Same model? and cannot reproduce 6.7%

#2
by fayejf - opened

Looks like the listed three models are identical. Could you confirm it please?
Also I cannot reproduce 6.7% on average. I've just follow the instruction and for loop (so bs=1) on development set which include 1609 samples. But have

error rate (%)
models params 0...5 5...20 average
speechbrain/lang-id-voxlingua107-ecapa 21.1M 11.9 6.7 7.5
TalTechNLP/voxlingua107-epaca-tdnn-ce 21.1M 11.9 6.7 7.5
TalTechNLP/voxlingua107-epaca-tdnn 21.1M 11.9 6.7 7.5

Did I miss anything? Thanks

fayejf changed discussion status to closed
fayejf changed discussion status to open

Sign up or log in to comment