Download evaluation/summary.csv from Qooladuntoye/fulfulde_tts_model: direct link, hf CLI and curl.
- Browser
- Download file 1.3 kB
-
https://huggingface.co/Qooladuntoye/fulfulde_tts_model/resolve/main/evaluation/summary.csv
- Command line
-
hf download hf://Qooladuntoye/fulfulde_tts_model/evaluation/summary.csv
-
curl -L -o summary.csv https://huggingface.co/Qooladuntoye/fulfulde_tts_model/resolve/main/evaluation/summary.csv
1.3 kB
| metric,key,description,direction,ft_mean,ft_std,base_mean,base_std,delta_ft_minus_base,n_samples | |
| MCD,mcd,Mel Cepstral Distortion,lower,230.0936409006308,71.800570911441,405.95989017362336,48.62109450182498,-175.86624927299258,50 | |
| Mel-L1,mel_l1,Log-mel L1,lower,1.6148298025131225,0.4707240161299903,2.981115484237671,0.6291589814260485,-1.3662856817245483,50 | |
| F0 RMSE (Hz),f0_rmse,Pitch RMSE,lower,80.822032081945,37.846587080920465,90.84999789930414,20.823659162816853,-10.027965817359146,50 | |
| STOI,stoi,Intelligibility (0-1),higher,0.2003183199785165,0.08715064051212826,0.20831588186503514,0.08601140543666418,-0.007997561886518623,50 | |
| PESQ,pesq,Perceptual quality,higher,1.148063564300537,0.2790061766168384,1.0619287085533142,0.06629712754457757,0.0861348557472228,50 | |
| Duration ratio,dur_ratio,Gen/ref duration,target=1,0.923657841122914,0.3749946479199864,0.6925102270025348,0.22976491805818186,0.2311476141203792,50 | |
| Speaker sim,spk_sim,Voice similarity (0-1),higher,0.6994895219802857,0.053617867085029566,0.5819840049743652,0.050223439481269797,0.11750551700592049,50 | |
| # NOTE: Both models evaluated on 50 randomly-sampled TRAINING utterances. | |
| # Fine-tuned model saw these during training -> metrics optimistically biased. | |
| # MCD computed on dB-scale MFCCs; STOI/PESQ with time-stretch alignment. | |