def testStateToOlabelUniqueSinglePath(self):
labels = [
[3, 4, 3],
[1, 0, 0],
]
num_labels = 8
# 3 frames, 2 batch, 8 states (4 label, 4 blank).
#
# There is only single valid path for each sequence because the frame
# lengths and the label lengths are the same.
states = [[[0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0],
[0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]],
[[0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0],
[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]],
[[0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0],
[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]]]
labels = ops.convert_to_tensor(labels)
states = math_ops.log(states)
olabel = ctc_ops._state_to_olabel_unique(
labels, num_labels, states, ctc_ops.ctc_unique_labels(labels))
olabel = math_ops.exp(olabel)
blank = olabel[:, :, 0]
self.assertAllClose(blank, [[0.0, 0.0], [0.0, 0.0], [0.0, 0.0]])
self.assertAllClose(olabel[:, :, 1:], [
[[0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0],
[1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]],
[[0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0],
[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]],
[[0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0],
[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]],
])
def testCtcLossDenseUniqueFastPathIsSameAsCtcLoss(self):
random_seed.set_random_seed(5)
batch_size = 8
num_labels = 6
label_length = 5
num_frames = 12
logits = random_ops.random_uniform(
[num_frames, batch_size, num_labels])
labels = random_ops.random_uniform([batch_size, label_length],
minval=1,
maxval=num_labels,
dtype=dtypes.int64)
label_lengths = random_ops.random_uniform([batch_size],
minval=2,
maxval=label_length,
dtype=dtypes.int64)
label_mask = array_ops.sequence_mask(label_lengths,
maxlen=label_length,
dtype=label_lengths.dtype)
labels *= label_mask
logit_lengths = [num_frames] * batch_size
ctc_loss = ctc_ops.ctc_loss_dense(
labels=labels,
logits=logits,
label_length=label_lengths,
logit_length=logit_lengths,
unique=ctc_ops.ctc_unique_labels(labels))
ctc_loss_grads = gradients_impl.gradients(ctc_loss, [logits])[0]
# Shift labels down by one (move blank from 0 to num_labels -1)
tf_ctc_loss_labels = math_ops.cast(labels, dtypes.int32) - 1
tf_nn_ctc_logits = array_ops.concat([
logits[:, :, 1:],
logits[:, :, 0:1],
],
axis=2)
tf_ctc_loss_labels = ctc_ops.dense_labels_to_sparse(
tf_ctc_loss_labels, label_lengths)
tf_nn_ctc_loss = ctc_ops.ctc_loss(labels=tf_ctc_loss_labels,
inputs=tf_nn_ctc_logits,
sequence_length=logit_lengths,
time_major=True)
tf_nn_ctc_grads = gradients_impl.gradients(tf_nn_ctc_loss, [logits])[0]
with self.cached_session() as sess:
for _ in range(32):
self.assertAllClose(*self.evaluate([ctc_loss, tf_nn_ctc_loss]))
self.assertAllClose(*self.evaluate(
[ctc_loss_grads, tf_nn_ctc_grads]),
rtol=2e-06,
atol=2e-06)
def testCtcLossDenseUniqueFastPathIsSameAsCtcLoss(self):
random_seed.set_random_seed(5)
batch_size = 8
num_labels = 6
label_length = 5
num_frames = 12
logits = random_ops.random_uniform([num_frames, batch_size, num_labels])
labels = random_ops.random_uniform(
[batch_size, label_length], minval=1, maxval=num_labels,
