kaldi.base

Functions

get_verbose_level	Returns the verbose level.
set_abort_on_assert_failure	Enables/disables the call to abort in case of assertion failure.
set_print_stack_trace_on_error	Enables/disables the printing of stack trace in case of an error.
set_verbose_level	Sets the verbose level.

Classes

Timer

Kaldi timer.

class kaldi.base.Timer

Kaldi timer.

elapsed() → float

Returns elapsed time since timer was reset.

new(set_timer:bool) → Timer

Creates and optionally resets the timer to current time.

reset()

Resets the timer to current time.

kaldi.base.get_verbose_level() → int

Returns the verbose level.

kaldi.base.set_abort_on_assert_failure(a:bool)

Enables/disables the call to abort in case of assertion failure.

kaldi.base.set_print_stack_trace_on_error(p:bool)

Enables/disables the printing of stack trace in case of an error.

kaldi.base.set_verbose_level(i:int)

Sets the verbose level.

kaldi.base.io

Functions

expect_token	Tries to read the given token and throws exception on failure.
flush	Flushes output stream.
get_stderr	Returns standard error stream.
get_stdin	Returns standard input stream.
get_stdout	Returns standard output stream.
init_kaldi_input_stream	Initializes an input stream by detecting the binary header.
init_kaldi_output_stream	Initializes an output stream by writing an optional binary header.
peek	Consumes whitespace (if binary == True) and returns the peek value.
peek_token	Returns the first character of the next token or -1 if EOF.
read	Reads and returns the contents of the stream as a bytes object.
read_bool	Reads a Boolean.
read_double	Reads a double precision float.
read_float	Reads a single precision float.
read_int	Reads an integer.
read_int_pair_vector	Reads a sequence of integer pairs.
read_int_vector	Reads a sequence of integers.
read_text	Reads and returns the contents of the stream as a unicode string.
read_token	Reads and returns the next token (throws exception on failure)
readline	Reads and returns a line from the input stream.
readline_text	Reads and returns a line from the input stream.
write	Writes bytes object to output stream.
write_bool	Writes a Boolean.
write_double	Writes a double precision float.
write_float	Writes a single precision float.
write_int	Writes an integer.
write_int_pair_vector	Writes a sequence of integer pairs.
write_int_vector	Writes a sequence of integers.
write_text	Writes unicode string to output stream.
write_token	Writes non-empty string of non-space characters.

Classes

ifstream	Input file stream.
iostream	Input/output stream.
istream	Input stream.
istringstream	Input string stream.
ofstream	Output file stream.
ostream	Output stream.
ostringstream	Output string stream.
stringstream	Input/output string stream.

kaldi.base.io.expect_token(is:istream, binary:bool, token:str)

Tries to read the given token and throws exception on failure.

kaldi.base.io.flush(os:ostream)

Flushes output stream.

kaldi.base.io.get_stderr() → ostream

Returns standard error stream.

kaldi.base.io.get_stdin() → istream

Returns standard input stream.

kaldi.base.io.get_stdout() → ostream

Returns standard output stream.

class kaldi.base.io.ifstream

Input file stream.

from_file(filename:str) → ifstream

Creates a new input file stream.

good() → bool

Checks whether the state of the stream is good.

kaldi.base.io.init_kaldi_input_stream(is:istream) -> (success:bool, binary:bool)

Initializes an input stream by detecting the binary header.

kaldi.base.io.init_kaldi_output_stream(os:ostream, binary:bool)

Initializes an output stream by writing an optional binary header.

class kaldi.base.io.iostream

Input/output stream.

good() → bool

Checks whether the state of the stream is good.

class kaldi.base.io.istream

Input stream.

good() → bool

Checks whether the state of the stream is good.

class kaldi.base.io.istringstream

Input string stream.

from_str(str:str) → istringstream

Creates a new input string stream.

good() → bool

Checks whether the state of the stream is good.

to_bytes() → bytes

Returns stream contents as a bytes object.

to_str() → str

Returns stream contents as a unicode string.

class kaldi.base.io.ofstream

Output file stream.

from_file(filename:str) → ofstream

Creates a new output file stream.

good() → bool

Checks whether the state of the stream is good.

class kaldi.base.io.ostream

Output stream.

good() → bool

Checks whether the state of the stream is good.

class kaldi.base.io.ostringstream

Output string stream.

from_str(str:str) → ostringstream

Creates a new output string stream.

good() → bool

Checks whether the state of the stream is good.

to_bytes() → bytes

Returns stream contents as a bytes object.

to_str() → str

Returns stream contents as a unicode string.

kaldi.base.io.peek(is:istream, binary:bool) → int

Consumes whitespace (if binary == True) and returns the peek value.

kaldi.base.io.peek_token(is:istream, binary:bool) → int

Returns the first character of the next token or -1 if EOF.

kaldi.base.io.read(is:istream) → bytes

Reads and returns the contents of the stream as a bytes object.

kaldi.base.io.read_bool(is:istream, binary:bool) → bool

Reads a Boolean.

kaldi.base.io.read_double(is:istream, binary:bool) → float

Reads a double precision float.

kaldi.base.io.read_float(is:istream, binary:bool) → float

Reads a single precision float.

kaldi.base.io.read_int(is:istream, binary:bool) → int

Reads an integer.

kaldi.base.io.read_int_pair_vector(is:istream, binary:bool) → list<tuple<int, int>>

Reads a sequence of integer pairs.

kaldi.base.io.read_int_vector(is:istream, binary:bool) → list<int>

Reads a sequence of integers.

kaldi.base.io.read_text(is:istream) → str

Reads and returns the contents of the stream as a unicode string.

kaldi.base.io.read_token(is:istream, binary:bool) → str

Reads and returns the next token (throws exception on failure)

kaldi.base.io.readline(is:istream) → bytes

Reads and returns a line from the input stream.

