k2
Contents:
Installation
Core concepts in k2
Python tutorials
Python API reference
version
k2
add_epsilon_self_loops
arc_sort
cat
closure
compose
compose_arc_maps
connect
convert_dense_to_fsa_vec
create_fsa_vec
create_sparse
ctc_graph
ctc_loss
ctc_topo
determinize
do_rnnt_pruning
expand_ragged_attributes
get_aux_labels
get_best_matching_stats
get_lattice
get_rnnt_logprobs
get_rnnt_logprobs_joint
get_rnnt_logprobs_pruned
get_rnnt_logprobs_smoothed
get_rnnt_prune_ranges
get_rnnt_prune_ranges_deprecated
index_add
index_fsa
index_select
intersect
intersect_dense
intersect_dense_pruned
intersect_device
invert
is_rand_equivalent
joint_mutual_information_recursion
levenshtein_alignment
levenshtein_graph
linear_fsa
linear_fsa_with_self_loops
linear_fst
linear_fst_with_self_loops
mutual_information_recursion
mwer_loss
one_best_decoding
properties_to_str
prune_on_arc_post
pruned_ranges_to_lattice
random_fsa
random_fsa_vec
random_paths
remove_epsilon
remove_epsilon_and_add_self_loops
remove_epsilon_self_loops
replace_fsa
reverse
rnnt_loss
rnnt_loss_pruned
rnnt_loss_simple
rnnt_loss_smoothed
shortest_path
simple_ragged_index_select
swoosh_l
swoosh_l_forward
swoosh_l_forward_and_deriv
swoosh_r
swoosh_r_forward
swoosh_r_forward_and_deriv
to_dot
to_str
to_str_simple
to_tensor
top_sort
trivial_graph
union
CtcLoss
forward
DecodeStateInfo
DenseFsaVec
__init__
_from_dense_fsa_vec
to
device
duration
DeterminizeWeightPushingType
name
value
Fsa
__getattr__
__getitem__
__init__
__setattr__
__str__
_get_arc_post
_get_backward_scores
_get_entering_arcs
_get_forward_scores
_get_tot_scores
_invalidate_cache_
as_dict
convert_attr_to_ragged
draw
from_openfst
from_str
get_arc_post
get_backward_scores
get_filler
get_forward_scores
get_tot_scores
invert
invert_
rename_tensor_attribute
requires_grad_
set_scores_stochastic
to
device
grad
num_arcs
properties
properties_str
requires_grad
shape
MWERLoss
forward
Nbest
from_lattice
intersect
top_k
total_scores
OnlineDenseIntersecter
__init__
decode
num_streams
RaggedShape
__eq__
__getitem__
__init__
__ne__
__repr__
__str__
compose
get_layer
index
max_size
numel
regular_ragged_shape
remove_axis
row_ids
row_splits
to
tot_size
tot_sizes
device
dim0
num_axes
RaggedTensor
__eq__
__getitem__
__getstate__
__init__
__ne__
__repr__
__setstate__
__str__
add
arange
argmax
cat
clone
index
logsumexp
max
min
normalize
numel
pad
remove_axis
remove_values_eq
remove_values_leq
requires_grad_
sort_
sum
to
to_str_simple
tolist
tot_size
unique
device
dim0
dtype
grad
is_cuda
num_axes
requires_grad
shape
values
RnntDecodingConfig
__init__
beam
decoder_history_len
max_contexts
max_states
vocab_size
RnntDecodingStream
__init__
__str__
RnntDecodingStreams
__init__
__str__
advance
format_output
get_contexts
terminate_and_flush_to_streams
SymbolTable
add
from_file
from_str
get
merge
to_file
ids
symbols
k2.ragged
cat
create_ragged_shape2
create_ragged_tensor
index
index_and_sum
random_ragged_shape
regular_ragged_shape
RaggedShape
__eq__
__getitem__
__init__
__ne__
__repr__
__str__
compose
get_layer
index
max_size
numel
regular_ragged_shape
remove_axis
row_ids
row_splits
to
tot_size
tot_sizes
device
dim0
num_axes
RaggedTensor
__eq__
__getitem__
__getstate__
__init__
__ne__
__repr__
__setstate__
__str__
add
arange
argmax
cat
clone
index
logsumexp
max
min
normalize
numel
pad
remove_axis
remove_values_eq
remove_values_leq
requires_grad_
sort_
sum
to
to_str_simple
tolist
tot_size
unique
device
dim0
dtype
grad
is_cuda
num_axes
