A dummy context manager - useful for having conditional scopes (such as @maybe_no_grad). Nothing happens in this scope.

Vectorized environment that runs multiple environments in parallel. It uses `multiprocessing` processes, and pipes for communication. Paramete

JPEG 2000 codec for numcodecs.

Module that can generate observation outputs by modality. Inputs are assumed to be flat (usually outputs from some hidden layer). Each observ

To ensure sending command to the robot with predictable latency this controller need its separate process (due to python GIL)

Conv1d --> GroupNorm --> Mish

This class allows networks to encode multiple observation dictionaries into a single flat, concatenated vector representation. It does this b

Pre-trained ResNetV1b Model, which produces the strides of 8 featuremaps at conv5. Parameters ---------- block : Block Class for

Class that wraps another valid torch distribution, such that sampled values from the base distribution are passed through a tanh layer. The c

To ensure sending command to the robot with predictable latency this controller need its separate process (due to python GIL)

To ensure sending command to the robot with predictable latency this controller need its separate process (due to python GIL)

Focal Loss, as described in https://arxiv.org/abs/1708.02002. It is essentially an enhancement to cross entropy loss and is useful for classif

Extension to MLP to accept multiple observation dictionaries as input and to output dictionaries of tensors. Inputs are specified as a dictio

Base class for simple Multi-Layer Perceptrons.

Zarr-based temporal datastructure. Assumes first dimension to be time. Only chunk in time dimension.

Works for DAVIS/YouTubeVOS/BL30K training For each sequence: - Pick three frames - Pick two objects - Apply some random transform

This class is used to read a video, one frame at a time

Applies a 1D convolution over an input signal composed of several input planes, optionally followed by batch normalization and ReLU activation.

Applies a 2D convolution (optionally with batch normalization and relu activation) over an input signal composed of several input planes.

Module wrapper that returns intermediate layers from a model It has a strong assumption that the modules have been registered into the m

Works for key/value pairs type storage e.g., working and long-term memory

Applies a linear transformation to the incoming data optionally followed by batch normalization and relu activation

This class is used to convert a indexed-mask to a one-hot representation. It also takes care of remapping non-continuous indices It has t

Module that processes inputs by observation key and then concatenates the processed observation keys together. Each key is processed with an

Custom dictionary class with the sole additional purpose of automatically registering new "keys" at runtime without breaking. This is mainly

A class that is a thin wrapper around a torch.nn.Parameter to make for easy saving and optimization.

Wraps @Algo object to make it easy to run policies in a rollout loop.

Compose multiple Modules together (defined above).

Track a series of values and provide access to smoothed values over a window or the global series average.