sonusai 0.18.9__py3-none-any.whl → 0.19.6__py3-none-any.whl

sonusai/utils/ranges.py

@@ -2,18 +2,18 @@ def expand_range(s: str, sort: bool = True) -> list[int]:
     """Returns a list of integers from a string input representing a range."""
     import re
 
-    clean_s = s.replace(':', '-')
-    clean_s = clean_s.replace(';', ',')
-    clean_s = re.sub(r' +', ',', clean_s)
-    clean_s = re.sub(r',+', ',', clean_s)
+    clean_s = s.replace(":", "-")
+    clean_s = clean_s.replace(";", ",")
+    clean_s = re.sub(r" +", ",", clean_s)
+    clean_s = re.sub(r",+", ",", clean_s)
 
     r: list[int] = []
-    for i in clean_s.split(','):
-        if '-' not in i:
+    for i in clean_s.split(","):
+        if "-" not in i:
             r.append(int(i))
         else:
-            l, h = map(int, i.split('-'))
-            r += range(l, h + 1)
+            lo, hi = map(int, i.split("-"))
+            r += range(lo, hi + 1)
 
     if sort:
         r = sorted(r)
@@ -23,12 +23,12 @@ def expand_range(s: str, sort: bool = True) -> list[int]:
 
 def consolidate_range(r: list[int]) -> str:
     """Returns a string representing a range from an input list of integers."""
-    from typing import Generator
+    from collections.abc import Generator
 
     def ranges(i: list[int]) -> Generator[tuple[int, int], None, None]:
         import itertools
 
-        for a, b in itertools.groupby(enumerate(i), lambda pair: pair[1] - pair[0]):
+        for _, b in itertools.groupby(enumerate(i), lambda pair: pair[1] - pair[0]):
             b_list = list(b)
             yield b_list[0][1], b_list[-1][1]
 
@@ -37,7 +37,7 @@ def consolidate_range(r: list[int]) -> str:
     for val in ls:
         entry = str(val[0])
         if val[0] != val[1]:
-            entry += f'-{val[1]}'
+            entry += f"-{val[1]}"
         result.append(entry)
 
-    return ', '.join(result)
+    return ", ".join(result)

sonusai/utils/read_predict_data.py

@@ -7,13 +7,12 @@ def read_predict_data(filename: str) -> Predict:
     """Read predict data from given HDF5 file and return it."""
     import h5py
 
-    from sonusai import SonusAIError
     from sonusai import logger
 
-    logger.debug(f'Reading prediction data from {filename}')
-    with h5py.File(filename, 'r') as f:
+    logger.debug(f"Reading prediction data from {filename}")
+    with h5py.File(filename, "r") as f:
         # prediction data is either [frames, num_classes], or [frames, timesteps, num_classes]
-        predict = np.array(f['predict'])
+        predict = np.array(f["predict"])
 
         if predict.ndim == 2:
             return predict
@@ -21,10 +20,13 @@ def read_predict_data(filename: str) -> Predict:
         if predict.ndim == 3:
             frames, timesteps, num_classes = predict.shape
 
-            logger.debug(f'Reshaping prediction data in {filename} 'f''
-                         f'from [{frames}, {timesteps}, {num_classes}] '
-                         f'to [{frames * timesteps}, {num_classes}]')
-            predict = np.reshape(predict, [frames * timesteps, num_classes], order='F')
+            logger.debug(
+                f"Reshaping prediction data in {filename} "
+                f""
+                f"from [{frames}, {timesteps}, {num_classes}] "
+                f"to [{frames * timesteps}, {num_classes}]"
+            )
+            predict = np.reshape(predict, [frames * timesteps, num_classes], order="F")
             return predict
 
-        raise SonusAIError(f'Invalid prediction data dimensions in {filename}')
+        raise RuntimeError(f"Invalid prediction data dimensions in {filename}")

sonusai/utils/reshape.py

@@ -1,5 +1,3 @@
-from typing import Optional
-
 import numpy as np
 
 from sonusai.mixture.datatypes import Feature
@@ -11,12 +9,14 @@ def get_input_shape(feature: Feature) -> tuple[int, ...]:
     return feature.shape[1:]
 