requires_grad
shape
values
k2
Python API reference
k2
Edit on GitHub
k2
add_epsilon_self_loops
arc_sort
cat
closure
compose
compose_arc_maps
connect
convert_dense_to_fsa_vec
create_fsa_vec
create_sparse
ctc_graph
ctc_loss
ctc_topo
determinize
do_rnnt_pruning
expand_ragged_attributes
get_aux_labels
get_best_matching_stats
get_lattice
get_rnnt_logprobs
get_rnnt_logprobs_joint
get_rnnt_logprobs_pruned
get_rnnt_logprobs_smoothed
get_rnnt_prune_ranges
get_rnnt_prune_ranges_deprecated
index_add
index_fsa
index_select
intersect
intersect_dense
intersect_dense_pruned
intersect_device
invert
is_rand_equivalent
joint_mutual_information_recursion
levenshtein_alignment
levenshtein_graph
linear_fsa
linear_fsa_with_self_loops
linear_fst
linear_fst_with_self_loops
mutual_information_recursion
mwer_loss
one_best_decoding
properties_to_str
prune_on_arc_post
pruned_ranges_to_lattice
random_fsa
random_fsa_vec
random_paths
remove_epsilon
remove_epsilon_and_add_self_loops
remove_epsilon_self_loops
replace_fsa
reverse
rnnt_loss
rnnt_loss_pruned
rnnt_loss_simple
rnnt_loss_smoothed
shortest_path
simple_ragged_index_select
swoosh_l
swoosh_l_forward
swoosh_l_forward_and_deriv
swoosh_r
swoosh_r_forward
swoosh_r_forward_and_deriv
to_dot
to_str
to_str_simple
to_tensor
top_sort
trivial_graph
union
CtcLoss
forward
DecodeStateInfo
DenseFsaVec
__init__
_from_dense_fsa_vec
to
device
duration
DeterminizeWeightPushingType
name
value
Fsa
__getattr__
__getitem__
__init__
__setattr__
__str__
_get_arc_post
_get_backward_scores
_get_entering_arcs
_get_forward_scores
_get_tot_scores
_invalidate_cache_
as_dict
convert_attr_to_ragged
draw
from_openfst
from_str
get_arc_post
get_backward_scores
get_filler
get_forward_scores
get_tot_scores
invert
invert_
rename_tensor_attribute
requires_grad_
set_scores_stochastic
to
device
grad
num_arcs
properties
properties_str
requires_grad
shape
MWERLoss
forward
Nbest
from_lattice
intersect
top_k
total_scores
OnlineDenseIntersecter
__init__
decode
num_streams
RaggedShape
__eq__
__getitem__
__init__
__ne__
__repr__
__str__
compose
get_layer
index
max_size
numel
regular_ragged_shape
remove_axis
row_ids
row_splits
to
tot_size
tot_sizes
device
dim0
num_axes
RaggedTensor
__eq__
__getitem__
__getstate__
__init__
__ne__
__repr__
__setstate__
__str__
add
arange
argmax
cat
clone
index
logsumexp
max
min
normalize
numel
pad
remove_axis
remove_values_eq
remove_values_leq
requires_grad_
sort_
sum
to
to_str_simple
tolist
tot_size
unique
device
dim0
dtype
grad
is_cuda
num_axes
requires_grad
shape
values
RnntDecodingConfig
__init__
beam
decoder_history_len
max_contexts
max_states
vocab_size
RnntDecodingStream
__init__
__str__
RnntDecodingStreams
__init__
__str__
advance
format_output
get_contexts
terminate_and_flush_to_streams
SymbolTable
add
from_file
from_str
get
merge
to_file
ids
symbols
k2.ragged
cat
create_ragged_shape2
create_ragged_tensor
index
index_and_sum
random_ragged_shape
regular_ragged_shape
RaggedShape
__eq__
__getitem__
__init__
__ne__
__repr__
__str__
compose
get_layer
index
max_size
numel
regular_ragged_shape
remove_axis
row_ids
row_splits
to
tot_size
tot_sizes
device
dim0
num_axes
RaggedTensor
__eq__
__getitem__
__getstate__
__init__
__ne__
__repr__
__setstate__
__str__
add
arange
argmax
cat
clone
index
logsumexp
max
min
normalize
numel
pad
remove_axis
remove_values_eq
remove_values_leq
requires_grad_
sort_
sum
to
to_str_simple
tolist
tot_size
unique
device
dim0
dtype
grad
is_cuda
num_axes
requires_grad
shape
values