If the stream is already at EOF, an empty bytes object is returned.

kaldi.base.io.readline_text(is:istream) → str

Reads and returns a line from the input stream.

If the stream is already at EOF, an empty unicode string is returned.

class kaldi.base.io.stringstream

Input/output string stream.

from_str(str:str) → stringstream

Creates a new input/output string stream.

good() → bool

Checks whether the state of the stream is good.

to_bytes() → bytes

Returns stream contents as a bytes object.

to_str() → str

Returns stream contents as a unicode string.

kaldi.base.io.write(os:ostream, b:bytes)

Writes bytes object to output stream.

kaldi.base.io.write_bool(os:ostream, binary:bool, t:bool)

Writes a Boolean.

kaldi.base.io.write_double(os:ostream, binary:bool, t:float)

Writes a double precision float.

kaldi.base.io.write_float(os:ostream, binary:bool, t:float)

Writes a single precision float.

kaldi.base.io.write_int(os:ostream, binary:bool, t:int)

Writes an integer.

kaldi.base.io.write_int_pair_vector(os:ostream, binary:bool, v:list<tuple<int, int>>)

Writes a sequence of integer pairs.

kaldi.base.io.write_int_vector(os:ostream, binary:bool, v:list<int>)

Writes a sequence of integers.

kaldi.base.io.write_text(os:ostream, s:str)

Writes unicode string to output stream.

kaldi.base.io.write_token(os:ostream, binary:bool, token:str)

Writes non-empty string of non-space characters.

kaldi.base.math

kaldi.base.math.DBL_EPSILON = 2.220446049250313e-16

kaldi.base.math.FLT_EPSILON = 1.1920929e-07

kaldi.base.math.M_PI = 3.141592653589793

kaldi.base.math.M_SQRT2 = 1.4142135623730951

kaldi.base.math.M_2PI = 6.283185307179586

kaldi.base.math.M_SQRT1_2 = 0.7071067811865476

kaldi.base.math.M_LOG_2PI = 1.8378770664093456

kaldi.base.math.M_LN2 = 0.6931471805599453

kaldi.base.math.M_LN10 = 2.302585092994046

kaldi.base.math.LOG_ZERO_FLOAT = -inf

kaldi.base.math.LOG_ZERO_DOUBLE = -inf

kaldi.base.math.MIN_LOG_DIFF_DOUBLE = -36.04365158081055

kaldi.base.math.MIN_LOG_DIFF_FLOAT = -15.942384719848633

Functions

approx_equal	Returns abs(a - b) <= relative_tolerance * (abs(a) + abs(b))
divide_rounding_down	Returns a / b, rounding towards negative infinity in all cases.
double_exp	Computes e raised to the given power.
double_hypot	Computes square root of the sum of the squares of two given numbers.
double_isfinite	Determines if the given floating point number has finite value.
double_isinf	Determines if the given floating point number is a positive or negative infinity.
double_isnan	Determines if the given floating point number is a not-a-number (NaN) value.
double_log	Computes the natural (base e) logarithm.
double_log1p	Computes natural logarithm of 1 plus the given number (ln(1+x))
double_log_add	Computes log(exp(x) + exp(y)) without losing precision.
double_log_sub	Computes log(exp(x) + exp(y)) without losing precision.
double_rand_gauss2	Returns a pair of gaussian random numbers using Box-Muller transform
double_sqr	Squares the given floating point number.
exp	Computes e raised to the given power.
factorize	Splits a positive integer into its prime factors.
gcd	Returns the greatest common divisor of two integers.
hypot	Computes square root of the sum of the squares of two given numbers.
isfinite	Determines if the given floating point number has finite value.
isinf	Determines if the given floating point number is a positive or negative infinity.
isnan	Determines if the given floating point number is a not-a-number (NaN) value.
lcm	Returns the least common multiple for x and y.
log	Computes the natural (base e) logarithm.
log1p	Computes natural logarithm of 1 plus the given number (ln(1+x))
log_add	Computes log(exp(x) + exp(y)) without losing precision.
log_sub	Computes log(exp(x) - exp(y)) without losing precision.
rand	Returns a random integer between 0 and RAND_MAX, inclusive.
rand_gauss	Returns a random number from standard normal distribution.
rand_gauss2	Returns a pair of gaussian random numbers using Box-Muller transform
rand_int	Returns a random integer in the given interval.
rand_poisson	Returns a random number from poisson distribution.
rand_prune	Prunes by preserving expectations.
rand_uniform	Returns a random number from uniform distribution (strictly between 0 and 1).
round_up_to_nearest_power_of_two	Rounds input integer up to the nearest power of 2.
sqr	Squares the given floating point number.
with_prob	Returns True with probability prob.

