Tracks statistics during segment processing. This class maintains counters for various types of segments and errors encountered during t

This class is from OpenAI's Whisper repository. The original version can be found at: https://github.com/openai/whisper/blob/ba3f3cd54b0e5b8c

This class is based on an implementation by OpenAI from the Whisper repository. The original version can be found at: https://github.com/open

This class is based on an implementation by OpenAI from the Whisper repository. The original version can be found at: https://github.com/open

This function is from OpenAI's Whisper repository. The original version can be found at: https://github.com/openai/whisper/blob/ba3f3cd54b0e5

This function is from OpenAI's Whisper repository. The original version can be found at: https://github.com/openai/whisper/blob/ba3f3cd54b0e5

PyTorch Dataset for ASR evaluation across multiple datasets and tasks. This dataset class provides a unified interface for loading and preprocess

Represents a transcript segment with associated metadata. This class encapsulates all information about a single segment including conte

A class to read WebVTT or SRT transcript files and extract transcript data. Attributes: file_path: Path to the transcript file (optional)

This class is based on an implementation by OpenAI from the Whisper repository. The original version can be found at: https://github.com/open

This class is based on an implementation by OpenAI from the Whisper repository. The original version can be found at: https://github.com/open

Represents a machine-generated transcript segment. This class is used for segments created by automated speech recognition systems and c

This class is from OpenAI's Whisper repository. The original version can be found at: https://github.com/openai/whisper/blob/ba3f3cd54b0e5b8c

This class is from OpenAI's Whisper repository. The original version can be found at: https://github.com/openai/whisper/blob/ba3f3cd54b0e5b8c

This class is from OpenAI's Whisper repository. The original version can be found at: https://github.com/openai/whisper/blob/ba3f3cd54b0e5b8c

This class is from OpenAI's Whisper repository. The original version can be found at: https://github.com/openai/whisper/blob/ba3f3cd54b0e5b8c

This class is from OpenAI's Whisper repository. The original version can be found at: https://github.com/openai/whisper/blob/ba3f3cd54b0e5b8c

This class is from OpenAI's Whisper repository. The original version can be found at: https://github.com/openai/whisper/blob/ba3f3cd54b0e5b8c

Dataset for loading audio segments and associated metadata for language identification. This dataset handles loading audio files, preprocess

Handles common audio processing operations for ASR evaluation. This utility class provides static methods for loading, preprocessing, and transfo

Dataset for YODAS (Yet another Open speech Dataset and Speech) training. Wraps HuggingFace datasets for YODAS training with timestamp-aware text

Dataset for audio and transcript segments Attributes: samples: List of sample dictionaries containing audio/transcript info n_tex

Dataset for audio and transcript segments Attributes: samples: List of sample dictionaries containing audio/transcript info n_tex

Thread-safe counter for multiprocessing environments. Uses a Manager to create a shared value that can be safely incremented across mult

This class is based on an implementation by OpenAI from the Whisper repository. Modifications were made to work with the training code.

This class is based on an implementation by OpenAI from the Whisper repository. Modifications were made to work with the training code.

Dataset for validation using HuggingFace speech recognition datasets. Supports multiple validation datasets including TED-LIUM, VoxPopuli, Common

Dataset loader for AMI Meeting Corpus. The AMI corpus consists of 100 hours of meeting recordings captured using multiple microphones. This l

Dataset loader for Artie Bias Corpus. The Artie Bias Corpus is designed to evaluate ASR systems for bias and fairness across different demogr

Represents an audio segment with optional timing information. This data class encapsulates audio file paths along with optional start and end tim

Abstract base class for dataset loaders. Defines the common interface for all dataset loaders in the evaluation pipeline. Each concrete loade

Dataset loader for CORAAL (Corpus of Regional African American Language). CORAAL is a corpus of African American Language featuring audio-aligned

Dataset loader for CORAAL long-form transcription variant. This loader handles the long-form version of the CORAAL corpus, designed for exten

Dataset loader for CallHome English corpus. CallHome consists of conversational telephone speech between family members and friends, represen

Dataset loader for CHiME-6 Challenge dataset. CHiME-6 focuses on distant multi-microphone conversational speech recognition in everyday home

Factory pattern implementation for creating dataset loaders. This factory class provides a centralized way to instantiate appropriate dataset

Dataset loader for FLEURS (Few-shot Learning Evaluation of Universal Representations of Speech). FLEURS is a multilingual speech dataset spanning

Dataset loader for Kincaid46 corpus. Kincaid46 is a specialized evaluation dataset containing 46 utterances designed to test specific aspects

Dataset loader for LibriSpeech test sets. LibriSpeech is a large corpus of approximately 1000 hours of English speech derived from LibriVox a

This class is from OpenAI's Whisper repository. The original version can be found at: https://github.com/openai/whisper/blob/ba3f3cd54b0e5b8c

This class is from OpenAI's Whisper repository. The original version can be found at: https://github.com/openai/whisper/blob/ba3f3cd54b0e5b8c

Dataset loader for Switchboard corpus. Switchboard contains conversational telephone speech between strangers discussing predetermined topics

Handles text cleaning operations for different datasets. This utility class provides dataset-specific text normalization and cleaning operati

Dataset loader for VoxPopuli dataset. VoxPopuli is a large-scale multilingual speech corpus for representation learning and speech recognitio

Dataset loader for Wall Street Journal (WSJ) corpus. The WSJ corpus contains read speech from the Wall Street Journal newspaper, representing

Handles Weights & Biases logging for ASR evaluation experiments. This utility class provides standardized WandB integration for tracking ASR