Fix reference-word counts in WER computation - #197
Merged
Merged
Conversation
Compared with `master`, this release includes 479 commits and 2,461 changed files. ## Highlights - Modernized package structure separating NumPy and PyTorch functionality. - Added PyTorch support for: - Waveform and pretrained-model x-vector extraction and training. - Wav2Vec2, WavLM, HuBERT, Whisper, and Wav2Vec2-BERT integrations. - Q-vector models and fine-tuning. - RNN-T and transducer training, decoding, and fine-tuning. - ConvNeXt, Conformer, Transformer, ResNet, and SpineNet architectures. - DINO self-supervised x-vector training. - DAC, streaming DAC, FreeVC, VAE/VQ-VAE, and voice anonymization. - Audio discriminators and speech-quality models. - Added improved trainers with distributed training, AMP, compilation, SWA, EMA, checkpointing, and multi-GPU diagnostics. - Added Hugging Face integrations, feature fusion, LoRA, and pretrained encoder fine-tuning. - Added tokenizer implementations and tokenizer-training utilities. - Added speech-quality metrics including SNR, LSD, PESQ, STOI, DNSMOS, and UTMOS. - Added verification, calibration, DET/DCF plotting, anonymization, VoxProfile, and adversarial-evaluation tools. - Added new dataset abstractions and preparation tools, including: - `HyperDataset` - Dataset cleaning - Trial/cohort splitting - Fold generation - Improved recording, segment, and metadata handling - Added data preparation support for LibriSpeech, LibriTTS, LibriTTS-R, MLS, Common Voice, VoxCeleb, SRE16/18/19/21/24, ASVspoof, DIHARD, JANUS, Voices Challenge, and other speech datasets. - Added and updated recipes for VoxCeleb, LibriSpeech ASR, LRE22, SRE24, and ASVspoof 2024. - Added `hyperion-submit` for Slurm and cluster
Update README build status badge
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
WER incorrectly included insertions in the reference-word count. Count reference tokens directly for corpus and per-utterance statistics, and exclude alignment placeholders ("*") from per-word reference totals.
Regression coverage checks error counts, corpus and per-utterance WER, and per-word reference totals.
Validation:
pytest tests/hyperion/metrics/test_wer.py