 
-def reshape_inputs(feature: Feature,
-                   batch_size: int,
-                   truth: Optional[Truth] = None,
-                   timesteps: int = 0,
-                   flatten: bool = False,
-                   add1ch: bool = False) -> tuple[Feature, Optional[Truth]]:
+def reshape_inputs(
+    feature: Feature,
+    batch_size: int,
+    truth: Truth | None = None,
+    timesteps: int = 0,
+    flatten: bool = False,
+    add1ch: bool = False,
+) -> tuple[Feature, Truth | None]:
     """Check SonusAI feature and truth data and reshape feature of size [frames, strides, feature_parameters] into
     one of several options:
 
@@ -38,16 +38,14 @@ def reshape_inputs(feature: Feature,
       feature       reshaped feature
       truth         reshaped truth
     """
-    from sonusai import SonusAIError
-
     frames, strides, feature_parameters = feature.shape
     if truth is not None:
         truth_frames, num_classes = truth.shape
         # Double-check correctness of inputs
         if frames != truth_frames:
-            raise SonusAIError('Frames in feature and truth do not match')
+            raise ValueError("Frames in feature and truth do not match")
     else:
-        num_classes = None
+        num_classes = 0
 
     if flatten:
         feature = np.reshape(feature, (frames, strides * feature_parameters))
@@ -64,12 +62,12 @@ def reshape_inputs(feature: Feature,
         fr2drop = frames_rem + bf_rem
         if fr2drop:
             if feature.ndim == 2:
-                feature = feature[0:-fr2drop, ]  # flattened input
+                feature = feature[0:-fr2drop,]  # flattened input
             elif feature.ndim == 3:
-                feature = feature[0:-fr2drop, ]  # un-flattened input
+                feature = feature[0:-fr2drop,]  # un-flattened input
 
             if truth is not None:
-                truth = truth[0:-fr2drop, ]
+                truth = truth[0:-fr2drop,]
 
         # Reshape
         if feature.ndim == 2:  # flattened input
@@ -88,9 +86,9 @@ def reshape_inputs(feature: Feature,
         # Drop frames if remainder exists (not fitting into a multiple of new number of sequences)
         fr2drop = feature.shape[0] % batch_size
         if fr2drop > 0:
-            feature = feature[0:-fr2drop, ]
+            feature = feature[0:-fr2drop,]
             if truth is not None:
-                truth = truth[0:-fr2drop, ]
+                truth = truth[0:-fr2drop,]
 
     # Add channel dimension if required for input to model (i.e. for cnn type input)
     if add1ch:
@@ -119,25 +117,20 @@ def get_num_classes_from_predict(predict: Predict, timesteps: int = 0) -> int:
     return dims[1]
 
 
-def reshape_outputs(predict: Predict,
-                    truth: Optional[Truth] = None,
-                    timesteps: int = 0) -> tuple[Predict, Optional[Truth]]:
+def reshape_outputs(predict: Predict, truth: Truth | None = None, timesteps: int = 0) -> tuple[Predict, Truth | None]:
     """Reshape model output data.
 
     truth and predict can be either [frames, num_classes], or [frames, timesteps, num_classes]
     In binary case, num_classes dim may not exist; detect this and set num_classes to 1.
     """
-    from sonusai import SonusAIError
-
-    if truth is not None:
-        if predict.shape != truth.shape:
-            raise SonusAIError('predict and truth shapes do not match')
+    if truth is not None and predict.shape != truth.shape:
+        raise ValueError("predict and truth shapes do not match")
 
     ndim = predict.ndim
     shape = predict.shape
 
     if not (0 < ndim <= 3):
-        raise SonusAIError(f'do not know how to reshape data with {ndim} dimensions')
+        raise ValueError(f"do not know how to reshape data with {ndim} dimensions")
 
     if ndim == 3 or (ndim == 2 and timesteps > 0):
         if ndim == 2:

sonusai/utils/seconds_to_hms.py

@@ -4,4 +4,4 @@ def seconds_to_hms(seconds: float) -> str:
     s = seconds - h * 3600
     m = int(s / 60)
     s = int(seconds - h * 3600 - m * 60)
-    return f'{h:d}:{m:02d}:{s:02d} (H:MM:SS)'
+    return f"{h:d}:{m:02d}:{s:02d} (H:MM:SS)"

