Split models (#24)

Resolves #15 
Resolves #16 
Resolves #18 
Resolves #19 
Resolves #20 
Resolves #21 
Resolves #23

.coveragerc

@@ -0,0 +1,36 @@
+[run]
+source = aligner
+omit =
+    *_rc.py
+    *_ui.py
+
+[report]
+exclude_lines =
+    pragma: no cover
+    if __name__ == .__main__.:
+
+    raise AssertionError
+    raise NotImplementedError
+    pass
+
+    if sys\.platform
+    if hasattr\(sys, 'frozen'\):
+
+    except ImportError:
+
+    if call_back
+    if stop_check
+
+    def run\(self\):
+
+    def calc\(self\):
+
+    def setResults
+
+    if dialog\.exec_():
+
+    if not skip_input.*:
+
+    if self.debug:
+
+    if debug:

.gitignore

@@ -14,6 +14,8 @@ thirdparty
 !thirdparty/*.md
 !thirdparty/*.py
 
+*.fst
+
 # C extensions
 *.so
 

.travis.yml

@@ -29,12 +29,6 @@ install:
   - which python
   - conda list
 
-before_script:
-  - ls $HOME/tools/kaldi
-  - cp -Rf $HOME/tools/kaldi/ thirdparty/bin/
-  - ls thirdparty/bin
-  - export LD_LIBRARY_PATH=$HOME/tools/kaldi:$LD_LIBRARY_PATH
-
 # Run test
 script:
   - travis_wait 120 "coverage run --source=aligner setup.py test"

README.md

@@ -8,3 +8,5 @@
 The Montreal Forced Aligner is a command line utility for performing forced alignment of speech datasets using Kaldi (http://kaldi-asr.org/).
 
 Please see the documentation http://montreal-forced-aligner.readthedocs.io for installation and usage.
+
+If you run into any issues, please check the [mailing list](https://groups.google.com/forum/#!forum/mfa-users) for fixes/workarounds or to post a new issue.

aligner/__init__.py

@@ -4,14 +4,14 @@
 __ver_tuple__ = (__ver_major__, __ver_minor__, __ver_patch__)
 __version__ = "%d.%d.%d" % __ver_tuple__
 
-__all__ = ['aligner', 'command_line', 'archive', 'corpus', 'config', 'dictionary', 'exceptions',
-            'helper', 'multiprocessing', 'textgrid', 'utils']
+__all__ = ['aligner', 'command_line', 'models', 'corpus', 'config', 'dictionary', 'exceptions',
+            'helper', 'multiprocessing', 'textgrid', 'utils', 'g2p']
 
 import aligner.aligner as aligner
 
 import aligner.command_line as command_line
 
-import aligner.archive as archive
+import aligner.models as models
 
 import aligner.corpus as corpus
 
@@ -28,3 +28,5 @@
 import aligner.textgrid as textgrid
 
 import aligner.utils as utils
+
+import aligner.g2p as g2p

aligner/aligner/base.py

@@ -14,7 +14,9 @@
 
 from ..exceptions import NoSuccessfulAlignments
 
-TEMP_DIR = os.path.expanduser('~/Documents/MFA')
+from .. import __version__
+
+from ..config import TEMP_DIR
 
 
 class BaseAligner(object):
@@ -63,7 +65,7 @@ def __init__(self, corpus, dictionary, output_directory,
         if self.corpus.num_jobs != num_jobs:
             num_jobs = self.corpus.num_jobs
         self.num_jobs = num_jobs
-        if temp_directory is None:
+        if not temp_directory:
             temp_directory = TEMP_DIR
         self.temp_directory = temp_directory
         self.call_back = call_back
@@ -78,6 +80,15 @@ def setup(self):
         self.corpus.initialize_corpus(self.dictionary)
         print(self.corpus.speaker_utterance_info())
 
