QuartzNet: Deep Automatic Speech Recognition with 1D Time-Channel Separable Convolutions

Benchmark Model Rank Results
human-pose-forecasting-on-human36m-absoluteQuartznet#10Average MPJPE (mm) @ 1000 ms: 168
speech-recognition-on-librispeech-test-cleanQuartzNet15x5#94Word Error Rate (WER): 2.69
speech-recognition-on-librispeech-test-otherQuartzNet15x5#101Word Error Rate (WER): 7.25