Classes

RandomState

Thread-safe random number generator state.

class kaldi.base.math.RandomState

Thread-safe random number generator state.

seed

Seed for generating random numbers

kaldi.base.math.approx_equal(a:float, b:float, relative_tolerance:float=default) → bool

Returns abs(a - b) <= relative_tolerance * (abs(a) + abs(b))

kaldi.base.math.divide_rounding_down(a:int, b:int) → int

Returns a / b, rounding towards negative infinity in all cases.

kaldi.base.math.double_exp(x:float) → float

Computes e raised to the given power.

kaldi.base.math.double_hypot(x:float, y:float) → float

Computes square root of the sum of the squares of two given numbers.

kaldi.base.math.double_isfinite(arg:float) → bool

Determines if the given floating point number has finite value.

kaldi.base.math.double_isinf(arg:float) → bool

Determines if the given floating point number is a positive or negative infinity.

kaldi.base.math.double_isnan(arg:float) → bool

Determines if the given floating point number is a not-a-number (NaN) value.

kaldi.base.math.double_log(x:float) → float

Computes the natural (base e) logarithm.

kaldi.base.math.double_log1p(x:float) → float

Computes natural logarithm of 1 plus the given number (ln(1+x))

kaldi.base.math.double_log_add(x:float, y:float) → float

Computes log(exp(x) + exp(y)) without losing precision.

kaldi.base.math.double_log_sub(x:float, y:float) → float

Computes log(exp(x) + exp(y)) without losing precision.

kaldi.base.math.double_rand_gauss2(state:RandomState=default) -> (a:float, b:float)

Returns a pair of gaussian random numbers using Box-Muller transform

kaldi.base.math.double_sqr(x:float) → float

Squares the given floating point number.

kaldi.base.math.exp(x:float) → float

Computes e raised to the given power.

kaldi.base.math.factorize(x)

Splits a positive integer into its prime factors.

Parameters:x (int) – positive integer Returns:Prime factors in increasing order, with duplication. Return type:List[int] Raises:ValueError if x <= 0

kaldi.base.math.gcd(m:int, n:int) → int

Returns the greatest common divisor of two integers.

kaldi.base.math.hypot(x:float, y:float) → float

Computes square root of the sum of the squares of two given numbers.

kaldi.base.math.isfinite(arg:float) → bool

Determines if the given floating point number has finite value.

kaldi.base.math.isinf(arg:float) → bool

Determines if the given floating point number is a positive or negative infinity.

kaldi.base.math.isnan(arg:float) → bool

Determines if the given floating point number is a not-a-number (NaN) value.

kaldi.base.math.lcm(x, y)

Returns the least common multiple for x and y.

Parameters:

• x (int) – first positive integer
• y (int) – second positive integer

Raises:

ValueError if x <= 0 or y <= 0

kaldi.base.math.log(x:float) → float

Computes the natural (base e) logarithm.

kaldi.base.math.log1p(x:float) → float

Computes natural logarithm of 1 plus the given number (ln(1+x))

kaldi.base.math.log_add(x:float, y:float) → float

Computes log(exp(x) + exp(y)) without losing precision.

kaldi.base.math.log_sub(x:float, y:float) → float

Computes log(exp(x) - exp(y)) without losing precision.

kaldi.base.math.rand(state:RandomState=default) → int

Returns a random integer between 0 and RAND_MAX, inclusive.

kaldi.base.math.rand_gauss(state:RandomState=default) → float

Returns a random number from standard normal distribution.

kaldi.base.math.rand_gauss2(state:RandomState=default) -> (a:float, b:float)

Returns a pair of gaussian random numbers using Box-Muller transform

kaldi.base.math.rand_int(first, last, state=None)

Returns a random integer in the given interval.

Parameters:

• first (int) – Lower bound
• last (int) – Upper bound (inclusive)
• state (RandomState or None) – RNG state

Raises:

ValueError if first >= last

kaldi.base.math.rand_poisson(lambda:float, state:RandomState=default) → int

Returns a random number from poisson distribution.

kaldi.base.math.rand_prune(post:float, prune_tresh:float, state:RandomState=default) → float

Prunes by preserving expectations.

This is typically used to prune posteriors.

kaldi.base.math.rand_uniform(state:RandomState=default) → float

Returns a random number from uniform distribution (strictly between 0 and 1).

kaldi.base.math.round_up_to_nearest_power_of_two(n)

Rounds input integer up to the nearest power of 2.

Parameters:n (int) – Positive integer Raises:ValueError if n <= 0

kaldi.base.math.sqr(x:float) → float

Squares the given floating point number.

kaldi.base.math.with_prob(prob)

Returns True with probability prob.

Parameters:prob (float) – probability of True, 0 <= prob <= 1 Raises:ValueError – If prob is negative or greater than 1.0.