dtype=dtypes.int64)
label_lengths = random_ops.random_uniform(
[batch_size], minval=2, maxval=label_length, dtype=dtypes.int64)
label_mask = array_ops.sequence_mask(
label_lengths, maxlen=label_length, dtype=label_lengths.dtype)
labels *= label_mask
logit_lengths = [num_frames] * batch_size
ctc_loss = ctc_ops.ctc_loss_dense(
labels=labels,
logits=logits,
label_length=label_lengths,
logit_length=logit_lengths,
unique=ctc_ops.ctc_unique_labels(labels))
ctc_loss_grads = gradients_impl.gradients(ctc_loss, [logits])[0]
# Shift labels down by one (move blank from 0 to num_labels -1)
tf_ctc_loss_labels = math_ops.cast(labels, dtypes.int32) - 1
tf_nn_ctc_logits = array_ops.concat([
logits[:, :, 1:],
logits[:, :, 0:1],
], axis=2)
tf_ctc_loss_labels = ctc_ops.dense_labels_to_sparse(
tf_ctc_loss_labels, label_lengths)
tf_nn_ctc_loss = ctc_ops.ctc_loss(
labels=tf_ctc_loss_labels,
inputs=tf_nn_ctc_logits,
sequence_length=logit_lengths,
time_major=True)
tf_nn_ctc_grads = gradients_impl.gradients(tf_nn_ctc_loss, [logits])[0]
with self.cached_session() as sess:
for _ in range(32):
self.assertAllClose(*self.evaluate([ctc_loss, tf_nn_ctc_loss]))
self.assertAllClose(
*self.evaluate([ctc_loss_grads, tf_nn_ctc_grads]),
rtol=2e-06,
atol=2e-06)
def testUnique(self):
labels = [
[3, 4, 4, 3],
[1, 1, 1, 0],
]
unique, idx = ctc_ops.ctc_unique_labels(labels)
self.assertAllEqual([
[3, 4, 0, 0],
[1, 0, 0, 0],
], unique)
self.assertAllEqual([
[0, 1, 1, 0],
[0, 0, 0, 1],
], idx)
def testUnique(self):
labels = [
[3, 4, 4, 3],
[1, 1, 1, 0],
]
unique, idx = ctc_ops.ctc_unique_labels(labels)
self.assertAllEqual([
[3, 4, 0, 0],
[1, 0, 0, 0],
], unique)
self.assertAllEqual([
[0, 1, 1, 0],
[0, 0, 0, 1],
], idx)
def testStateToOlabelUnique(self):
labels = [
[3, 4, 3, 4],
[1, 1, 1, 0],
]
num_labels = 8
# 3 frames, 2 batch, 10 states (5 label, 5 blank).
states = [
[[0.11, 0.12, 0.13, 0.14, 0.15, 0.16, 0.17, 0.18, 0.19, 0.20],
[0.21, 0.22, 0.23, 0.24, 0.25, 0.26, 0.27, 0.28, 0.29, 0.30]],
[[1.1, 1.2, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9, 2.0],
[2.1, 2.2, 2.3, 2.4, 2.5, 2.6, 2.7, 2.8, 2.9, 3.0]],
[[11.0, 12.0, 13.0, 14.0, 15.0, 16.0, 17.0, 18.0, 19.0, 20.0],
[21.0, 22.0, 23.0, 24.0, 25.0, 26.0, 27.0, 28.0, 29.0, 30.0]],
]
labels = ops.convert_to_tensor(labels)
states = math_ops.log(states)
olabel = ctc_ops._state_to_olabel_unique(
labels, num_labels, states, ctc_ops.ctc_unique_labels(labels))
olabel = math_ops.exp(olabel)
blank = olabel[:, :, 0]
self.assertAllClose(blank, [[
0.16 + 0.17 + 0.18 + 0.19 + 0.20, 0.26 + 0.27 + 0.28 + 0.29 + 0.30
], [1.6 + 1.7 + 1.8 + 1.9 + 2.0, 2.6 + 2.7 + 2.8 + 2.9 + 3.0],
[
16.0 + 17.0 + 18.0 + 19.0 + 20.0,
26.0 + 27.0 + 28.0 + 29.0 + 30.0
]])
self.assertAllClose(olabel[:, :, 1:], [
[[0.0, 0.0, 0.12 + 0.14, 0.13 + 0.15, 0.0, 0.0, 0.0],
[0.22 + 0.23 + 0.24, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]],
[[0.0, 0.0, 1.2 + 1.4, 1.3 + 1.5, 0.0, 0.0, 0.0],
[2.2 + 2.3 + 2.4, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]],
[[0.0, 0.0, 12.0 + 14.0, 13.0 + 15.0, 0.0, 0.0, 0.0],
[22.0 + 23.0 + 24.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]],
])
def testStateToOlabelUnique(self):
labels = [
[3, 4, 3, 4],
[1, 1, 1, 0],
]
