r""" Outputs: `Tuple` comprising various elements depending on the configuration (config) and inputs: **last_hidden_state**: ``torch.Float

Subclass torch's LayerNorm to handle fp16.

r"""A placeholder identity operator that is argument-insensitive.

r""" :class:`~pytorch_transformers.BertConfig` is the configuration class to store the configuration of a `BertModel`. Argum

Construct the embeddings from word, position and token_type embeddings.

Expand from BertModel to handle image region features as input

r""" Takes in observations and produces an embedding of the rgb component. Args: observation_space: The observation_space of the

Modified from BertAttention to add support for output_hidden_states.

Modified from BertEncoder to add support for output_hidden_states.

Modified from BertLayer to add support for output_hidden_states.

Modified from BertSelfAttention to add support for output_hidden_states.

Just to compute depth equirect with height same as equ_h

Compute SQuAD 2.0 answer class from classification and start tokens hidden states.

Compute SQuAD end_logits from sequence hidden states and start token hidden state.

Compute SQuAD start_logits from sequence hidden states.

r"""Utility class for extracting the action on the shortest path to the goal. Args: sim: HabitatSim instance. goal_radius:

Modified from BertForMultipleChoice to support oscar training.

r"""A base trainer for VLN-CE imitation learning.

Construct the embeddings from word, position and token_type embeddings.

r""" **masked_lm_labels**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size, sequence_length)``: Labels for computing the

r""" Inputs: **input_ids**: ``torch.LongTensor`` of shape ``(batch_size, num_choices, sequence_length)``: Indices of input seq

r""" **next_sentence_label**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size,)``: Labels for computing the next sequence

r""" **masked_lm_labels**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size, sequence_length)``: Labels for computing the

r""" **start_positions**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size,)``: Labels for position (index) of the start o

r""" **labels**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size,)``: Labels for computing the sequence classification/re

r""" **labels**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size, sequence_length)``: Labels for computing the token clas

An abstract class to handle weights initialization and a simple interface for dowloading and loading pretrained models.

An observation transformer that center crops your input on a per-sensor basis.

r"""This is an experimental use of ObservationTransformer that converts a cubemap output to an equirectangular one through projection. This needs

r"""The agents current location in the global coordinate frame Args: sim: reference to the simulator for calculating task observations.

r"""NDTW (Normalized Dynamic Time Warping) ref: Effective and General Evaluation for Instruction Conditioned Navigation using Dynamic Tim