Conformer: Convolution-augmented Transformer for Speech Recognition

Benchmark Model Rank Results
human-pose-forecasting-on-human36m-absoluteConformer#2Average MPJPE (mm) @ 1000 ms: 147
speech-recognition-on-aminvidia/stt_en_conformer_ctc_large#36Word Error Rate (WER): 15.2
speech-recognition-on-aminvidia/stt_en_conformer_transducer_small#52Word Error Rate (WER): 19.65
speech-recognition-on-aminvidia/stt_en_conformer_ctc_small#53Word Error Rate (WER): 19.91
speech-recognition-on-ami-cleanednvidia/stt_en_conformer_ctc_large#33Word Error Rate (WER): 13.2
speech-recognition-on-ami-cleanednvidia/stt_en_conformer_transducer_small#54Word Error Rate (WER): 17.19
speech-recognition-on-ami-cleanednvidia/stt_en_conformer_ctc_small#56Word Error Rate (WER): 17.8
speech-recognition-on-earnings-22nvidia/stt_en_conformer_ctc_large#48Word Error Rate (WER): 14.95
speech-recognition-on-earnings-22nvidia/stt_en_conformer_transducer_small#52Word Error Rate (WER): 17.18
speech-recognition-on-earnings-22nvidia/stt_en_conformer_ctc_small#55Word Error Rate (WER): 18.13
speech-recognition-on-gigaspeech-cleanednvidia/stt_en_conformer_ctc_large#48Word Error Rate (WER): 10.18
speech-recognition-on-gigaspeech-cleanednvidia/stt_en_conformer_transducer_small#53Word Error Rate (WER): 12.26
speech-recognition-on-gigaspeech-cleanednvidia/stt_en_conformer_ctc_small#57Word Error Rate (WER): 13.17
speech-recognition-on-gigaspeech-testnvidia/stt_en_conformer_ctc_large#49Word Error Rate (WER): 10.31
speech-recognition-on-gigaspeech-testnvidia/stt_en_conformer_transducer_small#56Word Error Rate (WER): 12.43
speech-recognition-on-gigaspeech-testnvidia/stt_en_conformer_ctc_small#59Word Error Rate (WER): 13.32
speech-recognition-on-librispeech-test-cleannvidia/stt_en_conformer_ctc_large#45Word Error Rate (WER): 1.66
speech-recognition-on-librispeech-test-cleanConformer(L)#63Word Error Rate (WER): 1.9
speech-recognition-on-librispeech-test-cleanConformer(M)#68Word Error Rate (WER): 2
speech-recognition-on-librispeech-test-cleanConformer(S)#76Word Error Rate (WER): 2.1
speech-recognition-on-librispeech-test-cleannvidia/stt_en_conformer_transducer_small#90Word Error Rate (WER): 2.38
speech-recognition-on-librispeech-test-cleannvidia/stt_en_conformer_ctc_small#104Word Error Rate (WER): 3.12
speech-recognition-on-librispeech-test-othernvidia/stt_en_conformer_ctc_large#50Word Error Rate (WER): 3.73
speech-recognition-on-librispeech-test-otherConformer(L)#54Word Error Rate (WER): 3.9
speech-recognition-on-librispeech-test-otherConformer(M)#66Word Error Rate (WER): 4.3
speech-recognition-on-librispeech-test-otherConformer(S)#84Word Error Rate (WER): 5.0
speech-recognition-on-librispeech-test-othernvidia/stt_en_conformer_transducer_small#93Word Error Rate (WER): 6.02
speech-recognition-on-librispeech-test-othernvidia/stt_en_conformer_ctc_small#102Word Error Rate (WER): 7.41
speech-recognition-on-spgispeechnvidia/stt_en_conformer_ctc_large#51Word Error Rate (WER): 4.79
speech-recognition-on-spgispeechnvidia/stt_en_conformer_transducer_small#56Word Error Rate (WER): 5.96
speech-recognition-on-spgispeechnvidia/stt_en_conformer_ctc_small#58Word Error Rate (WER): 7.01
speech-recognition-on-voxpopulinvidia/stt_en_conformer_ctc_large#26Word Error Rate (WER): 6.52
speech-recognition-on-voxpopulinvidia/stt_en_conformer_transducer_small#38Word Error Rate (WER): 7.67
speech-recognition-on-voxpopulinvidia/stt_en_conformer_ctc_small#50Word Error Rate (WER): 8.75
speech-recognition-on-voxpopuli-cleanednvidia/stt_en_conformer_ctc_large#27Word Error Rate (WER): 3.73
speech-recognition-on-voxpopuli-cleanednvidia/stt_en_conformer_transducer_small#43Word Error Rate (WER): 4.29
speech-recognition-on-voxpopuli-cleanednvidia/stt_en_conformer_ctc_small#54Word Error Rate (WER): 4.6