sonusai/utils/stacked_complex.py

@@ -12,14 +12,12 @@ def stack_complex(unstacked: np.ndarray) -> np.ndarray:
     :return: A stacked array
     :raises TypeError:
     """
-    from sonusai import SonusAIError
-
     if not unstacked.ndim > 1:
-        raise SonusAIError('unstacked must have more than 1 dimension')
+        raise ValueError("unstacked must have more than 1 dimension")
 
     shape = list(unstacked.shape)
     shape[-1] = shape[-1] * 2
-    stacked = np.empty(shape, dtype=np.complex64)
+    stacked = np.empty(shape, dtype=np.float32)
     half = unstacked.shape[-1]
     stacked[..., :half] = np.real(unstacked)
     stacked[..., half:] = np.imag(unstacked)
@@ -35,13 +33,11 @@ def unstack_complex(stacked: np.ndarray) -> np.ndarray:
     :return: An unstacked complex array
     :raises TypeError:
     """
-    from sonusai import SonusAIError
-
     if not stacked.ndim > 1:
-        raise SonusAIError('stacked must have more than 1 dimension')
+        raise ValueError("stacked must have more than 1 dimension")
 
     if stacked.shape[-1] % 2 != 0:
-        raise SonusAIError('last dimension of stacked must be a multiple of 2')
+        raise ValueError("last dimension of stacked must be a multiple of 2")
 
     half = stacked.shape[-1] // 2
     unstacked = 1j * stacked[..., half:]
@@ -58,13 +54,11 @@ def stacked_complex_real(stacked: np.ndarray) -> np.ndarray:
     :return: The real elements
     :raises TypeError:
     """
-    from sonusai import SonusAIError
-
     if not stacked.ndim > 1:
-        raise SonusAIError('stacked must have more than 1 dimension')
+        raise ValueError("stacked must have more than 1 dimension")
 
     if stacked.shape[-1] % 2 != 0:
-        raise SonusAIError('last dimension of stacked must be a multiple of 2')
+        raise ValueError("last dimension of stacked must be a multiple of 2")
 
     half = stacked.shape[-1] // 2
     return stacked[..., :half]
@@ -78,13 +72,11 @@ def stacked_complex_imag(stacked: np.ndarray) -> np.ndarray:
     :return: The imaginary elements
     :raises TypeError:
     """
-    from sonusai import SonusAIError
-
     if not stacked.ndim > 1:
-        raise SonusAIError('stacked must have more than 1 dimension')
+        raise ValueError("stacked must have more than 1 dimension")
 
     if stacked.shape[-1] % 2 != 0:
-        raise SonusAIError('last dimension of stacked must be a multiple of 2')
+        raise ValueError("last dimension of stacked must be a multiple of 2")
 
     half = stacked.shape[-1] // 2
     return stacked[..., half:]

sonusai/utils/stratified_shuffle_split.py

@@ -1,14 +1,14 @@
-from typing import Optional
-
 import numpy as np
 
 from sonusai.mixture import MixtureDatabase
 
 
-def stratified_shuffle_split_mixid(mixdb: MixtureDatabase,
-                                   vsplit: float = 0.2,
-                                   nsplit: int = 0,
-                                   rnd_seed: Optional[int] = 0) -> tuple[list[int], list[int], np.ndarray, np.ndarray]:
+def stratified_shuffle_split_mixid(
+    mixdb: MixtureDatabase,
+    vsplit: float = 0.2,
+    nsplit: int = 0,
+    rnd_seed: int | None = 0,
+) -> tuple[list[int], list[int], np.ndarray, np.ndarray]:
     """
     Create a training and test/validation list of mixture IDs from all mixtures in a mixture database.
     The test/validation split is specified by vsplit (0.0 to 1.0), default 0.2.
@@ -35,20 +35,18 @@ def stratified_shuffle_split_mixid(mixdb: MixtureDatabase,
     import random
     from copy import deepcopy
 
