batchalign 0.7.21__tar.gz → 0.7.21.post1__tar.gz

{batchalign-0.7.21/batchalign.egg-info → batchalign-0.7.21.post1}/PKG-INFO

@@ -1,6 +1,6 @@
-Metadata-Version: 2.2
+Metadata-Version: 2.4
 Name: batchalign
-Version: 0.7.21
+Version: 0.7.21.post1
 Summary: Python Speech Language Sample Analysis
 Author: Brian MacWhinney, Houjun Liu
 Author-email: macw@cmu.edu, houjun@cmu.edu
@@ -17,6 +17,8 @@ Requires-Dist: pydub
 Requires-Dist: plotly>=5.3.0
 Requires-Dist: transformers>=4.38.2
 Requires-Dist: tokenizers>=0.14.1
+Requires-Dist: numba>=0.61.0
+Requires-Dist: numpy<=2.2
 Requires-Dist: pycountry>=22.3
 Requires-Dist: stanza[transformers]>=1.10.1
 Requires-Dist: scipy~=1.11
@@ -35,8 +37,11 @@ Requires-Dist: tiktoken
 Requires-Dist: blobfile
 Requires-Dist: sentencepiece
 Requires-Dist: googletrans
-Requires-Dist: openai-whisper>=20240930
+Requires-Dist: openai-whisper
+Requires-Dist: llvmlite>=0.44.0
 Requires-Dist: praat-parselmouth==0.4.6
+Requires-Dist: pyannote.audio
+Requires-Dist: onnxruntime
 Provides-Extra: dev
 Requires-Dist: pytest; extra == "dev"
 Provides-Extra: train
@@ -49,6 +54,7 @@ Dynamic: author-email
 Dynamic: classifier
 Dynamic: description
 Dynamic: description-content-type
+Dynamic: license-file
 Dynamic: provides-extra
 Dynamic: requires-dist
 Dynamic: summary

{batchalign-0.7.21 → batchalign-0.7.21.post1}/batchalign/cli/dispatch.py

@@ -163,9 +163,11 @@ def _dispatch(command, lang, num_speakers,
                                    callback=lambda *args:progress_callback(file, *args),
                                    **kw)
                 msgs = [escape(str(i.message)).strip() for i in w]
+                msgs = [i for i in msgs if "torchaudio" not in i.lower()]
                 # write the format, as needed
                 writer(doc, output)
                 # print any warnings
+
                 if len(msgs) > 0:
                     if ctx.obj["verbose"] > 1:
                         Console().print(f"\n[bold yellow]WARN[/bold yellow] on {file}:\n","\n".join(msgs)+"\n")

{batchalign-0.7.21 → batchalign-0.7.21.post1}/batchalign/pipelines/__init__.py

@@ -13,4 +13,6 @@ from .analysis import EvaluationEngine
 from .utterance import StanzaUtteranceEngine
 
 from .translate import SeamlessTranslationModel, GoogleTranslateEngine
-from .avqi import AVQIEngine
+from .avqi import AVQIEngine
+
+from .diarization import PyannoteEngine

{batchalign-0.7.21 → batchalign-0.7.21.post1}/batchalign/pipelines/asr/rev.py

@@ -28,9 +28,9 @@ class RevEngine(BatchalignEngine):
         # if there is no utterance segmentation scheme, we only
         # run ASR
         if self.__engine:
-            return [ Task.ASR, Task.SPEAKER_RECOGNITION, Task.UTTERANCE_SEGMENTATION ]
+            return [ Task.ASR, Task.UTTERANCE_SEGMENTATION ]
         else:
-            return [ Task.ASR, Task.SPEAKER_RECOGNITION ]
+            return [ Task.ASR ]
 
     def __init__(self, key:str=None, lang="eng", num_speakers=2):
 

