3.0.5 (#791)

bin/mfa_update

@@ -0,0 +1,18 @@
+#!/usr/bin/env python
+
+import os
+import shutil
+import subprocess
+from importlib.util import find_spec
+
+anchor_found = find_spec("anchor") is not None
+
+conda_path = shutil.which("conda")
+mamba_path = shutil.which("mamba")
+if mamba_path is None:
+    print("No mamba found, installing first...")
+    subprocess.call([conda_path, "install", "-c", "conda-forge", "-y", "mamba"], env=os.environ)
+package_list = ["montreal-forced-aligner", "kalpy", "kaldi=*=cpu*"]
+if anchor_found:
+    package_list.append("anchor-annotator")
+subprocess.call([mamba_path, "update", "-c", "conda-forge", "-y"] + package_list, env=os.environ)

docs/source/changelog/changelog_3.0.rst

@@ -5,6 +5,13 @@
 3.0 Changelog
 *************
 
+3.0.5
+=====
+
+- Added :code:`mfa_update` command to better sync changes across dependencies
+- Updated how calculated properties are loaded to fix crashes in Anchor
+- Change when alignments are analyzed in training
+
 3.0.4
 =====
 

docs/source/installation.rst

@@ -31,7 +31,11 @@ General installation
 Updating Montreal Forced Aligner
 --------------------------------
 
-To install the latest version, please run either :code:`conda update -c conda-forge montreal-forced-aligner --update-deps` or  :code:`mamba update -c conda-forge montreal-forced-aligner --update-deps` if you have mamba installed.
+To install the latest version, please run either :code:`conda update -c conda-forge montreal-forced-aligner kalpy kaldi=*=cpu* --update-deps` or  :code:`mamba update -c conda-forge montreal-forced-aligner kalpy kaldi=*=cpu* --update-deps` if you have mamba installed.
+
+.. versionadded:: 3.0.5
+
+   MFA version 3.0.5 onward has a helper utility for updating to new versions.  Run :code:`mfa_update` to fetch the latest versions of MFA, Kalpy, and Kaldi.
 
 Installing SpeechBrain
 ----------------------

montreal_forced_aligner/acoustic_modeling/trainer.py

@@ -646,7 +646,6 @@ def export_files(
             Flag for including the original text of the corpus files as a tier
         """
         self.align()
-        self.analyze_alignments()
         super(TrainableAligner, self).export_files(
             output_directory, output_format, include_original_text
         )
@@ -708,6 +707,7 @@ def align(self) -> None:
                     {"done": True}
                 )
                 session.commit()
+            self.analyze_alignments()
         except Exception as e:
             with self.session() as session:
                 session.query(CorpusWorkflow).filter(CorpusWorkflow.id == wf.id).update(

montreal_forced_aligner/db.py

@@ -20,6 +20,7 @@
 from praatio import textgrid
 from praatio.utilities.constants import Interval
 from sqlalchemy import Boolean, Column, DateTime, Enum, Float, ForeignKey, Integer, String
+from sqlalchemy.ext.hybrid import hybrid_property
 from sqlalchemy.ext.orderinglist import ordering_list
 from sqlalchemy.orm import Bundle, declarative_base, joinedload, relationship
 
