athena.metrics

some metrics

Module Contents

Classes

Accuracy	Accuracy
CharactorAccuracy	CharactorAccuracy
Seq2SeqSparseCategoricalAccuracy	Seq2SeqSparseCategoricalAccuracy
CTCAccuracy	CTCAccuracy
ClassificationAccuracy	ClassificationAccuracy
EqualErrorRate	EqualErrorRate

class athena.metrics.Accuracy(name='Accuracy', rank_size=1)

Accuracy Base class for Accuracy calculation

reset_states(self)

reset num_err and num_total to zero

update_state(self, predictions, samples, logit_length=None)

Accumulate errors and counts

__call__(self, logits, samples, logit_length=None)

result(self)

returns word-error-rate calculated as num_err/num_total

class athena.metrics.CharactorAccuracy(name='CharactorAccuracy', rank_size=1)

Bases: athena.metrics.Accuracy

CharactorAccuracy Base class for Word Error Rate calculation

update_state(self, predictions, samples, logit_length=None)

Accumulate errors and counts

class athena.metrics.Seq2SeqSparseCategoricalAccuracy(eos, name='Seq2SeqSparseCategoricalAccuracy')

Bases: athena.metrics.CharactorAccuracy

Seq2SeqSparseCategoricalAccuracy Inherits CharactorAccuracy and implements Attention accuracy calculation

__call__(self, logits, samples, logit_length=None)

Accumulate errors and counts

class athena.metrics.CTCAccuracy(name='CTCAccuracy')

Bases: athena.metrics.CharactorAccuracy

CTCAccuracy Inherits CharactorAccuracy and implements CTC accuracy calculation

__call__(self, logits, samples, logit_length=None)

Accumulate errors and counts, logit_length is the output length of encoder

class athena.metrics.ClassificationAccuracy(name='ClassificationAccuracy', rank_size=1)

Bases: athena.metrics.Accuracy

ClassificationAccuracy Implements top-1 accuracy calculation for speaker classification (closed-set speaker recognition)

update_state(self, predictions, samples, logit_length=None)

Accumulate errors and counts

class athena.metrics.EqualErrorRate(name='EqualErrorRate')

EqualErrorRate Implements Equal Error Rate (EER) calculation for speaker verification (open-set speaker recognition)

reset_states(self)

reset predictions and labels

update_state(self, predictions, samples, logit_length=None)

append new predictions and labels

__call__(self, logits, samples, logit_length=None)

result(self)

calculate equal error rate