num_labels = 8
# 3 frames, 2 batch, 10 states (5 label, 5 blank).
states = [
[[0.11, 0.12, 0.13, 0.14, 0.15, 0.16, 0.17, 0.18, 0.19, 0.20],
[0.21, 0.22, 0.23, 0.24, 0.25, 0.26, 0.27, 0.28, 0.29, 0.30]],
[[1.1, 1.2, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9, 2.0],
[2.1, 2.2, 2.3, 2.4, 2.5, 2.6, 2.7, 2.8, 2.9, 3.0]],
[[11.0, 12.0, 13.0, 14.0, 15.0, 16.0, 17.0, 18.0, 19.0, 20.0],
[21.0, 22.0, 23.0, 24.0, 25.0, 26.0, 27.0, 28.0, 29.0, 30.0]],
]
labels = ops.convert_to_tensor(labels)
states = math_ops.log(states)
olabel = ctc_ops._state_to_olabel_unique(
labels, num_labels, states, ctc_ops.ctc_unique_labels(labels))
olabel = math_ops.exp(olabel)
blank = olabel[:, :, 0]
self.assertAllClose(blank, [
[0.16 + 0.17 + 0.18 + 0.19 + 0.20,
0.26 + 0.27 + 0.28 + 0.29 + 0.30],
[1.6 + 1.7 + 1.8 + 1.9 + 2.0,
2.6 + 2.7 + 2.8 + 2.9 + 3.0],
[16.0 + 17.0 + 18.0 + 19.0 + 20.0,
26.0 + 27.0 + 28.0 + 29.0 + 30.0]])
self.assertAllClose(olabel[:, :, 1:], [
[[0.0, 0.0, 0.12 + 0.14, 0.13 + 0.15, 0.0, 0.0, 0.0],
[0.22 + 0.23 + 0.24, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]],
[[0.0, 0.0, 1.2 + 1.4, 1.3 + 1.5, 0.0, 0.0, 0.0],
[2.2 + 2.3 + 2.4, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]],
[[0.0, 0.0, 12.0 + 14.0, 13.0 + 15.0, 0.0, 0.0, 0.0],
[22.0 + 23.0 + 24.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]],
])
def testCtcLossDenseUniqueFastPathWithBlankIndexIsSameAsCtcLoss(self):
random_seed.set_random_seed(5)
batch_size = 8
num_labels = 6
label_length = 5
num_frames = 12
logits = random_ops.random_uniform(
[num_frames, batch_size, num_labels])
labels = random_ops.random_uniform([batch_size, label_length],
minval=0,
maxval=num_labels - 1,
dtype=dtypes.int64)
label_lengths = random_ops.random_uniform([batch_size],
minval=2,
maxval=label_length,
dtype=dtypes.int64)
label_mask = array_ops.sequence_mask(label_lengths,
maxlen=label_length,
dtype=label_lengths.dtype)
labels *= label_mask
logit_lengths = [num_frames] * batch_size
tf_ctc_loss_labels = math_ops.cast(labels, dtypes.int32)
tf_ctc_loss_labels = ctc_ops.dense_labels_to_sparse(
tf_ctc_loss_labels, label_lengths)
tf_nn_ctc_loss = ctc_ops.ctc_loss(labels=tf_ctc_loss_labels,
inputs=logits,
sequence_length=logit_lengths,
time_major=True)
tf_nn_ctc_grads = gradients_impl.gradients(tf_nn_ctc_loss, [logits])[0]
# Shift the blank logits/labels to be somewhere in the middle.
blank_index = 2
shifted_logits = array_ops.concat([
logits[:, :, :blank_index],
logits[:, :, -1:],
logits[:, :, blank_index:-1],
],
axis=2)
shifted_labels = array_ops.where_v2(labels < blank_index, labels,
labels + 1)
ctc_loss = ctc_ops.ctc_loss_dense(
labels=shifted_labels,
logits=shifted_logits,
label_length=label_lengths,
logit_length=logit_lengths,
blank_index=blank_index,
unique=ctc_ops.ctc_unique_labels(shifted_labels))
ctc_loss_grads = gradients_impl.gradients(ctc_loss, [logits])[0]
with self.cached_session() as sess:
for _ in range(32):
self.assertAllClose(*self.evaluate([ctc_loss, tf_nn_ctc_loss]))
self.assertAllClose(*self.evaluate(
[ctc_loss_grads, tf_nn_ctc_grads]),
rtol=2e-06,
atol=2e-06)
