cjm-transcription-plugin-voxtral-hf 0.0.7

Mistral Voxtral plugin for the cjm-transcription-plugin-system library - provides local speech-to-text transcription through 🤗 Transformers with configurable model selection and parameter control.

cjm-transcription-plugin-voxtral-hf

Install

pip install cjm_transcription_plugin_voxtral_hf

Project Structure

nbs/
└── plugin.ipynb # Plugin implementation for Mistral Voxtral transcription through Hugging Face Transformers

Total: 1 notebook

Module Dependencies

graph LR
    plugin[plugin<br/>Voxtral HF Plugin]

No cross-module dependencies detected.

CLI Reference

No CLI commands found in this project.

Module Overview

Detailed documentation for each module in the project:

Voxtral HF Plugin (plugin.ipynb)

Plugin implementation for Mistral Voxtral transcription through Hugging Face Transformers

Import

from cjm_transcription_plugin_voxtral_hf.plugin import (
    VoxtralHFPluginConfig,
    VoxtralHFPlugin
)

Functions

@patch
def supports_streaming(
    self:VoxtralHFPlugin
) -> bool:  # True if streaming is supported
    "Check if this plugin supports streaming transcription."

@patch
def execute_stream(
    self:VoxtralHFPlugin,
    audio: Union[AudioData, str, Path],  # Audio data or path to audio file
    **kwargs  # Additional plugin-specific parameters
) -> Generator[str, None, TranscriptionResult]:  # Yields text chunks, returns final result
    "Stream transcription results chunk by chunk."

Classes

@dataclass
class VoxtralHFPluginConfig:
    "Configuration for Voxtral HF transcription plugin."
    
    model_id: str = field(...)
    device: str = field(...)
    dtype: str = field(...)
    language: Optional[str] = field(...)
    max_new_tokens: int = field(...)
    do_sample: bool = field(...)
    temperature: float = field(...)
    top_p: float = field(...)
    streaming: bool = field(...)
    trust_remote_code: bool = field(...)
    cache_dir: Optional[str] = field(...)
    compile_model: bool = field(...)
    load_in_8bit: bool = field(...)
    load_in_4bit: bool = field(...)

class VoxtralHFPlugin:
    def __init__(self):
        """Initialize the Voxtral HF plugin with default configuration."""
        self.logger = logging.getLogger(f"{__name__}.{type(self).__name__}")
        self.config: VoxtralHFPluginConfig = None
    "Mistral Voxtral transcription plugin via Hugging Face Transformers."
    
    def __init__(self):
            """Initialize the Voxtral HF plugin with default configuration."""
            self.logger = logging.getLogger(f"{__name__}.{type(self).__name__}")
            self.config: VoxtralHFPluginConfig = None
        "Initialize the Voxtral HF plugin with default configuration."
    
    def name(
            self
        ) -> str:  # Plugin name identifier
        "Get the plugin name identifier."
    
    def version(
            self
        ) -> str:  # Plugin version string
        "Get the plugin version string."
    
    def supported_formats(
            self
        ) -> List[str]:  # List of supported audio formats
        "Get the list of supported audio file formats."
    
    def get_current_config(
            self
        ) -> VoxtralHFPluginConfig:  # Current configuration dataclass
        "Return current configuration."
    
    def initialize(
            self,
            config: Optional[Any] = None  # Configuration dataclass, dict, or None
        ) -> None
        "Initialize the plugin with configuration."
    
    def execute(
            self,
            audio: Union[AudioData, str, Path],  # Audio data or path to audio file to transcribe
            **kwargs  # Additional arguments to override config
        ) -> TranscriptionResult:  # Transcription result with text and metadata
        "Transcribe audio using Voxtral."
    
    def is_available(
            self
        ) -> bool:  # True if Voxtral and its dependencies are available
        "Check if Voxtral is available."
    
    def cleanup(
            self
        ) -> None
        "Clean up resources with aggressive memory management."