hueso/gajim-plugins
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Improve multi modul support and refactor

Browse Source
This commit is contained in:
mesonium
2024-07-02 20:05:45 +02:00
committed by hueso
parent aff69e5b40
commit b0adecef7a
6 changed files with 111 additions and 92 deletions
Download Patch File Download Diff File Expand all files Collapse all files

139
stt_voice_messages/gtk/config_dialog.py
View File

@@ -15,6 +15,7 @@
from __future__ import annotations from __future__ import annotations
from dataclasses import asdict
import logging import logging
import typing import typing
from pathlib import Path from pathlib import Path
@@ -40,45 +41,51 @@ if TYPE_CHECKING:
log = logging.getLogger('gajim.p.sttvm_config_dialog') log = logging.getLogger('gajim.p.sttvm_config_dialog')
SUPPORTED_MODELS: dict[str, dict[str, typing.Union[list[str], Any, str]]] = {
'model_openaiwhisper': { @dataclass
'moduls': ['whisper'], class Model:
'class': openai_whisper.WhisperModel, name: str
'name': 'OpenAI Whisper' required_moduls: list[str]
}, klass: object
'model_ctranslate2': { config: Any
'moduls': ['ctranslate2'], instance: typing.Optional[object] = None
'class': None,
'name': _('CTranslate2')
}, SUPPORTED_MODELS: dict[str, Model] = {
'model_faster-whisper': { 'model_openaiwhisper': Model('OpenAI Whisper',
'moduls': ['faster-whisper'], ['whisper'],
'class': None, openai_whisper.WhisperModel,
'name:': _('Faster-Whisper') OpenAIWhisperSettings),
}, 'model_ctranslate2': Model('CTranslate2',
'model_distill': { ['ctranslate2'],
'moduls': ['transformers', 'accelerate', 'datasets[audio]'], None,
'class': None, None),
'name': _('Distill') 'model_faster-whisper': Model('Fast-Whisper',
} ['faster-whisper'],
None,
None),
'model_distill': Model('Distill',
['transformers', 'accelerate', 'datasets[audio]'],
None,
None)
} }
class Configuration: class Configuration:
def __init__(self, plugin: STTVoiceMessagesPlugin): def __init__(self, plugin: STTVoiceMessagesPlugin):
self._plugin = plugin self._plugin = plugin
self._openaiwhisper_settings = OpenAIWhisperSettings()
self._available_models: dict[ self._available_models: dict[str, Model] = {}
str, dict[str, typing.Union[list[str], Any, str]]] = {}
self.check_available_moduls() self.check_available_moduls()
log.debug('config = %s', self._plugin.config['model_openaiwhisper'])
@property @property
def plugin(self) -> STTVoiceMessagesPlugin: def plugin(self) -> STTVoiceMessagesPlugin:
return self._plugin return self._plugin
@property @property
def available_models(self) -> dict[ def available_models(self) -> dict[str, Model]:
str, dict[str, typing.Union[list[str], Any, str]]]:
return self._available_models return self._available_models
def on_setting(self, value: Any, data: Any) -> None: def on_setting(self, value: Any, data: Any) -> None:
@@ -86,30 +93,35 @@ class Configuration:
value.strip() value.strip()
log.debug('plugin config before:\n %s', self.plugin.config.data) log.debug('plugin config before:\n %s', self.plugin.config.data)
# TODO: Is 'modelname_key = value' a good design?
self.plugin.config[data] = value self.plugin.config[data] = value
# TODO: Apply setting only to specific instance
self._plugin.config['model_instance'].on_setting(data, value)
log.debug('plugin config after:\n %s', self.plugin.config.data) log.debug('plugin config after:\n %s', self.plugin.config.data)
def on_set_model(self, value: Any, data: Any) -> None: def on_config_model(self, model: str, value: Any, data: Any) -> None:
if isinstance(value, str): if isinstance(value, str):
value.strip() value.strip()
log.debug('plugin config before:\n %s', self.plugin.config.data[model])
setattr(self.plugin.config.data[model], data, value)
log.debug('plugin config after:\n %s', self.plugin.config.data[model])
self._plugin.config.data[model].instance.set_config(self.plugin.config.data[model])
def on_set_model(self, model: Any) -> None:
if isinstance(model, str):
model.strip()
log.debug('plugin config before:\n %s', self.plugin.config.data) log.debug('plugin config before:\n %s', self.plugin.config.data)
self._available_models[value]['model_instance'] = \ if (self.plugin.config.data[model].instance is None and
self._available_models[value]['class']() self._available_models[model].klass is not None):
self.plugin.config.data[model].instance = \
self._available_models[model].klass()
else:
return
self.plugin.config['model_class'] = self._available_models[value][ self.plugin.config['model'] = model
'class']
self.plugin.config['model_instance'] = self._available_models[value][
'model_instance']
self.on_setting(value, data)