batchalign-0.7.21.post1/batchalign/pipelines/diarization/__init__.py

@@ -0,0 +1 @@
+from .pyannote import PyannoteEngine

batchalign-0.7.21.post1/batchalign/pipelines/diarization/pyannote.py

@@ -0,0 +1,103 @@
+# system utils
+import glob, os, re
+from itertools import groupby
+
+# pathing tools
+from pathlib import Path
+
+# UD tools
+import stanza
+
+import copy
+
+from stanza.utils.conll import CoNLL
+from stanza import Document, DownloadMethod
+from stanza.models.common.doc import Token
+from stanza.pipeline.core import CONSTITUENCY
+from stanza import DownloadMethod
+from torch import heaviside
+
+from stanza.pipeline.processor import ProcessorVariant, register_processor_variant
+from stanza.resources.common import download_resources_json, load_resources_json, get_language_resources
+
+# the loading bar
+from tqdm import tqdm
+
+from bdb import BdbQuit
+
+from nltk import word_tokenize
+from collections import defaultdict
+
+import warnings
+
+from stanza.utils.conll import CoNLL
+
+# Oneliner of directory-based glob and replace
+globase = lambda path, statement: glob.glob(os.path.join(path, statement))
+repath_file = lambda file_path, new_dir: os.path.join(new_dir, pathlib.Path(file_path).name)
+
+
+from batchalign.document import *
+from batchalign.constants import *
+from batchalign.pipelines.base import *
+from batchalign.formats.chat.parser import chat_parse_utterance
+
+from batchalign.utils.dp import *
+
+from pyannote.audio import Pipeline
+
+import logging
+L = logging.getLogger("batchalign")
+
+import pycountry
+
+class PyannoteEngine(BatchalignEngine):
+    tasks = [ Task.SPEAKER_RECOGNITION ]
+    status_hook = None
+
+    def __init__(self, num_speakers=2):
+        self.pipe = Pipeline.from_pretrained("talkbank/dia-fork")
+        self.num_speakers = num_speakers
+
+    def process(self, doc):
+        assert doc.media != None and doc.media.url != None, f"We cannot diarize something that doesn't have a media path! Provided media tier='{doc.media}'"
+        res = self.pipe(doc.media.url, num_speakers=self.num_speakers)
+
+        speakers = list(set([int(i[-1].split("_")[-1])
+                            for i in res.itertracks(yield_label=True)]))
+        corpus = doc.tiers[0].corpus
+        lang = doc.tiers[0].lang
+        tiers = {
+            i:
+            Tier(
+                lang=lang, corpus=corpus,
+                id="PAR"+str(i), name="Participant",
+                birthday="",
+            )
+            for i in speakers
+        }
+
+        for i in doc.content:
+            if not isinstance(i, Utterance):
+                continue
+            if i.alignment is None:
+                continue
+            start,end = i.alignment
+            if start is None or end is None:
+                continue
+
+            for (a,b),_,speaker in res.itertracks(yield_label=True):
+                speaker_id = int(speaker.split("_")[-1])
+                tier = tiers.get(speaker_id)
+                # we set the end time of the utterance as the
+                # *LAST* segment it ends before
+                # i.e. [seg_end, ....., ut_end]
+                # like that 
+                if b <= end/1000 and tier:
+                    i.tier = tier
+
+        # doc.tiers = list(tiers.values())
+
+        return doc
+
+

{batchalign-0.7.21 → batchalign-0.7.21.post1}/batchalign/pipelines/dispatch.py

@@ -7,7 +7,7 @@ from batchalign import (WhisperEngine, WhisperFAEngine, StanzaEngine, RevEngine,
                         NgramRetraceEngine, DisfluencyReplacementEngine, WhisperUTREngine,
                         RevUTREngine, EvaluationEngine, WhisperXEngine, NemoSpeakerEngine,
                         StanzaUtteranceEngine, CorefEngine, Wave2VecFAEngine, SeamlessTranslationModel,
-                        GoogleTranslateEngine, OAIWhisperEngine)
+                        GoogleTranslateEngine, OAIWhisperEngine, PyannoteEngine)
 from batchalign import BatchalignPipeline
 from batchalign.models import resolve
 
@@ -22,7 +22,7 @@ DEFAULT_PACKAGES = {
     "asr": "whisper_oai",
     "utr": "whisper_utr",
     "fa": "whisper_fa",
-    "speaker": "nemo_speaker",
+    "speaker": "pyannote",
     "morphosyntax": "stanza",
     "disfluency": "replacement",
     "retracing": "ngram",
@@ -72,6 +72,8 @@ def dispatch_pipeline(pkg_str, lang, num_speakers=None, **arg_overrides):
     # if asr is in engines but disfluency or retracing is not
     # add them
     if "asr" in packages:
+        if "speaker" not in packages:
+            packages.append("speaker")
         if "disfluency" not in packages:
             packages.append("disfluency")
         if "retracing" not in packages:
@@ -137,6 +139,9 @@ def dispatch_pipeline(pkg_str, lang, num_speakers=None, **arg_overrides):
             engines.append(GoogleTranslateEngine())
         elif engine == "whisper_oai":
             engines.append(OAIWhisperEngine())
+        elif engine == "pyannote":
+            engines.append(PyannoteEngine())
+
 
