MCR-RNNT Dual-Mode Unified Transducer Training

논문 컨텍스트

Paper title: Reducing the Offline-Streaming Gap for Unified ASR Transducer with Consistency Regularization Abstract: Unification of automatic speech recognition (ASR) systems reduces development and maintenance costs, but training a single model to perform well in both offline and low-latency streaming settings remains challenging. We present a Unified ASR framework for Transducer (RNNT) training that supports both offline and streaming decoding within a single model, using chunk-limited attention with right context and dynamic chunked convolutions. To further close the gap between offline and streaming performance, we introduce an efficient Triton implementation of mode-consistency regularization for RNNT (MCR-RNNT), which encourages agreement across training modes. Experiments show that the proposed approach improves streaming accuracy at low latency while preserving offline performance and scaling to larger model sizes and training datasets. The proposed Unified ASR framework and the English model checkpoint are open-sourced. Passages referencing this figure: RNNT str , \mathcal{L}_{\text{DM}}=\alpha\,\mathcal{L}_{\text{RNNT}}^{\text{off}}+(1-\alpha)\,\mathcal{L}_{\text{RNNT}}^{\text{str}}, (2) where α ∈ [ 0 , 1 ] \alpha\in[0,1] represents the offline mode weight (by analogy with p off p_{\text{off}} in SM). This approach doubles the computational resources per training step compared to SM, but more directly couples the two modes during optimization. Figure 1: Unified Transducer training in dual mode with mode-consistency regularization (MCR-RNNT) loss. Table 1: Comparison of Average WER (%) on Open ASR Leaderboard for offline and streaming inference mode with various latency constraints. Left context was set to 5.6s (70 frames). Latency is defined as the sum of the chunk and the right context size: 2.08s=1.04+1.04, 1.12s=0.56+0.56, 0.56s=0.16