NeMo: a toolkit for building AI applications using Neural Modules

Benchmark Model Rank Results
speech-recognition-on-common-voice-englishcanary-1b-flash#2Test WER: 6.99%
speech-recognition-on-common-voice-englishcanary-1b#4Test WER: 7.97%
speech-recognition-on-common-voice-englishConformerCTC-L#5Test WER: 8.0%
speech-recognition-on-common-voice-frenchcanary-1b-flash#1Test WER: 6.15%
speech-recognition-on-common-voice-frenchConformerCTC-L (4-gram)#3Test WER: 9.16%
speech-recognition-on-common-voice-frenchConformerCTC-L (no-LM)#5Test WER: 9.63%
speech-recognition-on-common-voice-germancanary-1b-flash#4Test WER: 4.09%
speech-recognition-on-common-voice-germanConformerCTC-L (4-gram)#6Test WER: 6.03%
speech-recognition-on-common-voice-germanConformerCTC-L (no LM)#9Test WER: 6.68%
speech-recognition-on-common-voice-spanishcanary-1b-flash#1Test WER: 3.62%
speech-recognition-on-common-voice-spanishConformerCTC-L (4-gram)#2Test WER: 5.5%
speech-recognition-on-common-voice-spanishConformerCTC-L (no LM)#5Test WER: 6.9%