log.debug('plugin config after:\n %s', self.plugin.config.data) log.debug('plugin config after:\n %s', self.plugin.config.data)
@staticmethod def check_available_moduls(self):
def is_module_available(module: str) -> bool: def is_module_available(module: str) -> bool:
try: try:
__import__(module) __import__(module)
@@ -121,23 +133,21 @@ class Configuration:
log.debug(str(ex)) log.debug(str(ex))
return False return False
def check_available_moduls(self):
for model in SUPPORTED_MODELS: for model in SUPPORTED_MODELS:
available = True available = True
for modul in SUPPORTED_MODELS[model]['moduls']: for modul in SUPPORTED_MODELS[model].required_moduls:
if not self.is_module_available(modul): if not is_module_available(modul):
available = False available = False
continue continue
if available: if available:
self._available_models[model] = SUPPORTED_MODELS[model] self._available_models[model] = SUPPORTED_MODELS[model]
if SUPPORTED_MODELS[model].config is not None:
log.debug('created config for model = %s: %s', model, self._available_models[model])
log.debug('plugin config for model = %s', self.plugin.config[model])
self.plugin.config.data[model].instance = None
self._available_models[model].config = self.plugin.config[model]
if (self.plugin.config.data['model_class'] is None self.on_set_model(self._plugin.config['model'])
and len(self._available_models) > 0):
model = list(self._available_models)[0]
self.on_set_model(model, 'model')
log.debug('Choose first available model!')
else:
log.debug('Available model already chosen!')
log.debug('models = %s', self._available_models) log.debug('models = %s', self._available_models)
@@ -188,6 +198,9 @@ class STTVoiceMessagesConfigDialog(Gtk.ApplicationWindow):
self.show_all() self.show_all()
############################################################################
# General Settings
############################################################################
class STTBehaviour(PreferenceBox): class STTBehaviour(PreferenceBox):
def __init__(self, config_dialog: STTVoiceMessagesConfigDialog) -> None: def __init__(self, config_dialog: STTVoiceMessagesConfigDialog) -> None:
settings = [ settings = [
@@ -206,7 +219,7 @@ class STTVoiceMessagesConfigDialog(Gtk.ApplicationWindow):
models: list[tuple[str, str]] = [] models: list[tuple[str, str]] = []
for key, value in config_dialog.config.available_models.items(): for key, value in config_dialog.config.available_models.items():
models.append( models.append(
(key, str(value['name'])) (key, str(value.name))
) )
settings = [ settings = [
@@ -222,29 +235,37 @@ class STTVoiceMessagesConfigDialog(Gtk.ApplicationWindow):
PreferenceBox.__init__(self, settings) PreferenceBox.__init__(self, settings)
############################################################################
# OpenAI Whisper Settings
############################################################################
class OpenAIWhisperGeneral(PreferenceBox): class OpenAIWhisperGeneral(PreferenceBox):
def __init__(self, config_dialog: STTVoiceMessagesConfigDialog) -> None: def __init__(self, config_dialog: STTVoiceMessagesConfigDialog) -> None:
self._model = 'model_openaiwhisper'
self._config_dialog = config_dialog
settings = [ settings = [
Setting(SettingKind.POPOVER, Setting(SettingKind.POPOVER,
_('Language Model Size'), _('Language Model Size'),
SettingType.VALUE, SettingType.VALUE,
value=config_dialog.plugin.config[ value=config_dialog.config.available_models[self._model].config.model_size,
'whisperai_model_size'], data='model_size',
data='whisperai_model_size', callback=self._set_config,
callback=config_dialog.config.on_setting,
props={'entries': whisper.available_models()}), props={'entries': whisper.available_models()}),
Setting(SettingKind.SWITCH, Setting(SettingKind.SWITCH,
_('Translate'), _('Translate'),
SettingType.VALUE, SettingType.VALUE,
value=config_dialog.plugin.config[ value=config_dialog.config.available_models[self._model].config.translate_to_english,
'whisperai_translate'], data='translate_to_english',
data='whisperai_translate', callback=self._set_config)
callback=config_dialog.config.on_setting)
] ]
PreferenceBox.__init__(self, settings) PreferenceBox.__init__(self, settings)
def _set_config(self, value: Any, data: Any):
self._config_dialog.config.on_config_model(self._model, value, data)
def _add_prefs(self, prefs: list[tuple[str, type[PreferenceBox]]]): def _add_prefs(self, prefs: list[tuple[str, type[PreferenceBox]]]):
for ui_name, klass in prefs: for ui_name, klass in prefs:
pref_box = getattr(self._ui, ui_name) pref_box = getattr(self._ui, ui_name)

