Computes recall at 1, 5, and 10 given a similarity matrix S. By convention, rows of S are assumed to correspond to images and columns are cap

Assumes image_outputs is a (batchsize, embedding_dim, rows, height) tensor Assumes audio_outputs is a (batchsize, embedding_dim, 1, time) ten

Assumes image_outputs is a (batchsize, embedding_dim, rows, height) tensor Assumes audio_outputs is a (batchsize, embedding_dim, 1, time) ten

List msg for verbose function

List msg for verbose function

Sinusoid position encoding table

Sinusoid position encoding table

Sinusoid position encoding table

load progress pkl file Args: prog_pkl(str): path to progress pkl file Return: progress(list): epoch(int):

perform preemphasis on the input signal. :param signal: The signal to filter. :param coeff: The preemphasis coefficient. 0 is none, default 0.

Computes the triplet margin ranking loss for each anchor image/caption pair The impostor image/caption is randomly sampled from the minibatch

Input: Similarity matrix S as an autograd.Variable Output: The one-way triplet loss from rows of S to columns of S. Imposters are randoml

Input: Similarity matrix S as an autograd.Variable Output: The one-way triplet loss from rows of S to columns of S. Impostors are taken t