Args: level(str): log level string of loguru. Default value: "INFO". caller_names(tuple): caller names of redirected

Abstract model

Abstract model

transformer agent

transformer encoder for language inputs

transformer encoder for language, frames and action inputs

adjust optimizer learning rate w.r.t the schedulers

Run all_gather on arbitrary picklable data (not necessarily tensors) Args: data: any picklable object Returns: list[data]

compress mask array

loss function for Seq2Seq agent

compute f1 and extract match scores for agent output

compute exact matching and f1 score for action predictions

compute precision of predictions for interaction object classes

#Takes in the response object returned by action inference service and converts object class names to object IDs using the metadata #instr (N

create a scheduler for the learning rate

output processing for a VLN agent

Compress features by channel average pooling