iamjanvijay/rnnt_decoder_cuda

An efficient implementation of RNN-T Prefix Beam Search in C++/CUDA.

48
/ 100
Emerging

Optimized prefix beam search decoder for RNN-Transducer models, achieving ~100ms decoding latency for 5-second audio segments at beam size 10. Implements the full transduction algorithm in CUDA-accelerated C++, enabling real-time inference on GPU hardware. Supports configurable beam widths and vocabulary sizes, making it suitable for production speech recognition pipelines requiring low-latency online decoding.

No Package No Dependents
Maintenance 10 / 25
Adoption 8 / 25
Maturity 16 / 25
Community 14 / 25

How are scores calculated?

Stars

67

Forks

10

Language

Cuda

License

MIT

Last pushed

Jan 07, 2026

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/voice-ai/iamjanvijay/rnnt_decoder_cuda"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.