3
stt_voice_messages/gtk/sttbox.py
View File

@@ -57,7 +57,8 @@ class STTBox(Gtk.Box):
def _on_transcribe_clicked(self, _button: Gtk.Button) -> None: def _on_transcribe_clicked(self, _button: Gtk.Button) -> None:
log.debug('config.data = %s', self._config.data) log.debug('config.data = %s', self._config.data)
model = self._config.data['model_instance'] model_name = self._config.data['model']
model = self._config.data[model_name].instance
if model is None: if model is None:
return return

4
stt_voice_messages/models/model_settings.py
View File

@@ -19,5 +19,5 @@ from dataclasses import dataclass, field
@dataclass @dataclass
class OpenAIWhisperSettings: class OpenAIWhisperSettings:
whisperai_model_size: str = field(default='tiny', init=True) model_size: str = field(default='tiny', init=True)
translate_to_english: bool = field(default=False, init=True)

6
stt_voice_messages/models/model_template.py
View File

@@ -16,8 +16,6 @@
from abc import ABC, abstractmethod from abc import ABC, abstractmethod
from pathlib import Path from pathlib import Path
from gajim.gtk.const import Setting
from ..helper import Results from ..helper import Results
@@ -26,7 +24,3 @@ class Model(ABC):
@abstractmethod @abstractmethod
def transcribe(self, result: Results, audio_file: Path) -> str: def transcribe(self, result: Results, audio_file: Path) -> str:
return '' return ''
@abstractmethod
def on_setting(self, setting: Setting):
pass

11
stt_voice_messages/models/openai_whisper.py
View File

@@ -46,11 +46,10 @@ class WhisperModel(Model):
return self._result return self._result
def transcribe(self, result: Results, audio_file: Path) -> str: def transcribe(self, result: Results, audio_file: Path) -> str:
model = whisper.load_model(self._config['whisperai_model_size']) model = whisper.load_model(self._config.model_size)
log.debug('model size is used = %s', self._config['whisperai_model_size']) log.debug('model size is used = %s', self._config.model_size)
result.text = model.transcribe(audio_file)['text'] result.text = model.transcribe(audio_file)['text'] # pyright: ignore [reportAttributeAccessIssue]
def on_setting(self, key, value): def set_config(self, config: OpenAIWhisperSettings) -> None:
log.debug('key = %s, value = %s', key, value) self._config = config
self._config[key] = value

20
stt_voice_messages/stt_voice_messages.py
View File

@@ -22,6 +22,7 @@ from gajim.plugins.plugins_i18n import _
from .gtk.config_dialog import * from .gtk.config_dialog import *
from .gtk.sttbox import STTBox from .gtk.sttbox import STTBox
from .models.model_settings import *
log = logging.getLogger('gajim.p.stt_voice_messages') log = logging.getLogger('gajim.p.stt_voice_messages')
@@ -29,6 +30,17 @@ log = logging.getLogger('gajim.p.stt_voice_messages')
class STTVoiceMessagesPlugin(GajimPlugin): class STTVoiceMessagesPlugin(GajimPlugin):
def init(self) -> None: def init(self) -> None:
self.description = _('Transcribes voice messages to text.') self.description = _('Transcribes voice messages to text.')
self.config_default_values = {
'auto_transcribe': (False, ''),
'model': ('model_openaiwhisper', ''),
'model_openaiwhisper': (
OpenAIWhisperSettings(
model_size='tiny',
translate_to_english=False),
'')
}
self._config = Configuration(self) self._config = Configuration(self)
self._config.check_available_moduls() self._config.check_available_moduls()
self.config_dialog = partial(STTVoiceMessagesConfigDialog, self._config) self.config_dialog = partial(STTVoiceMessagesConfigDialog, self._config)
@@ -37,14 +49,6 @@ class STTVoiceMessagesPlugin(GajimPlugin):
'preview_audio': (self._on_preview_audio_created, None), 'preview_audio': (self._on_preview_audio_created, None),
} }
self.config_default_values = {
'auto_transcribe': (False, ''),
'model': ('', ''),
'model_class': (None, ''),
'whisperai_model_size': ('tiny', ''),
'whisperai_translate': (False, ''),
}
self._audio_file: str = '' self._audio_file: str = ''
self._preview_audio_widget = None self._preview_audio_widget = None
self._stt_box = None self._stt_box = None