+    @property
+    def meta(self):
+        data = {'phones':sorted(self.dictionary.nonsil_phones),
+                'version': __version__,
+                'architecture':'gmm-hmm',
+                'features':'mfcc+deltas',
+                }
+        return data
+
     @property
     def mono_directory(self):
         return os.path.join(self.temp_directory, 'mono')
@@ -124,7 +135,6 @@ def export_textgrids(self):
             model_directory = self.tri_directory
         elif os.path.exists(self.mono_final_model_path):
             model_directory = self.mono_directory
-        print('hello')
         convert_ali_to_textgrids(self.output_directory, model_directory, self.dictionary,
                                  self.corpus, self.num_jobs)
 

aligner/aligner/pretrained.py

@@ -1,13 +1,11 @@
 import os
 import shutil
 from tqdm import tqdm
-import time
 import re
 
 from .base import BaseAligner, TEMP_DIR, TriphoneFmllrConfig, TriphoneConfig
 
-from ..dictionary import Dictionary
-from ..corpus import load_scp, save_scp
+from ..exceptions import PronunciationAcousticMismatchError
 
 from ..multiprocessing import (align, calc_fmllr, test_utterances, thirdparty_binary, subprocess,
                                convert_ali_to_textgrids)
@@ -41,10 +39,12 @@ class PretrainedAligner(BaseAligner):
 
     Parameters
     ----------
-    archive : :class:`~aligner.archive.Archive`
-        Archive containing the acoustic model and pronunciation dictionary
     corpus : :class:`~aligner.corpus.Corpus`
         Corpus object for the dataset
+    dictionary : :class:`~aligner.dictionary.Dictionary`
+        Dictionary object for the pronunciation dictionary
+    acoustic_model : :class:`~aligner.models.AcousticModel`
+        Archive containing the acoustic model and pronunciation dictionary
     output_directory : str
         Path to directory to save TextGrids
     temp_directory : str, optional
@@ -56,7 +56,7 @@ class PretrainedAligner(BaseAligner):
         Specifies a call back function for alignment
     '''
 
-    def __init__(self, archive, corpus, output_directory,
+    def __init__(self, corpus, dictionary, acoustic_model, output_directory,
                  temp_directory=None, num_jobs=3, speaker_independent=False,
                  call_back=None, debug=False):
         self.debug = debug
@@ -66,11 +66,9 @@ def __init__(self, archive, corpus, output_directory,
         self.output_directory = output_directory
         self.corpus = corpus
         self.speaker_independent = speaker_independent
-        self.dictionary = Dictionary(archive.dictionary_path, os.path.join(temp_directory, 'dictionary'),
-                                     word_set=corpus.word_set, debug=debug)
-
+        self.dictionary = dictionary
         self.setup()
-        archive.export_triphone_model(self.tri_directory)
+        acoustic_model.export_triphone_model(self.tri_directory)
         log_dir = os.path.join(self.tri_directory, 'log')
         os.makedirs(log_dir, exist_ok=True)
 

aligner/aligner/trainable.py

@@ -14,7 +14,7 @@
 
 from .base import BaseAligner
 
-from ..archive import Archive
+from ..models import AcousticModel
 
 
 class TrainableAligner(BaseAligner):
@@ -55,13 +55,13 @@ def save(self, path):
         '''
         directory, filename = os.path.split(path)
         basename, _ = os.path.splitext(filename)
-        archive = Archive.empty(basename)
-        archive.add_triphone_model(self.tri_fmllr_directory)
-        archive.add_triphone_fmllr_model(self.tri_fmllr_directory)
-        archive.add_dictionary(self.dictionary)
+        acoustic_model = AcousticModel.empty(basename)
+        acoustic_model.add_meta_file(self)
+        acoustic_model.add_triphone_model(self.tri_fmllr_directory)
+        acoustic_model.add_triphone_fmllr_model(self.tri_fmllr_directory)
         os.makedirs(directory, exist_ok=True)
         basename, _ = os.path.splitext(path)
-        archive.dump(basename)
+        acoustic_model.dump(basename)
         print('Saved model to {}'.format(path))
 
     def _do_tri_training(self):