     L.debug(f"Done initalizing packages.")
     return BatchalignPipeline(*engines)

batchalign-0.7.21.post1/batchalign/version

@@ -0,0 +1,3 @@
+0.7.21-post.1
+August 28, 2025
+Pyannote support

{batchalign-0.7.21 → batchalign-0.7.21.post1/batchalign.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
-Metadata-Version: 2.2
+Metadata-Version: 2.4
 Name: batchalign
-Version: 0.7.21
+Version: 0.7.21.post1
 Summary: Python Speech Language Sample Analysis
 Author: Brian MacWhinney, Houjun Liu
 Author-email: macw@cmu.edu, houjun@cmu.edu
@@ -17,6 +17,8 @@ Requires-Dist: pydub
 Requires-Dist: plotly>=5.3.0
 Requires-Dist: transformers>=4.38.2
 Requires-Dist: tokenizers>=0.14.1
+Requires-Dist: numba>=0.61.0
+Requires-Dist: numpy<=2.2
 Requires-Dist: pycountry>=22.3
 Requires-Dist: stanza[transformers]>=1.10.1
 Requires-Dist: scipy~=1.11
@@ -35,8 +37,11 @@ Requires-Dist: tiktoken
 Requires-Dist: blobfile
 Requires-Dist: sentencepiece
 Requires-Dist: googletrans
-Requires-Dist: openai-whisper>=20240930
+Requires-Dist: openai-whisper
+Requires-Dist: llvmlite>=0.44.0
 Requires-Dist: praat-parselmouth==0.4.6
+Requires-Dist: pyannote.audio
+Requires-Dist: onnxruntime
 Provides-Extra: dev
 Requires-Dist: pytest; extra == "dev"
 Provides-Extra: train
@@ -49,6 +54,7 @@ Dynamic: author-email
 Dynamic: classifier
 Dynamic: description
 Dynamic: description-content-type
+Dynamic: license-file
 Dynamic: provides-extra
 Dynamic: requires-dist
 Dynamic: summary

{batchalign-0.7.21 → batchalign-0.7.21.post1}/batchalign.egg-info/SOURCES.txt

@@ -74,6 +74,8 @@ batchalign/pipelines/cleanup/retrace.py
 batchalign/pipelines/cleanup/support/filled_pauses.eng
 batchalign/pipelines/cleanup/support/replacements.eng
 batchalign/pipelines/cleanup/support/test.test
+batchalign/pipelines/diarization/__init__.py
+batchalign/pipelines/diarization/pyannote.py
 batchalign/pipelines/fa/__init__.py
 batchalign/pipelines/fa/wave2vec_fa.py
 batchalign/pipelines/fa/whisper_fa.py

{batchalign-0.7.21 → batchalign-0.7.21.post1}/batchalign.egg-info/requires.txt

@@ -7,6 +7,8 @@ pydub
 plotly>=5.3.0
 transformers>=4.38.2
 tokenizers>=0.14.1
+numba>=0.61.0
+numpy<=2.2
 pycountry>=22.3
 stanza[transformers]>=1.10.1
 scipy~=1.11
@@ -25,8 +27,11 @@ tiktoken
 blobfile
 sentencepiece
 googletrans
-openai-whisper>=20240930
+openai-whisper
+llvmlite>=0.44.0
 praat-parselmouth==0.4.6
+pyannote.audio
+onnxruntime
 
 [dev]
 pytest

{batchalign-0.7.21 → batchalign-0.7.21.post1}/setup.py

@@ -39,6 +39,8 @@ setup(
         "plotly>=5.3.0",
         "transformers>=4.38.2",
         "tokenizers>=0.14.1",
+        "numba>=0.61.0",
+        "numpy<=2.2",
         "pycountry>=22.3",
         "stanza[transformers]>=1.10.1",
         "scipy~=1.11",
@@ -57,8 +59,11 @@ setup(
         "blobfile",
         "sentencepiece",
         "googletrans",
-        "openai-whisper>=20240930",
-        "praat-parselmouth==0.4.6"  # Added parselmouth for AVQI
+        "openai-whisper",
+        "llvmlite>=0.44.0",
+        "praat-parselmouth==0.4.6",  # Added parselmouth for AVQI
+        "pyannote.audio",
+        "onnxruntime"
     ],
     extras_require={
         'dev': [
@@ -113,4 +118,4 @@ setup(
         "Development Status :: 3 - Alpha",
         "Topic :: Utilities"
     ]
-)
+)

batchalign-0.7.21/batchalign/version

@@ -1,3 +0,0 @@
-0.7.21
-August 26, 2025
-AVQI