-    from sonusai import SonusAIError
     from sonusai import logger
     from sonusai.mixture import get_class_count_from_mixids
-    from sonusai.mixture import get_truth_indices_for_target
 
     if vsplit < 0 or vsplit > 1:
-        raise SonusAIError('vsplit must be between 0 and 1')
+        raise ValueError("vsplit must be between 0 and 1")
 
     a_class_mixid: dict[int, list[int]] = {i + 1: [] for i in range(mixdb.num_classes)}
     for mixid, mixture in enumerate(mixdb.mixtures):
         class_count = get_class_count_from_mixids(mixdb, mixid)
-        if any(class_count) or mixdb.truth_mutex == 0:
-            for truth_index in get_truth_indices_for_target(mixdb.target_files[mixture.targets[0].file_id]):
-                a_class_mixid[truth_index].append(mixid)
+        if any(class_count):
+            for class_index in mixdb.target_files[mixture.targets[0].file_id].class_indices:
+                a_class_mixid[class_index].append(mixid)
         else:
             # no counts and mutex mode means this is all 'other' class
             a_class_mixid[mixdb.num_classes].append(mixid)
@@ -80,11 +78,11 @@ def stratified_shuffle_split_mixid(mixdb: MixtureDatabase,
             # randomize order
             random.shuffle(indices)
 
-        t_class_mixid[ci] = [a_class_mixid[ci + 1][ii] for ii in indices[0:t_num_mixid[ci]]]
-        v_class_mixid[ci] = [a_class_mixid[ci + 1][ii] for ii in indices[t_num_mixid[ci]:]]
+        t_class_mixid[ci] = [a_class_mixid[ci + 1][ii] for ii in indices[0 : t_num_mixid[ci]]]
+        v_class_mixid[ci] = [a_class_mixid[ci + 1][ii] for ii in indices[t_num_mixid[ci] :]]
 
     if np.any(~(t_num_mixid > 0)):
-        logger.warning(f'Some classes have zero coverage: {np.where(~(t_num_mixid > 0))[0]}')
+        logger.warning(f"Some classes have zero coverage: {np.where(~(t_num_mixid > 0))[0]}")
 
     # Stratify over non-zero classes
     nz_indices = np.where(t_num_mixid > 0)[0]
@@ -97,8 +95,10 @@ def stratified_shuffle_split_mixid(mixdb: MixtureDatabase,
     # 2nd stage stratify by class_count/min(class_count-n3) n2 times
     n2 = int(max(min_class - n0 - n3, 0))
 
-    logger.info(f'Stratifying training, x1 cnt {n0}: x(class_count/{n2}): x1 cnt {n3} x1, '
-                f'for {len(nz_indices)} populated classes')
+    logger.info(
+        f"Stratifying training, x1 cnt {n0}: x(class_count/{n2}): x1 cnt {n3} x1, "
+        f"for {len(nz_indices)} populated classes"
+    )
 
     # initialize source list
     tt = deepcopy(t_class_mixid)
@@ -116,13 +116,13 @@ def stratified_shuffle_split_mixid(mixdb: MixtureDatabase,
     # which will leave approx n3 remaining
     if n2 > 0:
         # should always be non-zero
-        min_class = np.min(t_num_mixid2 - n3)
+        min_class = int(np.min(t_num_mixid2 - n3))
         class_count = np.floor((t_num_mixid2 - n3) / min_class)
         # class_count = np.maximum(np.floor((t_num_mixid2 - n3) / n2),0) # Counts per class
         for _ in range(min_class):
             for ci in range(mixdb.num_classes):
                 if class_count[ci] > 0:
-                    for cc in range(int(class_count[ci])):
+                    for _ in range(int(class_count[ci])):
                         # append first
                         t_mixid.append(tt[ci][0])
                         del tt[ci][0]
@@ -133,10 +133,10 @@ def stratified_shuffle_split_mixid(mixdb: MixtureDatabase,
     t_mixid = _extract_remaining_mixids(mixdb, t_mixid, t_num_mixid2, tt)
 
     if len(t_mixid) != sum(t_num_mixid):
-        logger.warning('Final stratified training list length does not match starting list length.')
+        logger.warning("Final stratified training list length does not match starting list length.")
 
     if any(t_num_mixid2) or any(tt):
-        logger.warning('Remaining training mixid list not empty.')
+        logger.warning("Remaining training mixid list not empty.")
 
