Soft Dot Attention. Ref: http://www.aclweb.org/anthology/D15-1166 Adapted from PyTorch OPEN NMT.

A simple wrapper for a batch of MatterSim environments, using discretized viewpoints and pretrained features

Results submission format: [{'instr_id': string, 'trajectory':[(viewpoint_id, heading_rads, elevation_rads),] } ]

Implements the Room to Room navigation task, using discretized viewpoints and pretrained features

Class to tokenize and encode a sentence.

An agent based on an LSTM seq2seq model with attention.

An unrolled LSTM with attention over instructions for decoding navigation actions.

An unrolled LSTM with attention over instructions for decoding navigation actions.

An unrolled LSTM with attention over instructions for decoding navigation actions.

Base class for an R2R agent to generate and save trajectories.

Encodes navigation instructions, returning hidden state context (for attention methods) and a decoder initial state.