One-pass Multiple Conformer and Foundation Speech Systems Compression and Quantization Using An All-in-one Neural Model
We propose a novel one-pass multiple ASR systems joint compression and quantization
approach using an all-in-one neural model. A single compression cycle allows multiple …
approach using an all-in-one neural model. A single compression cycle allows multiple …
AdaKD: Dynamic Knowledge Distillation of ASR models using Adaptive Loss Weighting
S Ganguly, R Nayak, R Rao, U Deb, P AP - arXiv preprint arXiv …, 2024 - arxiv.org
Knowledge distillation, a widely used model compression technique, works on the basis of
transferring knowledge from a cumbersome teacher model to a lightweight student model …
transferring knowledge from a cumbersome teacher model to a lightweight student model …
Simul-Whisper: Attention-Guided Streaming Whisper with Truncation Detection
As a robust and large-scale multilingual speech recognition model, Whisper has
demonstrated impressive results in many low-resource and out-of-distribution scenarios …
demonstrated impressive results in many low-resource and out-of-distribution scenarios …
[PDF][PDF] Knowledge Distillation for End-to-End ASR in Resource-Constrained Environments
P Martin - paulsbitsandbytes.com
This dissertation explores knowledge distillation techniques for end-to-end Automatic
Speech Recognition (ASR) models in resource-constrained environments. The aim is to …
Speech Recognition (ASR) models in resource-constrained environments. The aim is to …