     # Now stratify the validation list, which is probably not as important, so use simple method
     # initialize source list
@@ -145,18 +145,20 @@ def stratified_shuffle_split_mixid(mixdb: MixtureDatabase,
     v_mixid = _extract_remaining_mixids(mixdb, [], v_num_mixid2, vv)
 
     if len(v_mixid) != sum(v_num_mixid):
-        logger.warning('Final stratified validation list length does not match starting lists length.')
+        logger.warning("Final stratified validation list length does not match starting lists length.")
 
     if any(v_num_mixid2) or any(vv):
-        logger.warning('Remaining validation mixid list not empty.')
+        logger.warning("Remaining validation mixid list not empty.")
 
     return t_mixid, v_mixid, t_num_mixid, v_num_mixid
 
 
-def _extract_remaining_mixids(mixdb: MixtureDatabase,
-                              mixid: list[int],
-                              num_mixid: np.ndarray,
-                              class_mixid: list[list[int]]) -> list[int]:
+def _extract_remaining_mixids(
+    mixdb: MixtureDatabase,
+    mixid: list[int],
+    num_mixid: np.ndarray,
+    class_mixid: list[list[int]],
+) -> list[int]:
     for _ in range(max(num_mixid)):
         for ci in range(mixdb.num_classes):
             if num_mixid[ci] > 0:

sonusai/utils/write_audio.py

@@ -3,7 +3,7 @@ from sonusai.mixture.datatypes import AudioT
 
 
 def write_audio(name: str, audio: AudioT, sample_rate: int = SAMPLE_RATE) -> None:
-    """ Write an audio file.
+    """Write an audio file.
 
     To write multiple channels, use a 2D array of shape [channels, samples].
     The bits per sample and PCM/float are determined by the data type.
@@ -17,7 +17,7 @@ def write_audio(name: str, audio: AudioT, sample_rate: int = SAMPLE_RATE) -> Non
     if data.dim() == 1:
         data = torch.reshape(data, (1, data.shape[0]))
     if data.dim() != 2:
-        raise ValueError(f'audio must be a 1D or 2D array')
+        raise ValueError("audio must be a 1D or 2D array")
 
     # Assuming data has more samples than channels, check if array needs to be transposed
     if data.shape[1] < data.shape[0]:

sonusai/utils/yes_or_no.py

@@ -1,8 +1,8 @@
 def yes_or_no(question: str) -> bool:
     """Wait for yes or no input"""
     while True:
-        reply = str(input(question + ' (y/n)?: ')).lower().strip()
-        if reply[:1] == 'y':
+        reply = str(input(question + " (y/n)?: ")).lower().strip()
+        if reply[:1] == "y":
             return True
-        if reply[:1] == 'n':
+        if reply[:1] == "n":
             return False

sonusai/vars.py

@@ -23,18 +23,18 @@ def main() -> None:
 
     from sonusai.mixture import DEFAULT_NOISE
 
-    print('Custom SonusAI variables:')
-    print('')
-    print(f'${{default_noise}}: {DEFAULT_NOISE}')
-    print('')
-    print('SonusAI recognized environment variables:')
-    print('')
-    print(f'DEEPGRAM_API_KEY {getenv("DEEPGRAM_API_KEY")}')
-    print(f'GOOGLE_SPEECH_API_KEY {getenv("GOOGLE_SPEECH_API_KEY")}')
-    print('')
-    items = ['DEEPGRAM_API_KEY', 'GOOGLE_SPEECH_API_KEY']
-    items += [item for item in environ.keys() if item.upper().startswith("AIXP_WHISPER_")]
-
-
-if __name__ == '__main__':
+    print("Custom SonusAI variables:")
+    print("")
+    print(f"${{default_noise}}: {DEFAULT_NOISE}")
+    print("")
+    print("SonusAI recognized environment variables:")
+    print("")
+    print(f"DEEPGRAM_API_KEY {getenv('DEEPGRAM_API_KEY')}")
+    print(f"GOOGLE_SPEECH_API_KEY {getenv('GOOGLE_SPEECH_API_KEY')}")
+    print("")
+    items = ["DEEPGRAM_API_KEY", "GOOGLE_SPEECH_API_KEY"]
+    items += [item for item in environ if item.upper().startswith("AIXP_WHISPER_")]
+
+
+if __name__ == "__main__":
     main()

