Serializes this instance to a JSON string.

angle: 0 ~ 360 degree

angle: 0 ~ 360 degree

angle: 0 ~ 360 degree

Construct a layernorm module in the TF style (epsilon inside the square root).

Build a resized Embedding Module from a provided token Embedding Module. Increasing the size will add newly initialized vectors at the en

Loads word embeddings from a pretrained embeddings file. PAD: index 0. [0.0, ... 0.0] UNK: index 1. mean of all R2R word embeddings: [

Episode length is currently not considered. We were previously batching episodes together with similar lengths. Not sure if we need t

Prunes heads of the model. heads_to_prune: dict of {layer_num: list of heads to prune in this layer} See base class PreTraine

r"""Helper method to get the scene name from an episode. Assumes the scene_id is formated /path/to/<scene_name>.<ext>

r'''get resulting distance to goal by executing a candidate action

r"""Uses the ground truth trajectories to create a teacher forcing datset for a given split. Loads both guide and follower episodes.

Instantiate NeRF's MLP model.

Constructs a `Config` from a Python dictionary of parameters.

Constructs a `BertConfig` from a json file of parameters.

Given a URL, look for the corresponding dataset in the local cache. If it's not there, download it. Then return the path to the cached file.

Determine the closest MP3D node to the agent's start position as given by a [x,z] position vector. Returns: node ID

Returns the next action along the shortest path.

Generate masks centered at mu of the given x and y range with the origin in the centre of the output Inputs: mu: tensor (N, 2)

Input (batch_size, 1, height, width)

Generate image of single frame from observation and info returned from a single environment step(). Args: observation: observation re

Takes a batch of waypoint predictions and converts them to global 2D Cartesian coordinates. `current_position` and `current_heading` are in the

Prune a Conv1D layer (a model parameters) to keep only entries in index. A Conv1D work as a Linear layer (see e.g. BERT) but the weights are

Transforms model's predictions to semantically meaningful values. Args: raw: [num_rays, num_samples along ray, 4]. Prediction from model

r"""Runs experiment given mode and config Args: exp_config: path to config file. run_type: "train" or "eval. opts: list o