VERSA
VERSADocs

Project statistics

Current metric collection and GitHub activity, with sources and counting methodology.

On this page

Last updated: 2026-09-20 19:27:56 UTC.

Current collection

Statistic Count
Discoverable metrics 96
GitHub stars 436
GitHub forks 53

Metrics by category

Category Metrics
Dependent 17
Distributional 4
Independent 65
Non-match 10

How we count

The metric total is the number of unique canonical names returned by VERSA’s source-based discovery registry for the published checkout. Aliases are excluded. Named Qwen prompt metrics are included separately; a metric returning multiple output scores is counted once. This is a count of discoverable metrics, not installed backends or model variants. Discovery reads source metadata without loading models; optional dependencies are still needed to run many metrics. It recognizes the metadata patterns supported by VERSA’s discovery code.

Stars and forks come from the official repository’s GitHub API at build time. These are dated snapshots, not real-time counters. The website refreshes daily and when relevant code or documentation changes. Scheduled runs can be delayed by GitHub. If collection fails, the previous successful website stays online with its original timestamp.

Sources

Metrics included in this count

Canonical name Category
arecho Independent
asr_match Dependent
asvspoof Independent
audiobox_aesthetics Independent
cdpam_distance Dependent
chroma_alignment Dependent
clap_score Distributional
discrete_speech Dependent
dpam_distance Dependent
emo_vad Independent
emotion Non-match
espnet_wer Non-match
estoi Dependent
fad Distributional
fwhisper_wer Non-match
hubert_wer Non-match
individual_fad Distributional
kid Distributional
lid Independent
log_wmse Dependent
mapss Dependent
mcd_f0 Dependent
multigauss Independent
nemo_wer Non-match
nisqa Independent
nomad Non-match
owsm_wer Non-match
pam Independent
pesq Dependent
pseudo_mos Independent
pysepm Dependent
qwen2_audio_channel_type Independent
qwen2_audio_language Independent
qwen2_audio_laughter_crying Independent
qwen2_audio_overlapping_speech Independent
qwen2_audio_pitch_range Independent
qwen2_audio_recording_quality Independent
qwen2_audio_singing_technique Independent
qwen2_audio_speaker_age Independent
qwen2_audio_speaker_count Independent
qwen2_audio_speaker_gender Independent
qwen2_audio_speaking_style Independent
qwen2_audio_speech_background_environment Independent
qwen2_audio_speech_clarity Independent
qwen2_audio_speech_emotion Independent
qwen2_audio_speech_impairment Independent
qwen2_audio_speech_purpose Independent
qwen2_audio_speech_rate Independent
qwen2_audio_speech_register Independent
qwen2_audio_speech_volume_level Independent
qwen2_audio_vocabulary_complexity Independent
qwen2_audio_voice_pitch Independent
qwen2_audio_voice_type Independent
qwen_omni_channel_type Independent
qwen_omni_language Independent
qwen_omni_laughter_crying Independent
qwen_omni_overlapping_speech Independent
qwen_omni_pitch_range Independent
qwen_omni_recording_quality Independent
qwen_omni_singing_technique Independent
qwen_omni_speaker_age Independent
qwen_omni_speaker_count Independent
qwen_omni_speaker_gender Independent
qwen_omni_speaking_style Independent
qwen_omni_speech_background_environment Independent
qwen_omni_speech_clarity Independent
qwen_omni_speech_emotion Independent
qwen_omni_speech_impairment Independent
qwen_omni_speech_purpose Independent
qwen_omni_speech_rate Independent
qwen_omni_speech_register Independent
qwen_omni_speech_volume_level Independent
qwen_omni_vocabulary_complexity Independent
qwen_omni_voice_pitch Independent
qwen_omni_voice_type Independent
scoreq_nr Independent
scoreq_ref Dependent
se_snr Independent
sheet_ssqa Independent
sigmos Independent
signal_metric Dependent
singer Non-match
songeval Independent
speaker Non-match
speaking_rate Independent
squim_no_ref Independent
squim_ref Dependent
srmr Independent
stoi Dependent
universa Independent
vad Independent
visqol Dependent
vqscore Independent
warpq Dependent
whisper_wer Non-match
wvmos Independent