# Copyright 2025 Daniil Shmelev
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# =========================================================================
from typing import Union
from pathlib import Path
import numpy as np
import torch
from .param_checks import check_type, check_non_neg, check_log_sig_method
from .error_codes import err_msg
from .dtypes import CPSIG_SIG_TO_LOG_SIG, CPSIG_BATCH_SIG_TO_LOG_SIG
from .sig_length import sig_length, log_sig_length
from .sig import sig
from .data_handlers import SigOutputHandler, DeviceToHost, SigInputHandler
from .load_siglib import CPSIG
######################################################
# Python wrappers
######################################################
[docs]
def set_cache_dir(
dir : str
):
"""
Sets the cache directory to use in ``pysiglib.prepare_log_sig``
when ``use_disk=True``. If the cache directory is not explicitly
set by a call to this function, a default directory will be used:
- Windows: ``%LOCALAPPDATA%``
- Linux: ``~/.cache``
- Mac: ``~/Library/Caches``
This function is not thread safe.
:param dir: Path to cache directory
:type dir: str
Example usage:
----------------
.. code-block::
import pysiglib
# Set cache dir to a folder "my_cache_dir" in the current working directory
pysiglib.set_cache_dir("./my_cache_dir")
pysiglib.prepare_log_sig(5, 3, lead_lag=True, method=2, use_disk=True)
X = torch.rand((32,100,5))
X_log_sig = pysiglib.log_sig(X, 3, lead_lag=True, method=2)
"""
check_type(dir, "dir", str)
p = Path(dir)
if not p.exists():
raise ValueError(f"Path does not exist: {p}")
if not p.is_dir():
raise ValueError(f"Path is not a directory: {p}")
err_code = CPSIG.set_cache_dir(dir.encode("utf-8"))
if err_code:
raise Exception("Error in pysiglib.prepare_log_sig: " + err_msg(err_code))
[docs]
def prepare_log_sig(
dimension : int,
degree : int,
method : int,
time_aug : bool = False,
lead_lag : bool = False,
use_disk : bool = False
):
"""
Prepares for log signature computations. For details concerning the ``method`` parameter,
see the page :doc:`Computing Log Signatures </pages/log_signatures/log_sig_methods>`.
This function is not thread safe.
:param dimension: Dimension of the underlying path(s).
:type dimension: int
:param degree: Truncation degree of the log signature.
:type degree: int
:param method: Method for the log signature computation. Must be one of `0`, `1` or `2`.
:type method: int
:param time_aug: Whether time augmentation will be used in the computation.
:type time_aug: bool
:param lead_lag: Whether the lead lag transform will be used in the computation.
:type lead_lag: bool
:param use_disk: If ``False``, will cache prepared objects in memory only.
If ``True``, will also save these objects in a cache directory to be
re-used for future runs. See additionally the documentation for
``pysiglib.set_cache_dir``.
:type use_disk: bool
Example usage:
----------------
.. code-block::
import pysiglib
pysiglib.prepare_log_sig(5, 3, lead_lag=True, method=2, use_disk=True)
X = torch.rand((32,100,5))
X_log_sig = pysiglib.log_sig(X, 3, lead_lag=True, method=2)
"""
check_type(dimension, "dimension", int)
check_type(degree, "degree", int)
check_type(method, "method", int)
check_log_sig_method(method)
check_type(time_aug, "time_aug", bool)
check_type(lead_lag, "lead_lag", bool)
if method == 0:
return
aug_dimension = (2 * dimension if lead_lag else dimension) + (1 if time_aug else 0)
err_code = CPSIG.prepare_log_sig(
aug_dimension,
degree,
method,
use_disk
)
if err_code:
raise Exception("Error in pysiglib.prepare_log_sig: " + err_msg(err_code))
[docs]
def clear_cache(
use_disk : bool = False
):
"""
Clears the cache generated by ``pysiglib.prepare_log_sig``.
:param use_disk: If ``False``, will clear the cache from memory only.
If ``True``, will also clear the cache directory.
See additionally the documentation for
``pysiglib.set_cache_dir``.
:type use_disk: bool
"""
err_code = CPSIG.clear_cache(use_disk)
if err_code:
raise Exception("Error in pysiglib.prepare_log_sig: " + err_msg(err_code))
def sig_to_log_sig_(data, result, data_dimension, degree, time_aug, lead_lag, method):
err_code = CPSIG_SIG_TO_LOG_SIG[data.dtype](
data.data_ptr,
result.data_ptr,
data_dimension,
degree,
time_aug,
lead_lag,
method
)
if err_code:
raise Exception("Error in pysiglib.sig_to_log_sig: " + err_msg(err_code))
return result.data
def batch_sig_to_log_sig_(data, result, data_dimension, degree, time_aug, lead_lag, method, n_jobs = 1):
err_code = CPSIG_BATCH_SIG_TO_LOG_SIG[data.dtype](
data.data_ptr,
result.data_ptr,
data.batch_size,
data_dimension,
degree,
time_aug,
lead_lag,
method,
n_jobs
)
if err_code:
raise Exception("Error in pysiglib.sig_to_log_sig: " + err_msg(err_code))
return result.data
[docs]
def sig_to_log_sig(
sig : Union[np.ndarray, torch.tensor],
dimension : int,
degree : int,
time_aug : bool = False,
lead_lag : bool = False,
method : int = 1,
n_jobs : int = 1
) -> Union[np.ndarray, torch.tensor]:
"""
Computes the log signature from the signature, using the specified method. For details,
see the page :doc:`Computing Log Signatures </pages/log_signatures/log_sig_methods>`.
