotates the input tensor by performing a cyclic horizontal shift along its width dimension. Args: input_z (torch.Tensor): Input tenso