{sonusai-0.18.9.dist-info → sonusai-0.19.6.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: sonusai
-Version: 0.18.9
+Version: 0.19.6
 Summary: Framework for building deep neural network models for sound, speech, and voice AI
 Home-page: https://aaware.com
 License: GPL-3.0-only
@@ -8,39 +8,38 @@ Author: Chris Eddington
 Author-email: chris@aaware.com
 Maintainer: Chris Eddington
 Maintainer-email: chris@aaware.com
-Requires-Python: >=3.9,<3.12
+Requires-Python: >=3.11,<3.12
 Classifier: License :: OSI Approved :: GNU General Public License v3 (GPLv3)
 Classifier: Programming Language :: Python :: 3
-Classifier: Programming Language :: Python :: 3.9
-Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
-Requires-Dist: PyYAML (>=6.0.1,<7.0.0)
-Requires-Dist: dataclasses-json (>=0.6.1,<0.7.0)
+Requires-Dist: dataclasses-json (>=0.6.7,<0.7.0)
 Requires-Dist: docopt (>=0.6.2,<0.7.0)
-Requires-Dist: h5py (>=3.11.0,<4.0.0)
-Requires-Dist: jiwer (>=3.0.3,<4.0.0)
-Requires-Dist: librosa (>=0.10.1,<0.11.0)
-Requires-Dist: matplotlib (>=3.8.0,<4.0.0)
+Requires-Dist: h5py (>=3.12.1,<4.0.0)
+Requires-Dist: jiwer (>=3.0.4,<4.0.0)
+Requires-Dist: librosa (>=0.10.2.post1,<0.11.0)
+Requires-Dist: matplotlib (>=3.9.2,<4.0.0)
 Requires-Dist: mgzip (>=0.2.1,<0.3.0)
-Requires-Dist: numpy (>=1.26.4,<2.0.0)
-Requires-Dist: onnx (>=1.14.1,<2.0.0)
-Requires-Dist: onnxruntime (>=1.16.1,<2.0.0)
-Requires-Dist: paho-mqtt (>=2.0.0,<3.0.0)
-Requires-Dist: pandas (>=2.1.1,<3.0.0)
+Requires-Dist: numpy (>=1,<2)
+Requires-Dist: onnx (>=1.17.0,<2.0.0)
+Requires-Dist: onnxruntime (>=1.19.2,<2.0.0)
+Requires-Dist: paho-mqtt (>=2.1.0,<3.0.0)
+Requires-Dist: pandas (>=2.2.3,<3.0.0)
 Requires-Dist: pesq (>=0.0.4,<0.0.5)
 Requires-Dist: praatio (>=6.2.0,<7.0.0)
-Requires-Dist: psutil (>=5,<6)
-Requires-Dist: pyaaware (>=1.5.7,<2.0.0)
+Requires-Dist: psutil (>=6.0.0,<7.0.0)
+Requires-Dist: pyaaware (>=1.5.18,<2.0.0)
 Requires-Dist: pyaudio (>=0.2.14,<0.3.0)
 Requires-Dist: pydub (>=0.25.1,<0.26.0)
-Requires-Dist: pystoi (>=0.4.0,<0.5.0)
-Requires-Dist: requests (>=2.31.0,<3.0.0)
+Requires-Dist: pystoi (>=0.4.1,<0.5.0)
+Requires-Dist: pyyaml (>=6.0.2,<7.0.0)
+Requires-Dist: requests (>=2.32.3,<3.0.0)
+Requires-Dist: rich (>=13.9.4,<14.0.0)
 Requires-Dist: samplerate (>=0.2.1,<0.3.0)
 Requires-Dist: soundfile (>=0.12.1,<0.13.0)
-Requires-Dist: sox (>=1.4.1,<2.0.0)
+Requires-Dist: sox (>=1.5.0,<2.0.0)
 Requires-Dist: torch (>=2.2,<2.3)
 Requires-Dist: torchaudio (>=2.2,<2.3)
-Requires-Dist: tqdm (>=4.66.1,<5.0.0)
+Requires-Dist: tqdm (>=4.66.5,<5.0.0)
 Description-Content-Type: text/x-rst
 
 SonusAI: Framework for simplified creation of deep NN models for sound, speech, and voice AI