:param sig: The signature or batch of signatures, given as a `numpy.ndarray` or `torch.tensor`.
For a single signature, this must be of shape ``sig_length``. For a batch of paths, this must
be of shape ``(batch_size, sig_length)``.
:type sig: numpy.ndarray | torch.tensor
:param dimension: Dimension of the underlying path(s).
:type dimension: int
:param degree: Truncation degree of the (log) signature(s).
:type degree: int
:param time_aug: Whether the signatures were computed with ``time_aug=True``.
:type time_aug: bool
:param lead_lag: Whether the signatures were computed with ``lead_lag=True``.
:type lead_lag: bool
:param method: Method to use for the log signature computation (`0`, `1` or `2`).
:type method: int
:param n_jobs: Number of threads to run in parallel.
If n_jobs = 1, the computation is run serially. If set to -1, all available threads
are used. For n_jobs below -1, (max_threads + 1 + n_jobs) threads are used. For example
if n_jobs = -2, all threads but one are used.
:type n_jobs: int
:return: Log signature or a batch of log signatures.
:rtype: numpy.ndarray | torch.tensor
Example usage:
----------------
.. code-block:: python
import torch
import pysiglib
pysiglib.prepare_log_sig(5, 3, lead_lag=True, method=2)
X = torch.rand((32,100,5))
X_sig = pysiglib.sig(X, 3, lead_lag=True)
X_log_sig = pysiglib.sig_to_log_sig(X_sig, 5, 3, lead_lag=True, method=2)
"""
check_type(dimension, "dimension", int)
check_non_neg(dimension, "dimension")
check_type(degree, "degree", int)
check_non_neg(degree, "degree")
check_type(time_aug, "time_aug", bool)
check_type(lead_lag, "lead_lag", bool)
check_type(method, "method", int)
check_log_sig_method(method)
# If path is on GPU, move to CPU
device_handler = DeviceToHost([sig], ["sig"])
sig = device_handler.data[0]
aug_dimension = (2 * dimension if lead_lag else dimension) + (1 if time_aug else 0)
sig_len = sig_length(aug_dimension, degree)
data = SigInputHandler(sig, sig_len, "sig")
log_sig_len = log_sig_length(aug_dimension, degree) if method else sig_length(aug_dimension, degree)
result = SigOutputHandler(data, log_sig_len)
if data.is_batch:
check_type(n_jobs, "n_jobs", int)
if n_jobs == 0:
raise ValueError("n_jobs cannot be 0")
res = batch_sig_to_log_sig_(data, result, dimension, degree, time_aug, lead_lag, method, n_jobs)
else:
res = sig_to_log_sig_(data, result, dimension, degree, time_aug, lead_lag, method)
if device_handler.device is not None:
res = res.to(device_handler.device)
return res
[docs]
def log_sig(
path : Union[np.ndarray, torch.tensor],
degree : int,
time_aug : bool = False,
lead_lag : bool = False,
end_time : float = 1.,
method : int = 1,
n_jobs : int = 1
) -> Union[np.ndarray, torch.tensor]:
"""
Computes the log signature using the specified method. For details,
see the page :doc:`Computing Log Signatures </pages/log_signatures/log_sig_methods>`.
:param path: The underlying path or batch of paths, given as a `numpy.ndarray` or `torch.tensor`.
For a single path, this must be of shape ``(length, dimension)``. For a batch of paths, this must
be of shape ``(batch_size, length, dimension)``.
:type path: numpy.ndarray | torch.tensor
:param degree: Truncation degree of the (log) signature(s).
:type degree: int
:param time_aug: If set to True, will compute the log signature of the time-augmented path, :math:`\\hat{x}_t := (t, x_t)`,
defined as the original path with an extra channel set to time, :math:`t`. This channel spans :math:`[0, t_L]`,
where :math:`t_L` is given by the parameter ``end_time``.
:type time_aug: bool
:param lead_lag: If set to True, will compute the log signature of the path after applying the lead-lag transformation.
:type lead_lag: bool
:param end_time: End time for time-augmentation, :math:`t_L`.
:type end_time: float
:param method: Method to use for the log signature computation (`0`, `1` or `2`).
:type method: int
:param n_jobs: Number of threads to run in parallel.
If n_jobs = 1, the computation is run serially. If set to -1, all available threads
are used. For n_jobs below -1, (max_threads + 1 + n_jobs) threads are used. For example
if n_jobs = -2, all threads but one are used.
:type n_jobs: int
:return: Log signature or a batch of log signatures.
:rtype: numpy.ndarray | torch.tensor
Example usage:
----------------
.. code-block:: python
import torch
import pysiglib
pysiglib.prepare_log_sig(5, 3, lead_lag=True, method=2)
X = torch.rand((32,100,5))
X_log_sig = pysiglib.log_sig(X, 3, lead_lag=True, method=2)
"""
sig_ = sig(path, degree, time_aug, lead_lag, end_time, True, n_jobs)
dimension = path.shape[-1]
log_sig_ = sig_to_log_sig(sig_, dimension, degree, time_aug, lead_lag, method, n_jobs)
return log_sig_