@@ -1393,7 +1394,9 @@ class Utterance(MfaSqlBase):
     id = Column(Integer, primary_key=True, autoincrement=True)
     begin = Column(Float, nullable=False, index=True)
     end = Column(Float, nullable=False)
-    duration = Column(Float, sqlalchemy.Computed('"end" - "begin"'), index=True)
+    _duration = sqlalchemy.orm.deferred(
+        Column("duration", Float, sqlalchemy.Computed('"end" - "begin"'), index=True)
+    )
     channel = Column(Integer, nullable=False)
     num_frames = Column(Integer)
     text = Column(String)
@@ -1418,11 +1421,13 @@ class Utterance(MfaSqlBase):
     xvector = Column(Vector(config.XVECTOR_DIMENSION), nullable=True)
     file_id = Column(Integer, ForeignKey("file.id"), index=True, nullable=False)
     speaker_id = Column(Integer, ForeignKey("speaker.id"), index=True, nullable=False)
-    kaldi_id = Column(
-        String,
-        sqlalchemy.Computed("CAST(speaker_id AS text)|| '-' ||CAST(id AS text)"),
-        unique=True,
-        index=True,
+    _kaldi_id = sqlalchemy.orm.deferred(
+        Column(
+            "kaldi_id",
+            String,
+            sqlalchemy.Computed("CAST(speaker_id AS text)|| '-' ||CAST(id AS text)"),
+            unique=True,
+        )
     )
     job_id = Column(Integer, ForeignKey("job.id"), index=True, nullable=True)
     file = relationship("File", back_populates="utterances", cascade_backrefs=False)
@@ -1449,6 +1454,22 @@ class Utterance(MfaSqlBase):
         ),
     )
 
+    @hybrid_property
+    def duration(self) -> float:
+        return self.end - self.begin
+
+    @duration.expression
+    def duration(cls):
+        return cls._duration
+
+    @hybrid_property
+    def kaldi_id(self) -> str:
+        return f"{self.speaker_id}-{self.id}"
+
+    @kaldi_id.expression
+    def kaldi_id(cls):
+        return cls._kaldi_id
+
     def __repr__(self) -> str:
         """String representation of the utterance object"""
         return f"<Utterance in {self.file_name} by {self.speaker_name} from {self.begin} to {self.end}>"
@@ -1751,7 +1772,9 @@ class PhoneInterval(MfaSqlBase):
     begin = Column(Float, nullable=False, index=True)
     end = Column(Float, nullable=False)
     phone_goodness = Column(Float, nullable=True)
-    duration = Column(Float, sqlalchemy.Computed('"end" - "begin"'))
+    _duration = sqlalchemy.orm.deferred(
+        Column("duration", Float, sqlalchemy.Computed('"end" - "begin"'))
+    )
 
     phone_id = Column(
         Integer, ForeignKey("phone.id", ondelete="CASCADE"), index=True, nullable=False
@@ -1777,6 +1800,14 @@ class PhoneInterval(MfaSqlBase):
         sqlalchemy.Index("phone_utterance_workflow_index", "utterance_id", "workflow_id"),
     )
 
+    @hybrid_property
+    def duration(self) -> float:
+        return self.end - self.begin
+
+    @duration.expression
+    def duration(cls):
+        return cls._duration
+
     def __repr__(self):
         return f"<PhoneInterval {self.phone.kaldi_label} ({self.workflow.workflow_type}) from {self.begin}-{self.end} for utterance {self.utterance_id}>"
 
@@ -1859,7 +1890,9 @@ class WordInterval(MfaSqlBase):
     id = Column(Integer, primary_key=True, autoincrement=True)
     begin = Column(Float, nullable=False, index=True)
     end = Column(Float, nullable=False)
-    duration = Column(Float, sqlalchemy.Computed('"end" - "begin"'))
+    _duration = sqlalchemy.orm.deferred(
+        Column("duration", Float, sqlalchemy.Computed('"end" - "begin"'))
+    )
 
     utterance_id = Column(
         Integer, ForeignKey("utterance.id", ondelete="CASCADE"), index=True, nullable=False
@@ -1891,6 +1924,14 @@ class WordInterval(MfaSqlBase):
         sqlalchemy.Index("word_utterance_workflow_index", "utterance_id", "workflow_id"),
     )
 
+    @hybrid_property
+    def duration(self) -> float:
+        return self.end - self.begin
+
+    @duration.expression
+    def duration(cls):
+        return cls._duration
+
     @classmethod
     def from_ctm(
         self, interval: CtmInterval, utterance: Utterance, workflow_id: int

setup.cfg

@@ -58,6 +58,8 @@ exclude = tests
 [options.entry_points]
 console_scripts =
     mfa = montreal_forced_aligner.command_line.mfa:mfa_cli
+scripts =
+    bin/mfa_update
 
 [options.extras_require]
 anchor =