Ahmadzei's picture
added 3 more tables for large emb model
5fa1a76
On the CommonVoice benchmark, XLSR shows a relative phoneme error rate reduction
of 72% compared to the best known results.