USM-Lite: Quantization and Sparsity Aware Fine-tuning for Speech Recognition with Universal Speech Models

Documenti analoghi

• USM RNN-T model weights binarization
  di: Rybakov, Oleg, et al.
  Pubblicazione: (2024)
• Extreme Encoder Output Frame Rate Reduction: Improving Computational Latencies of Large End-to-End Models
  di: Prabhavalkar, Rohit, et al.
  Pubblicazione: (2024)
• How to Estimate Model Transferability of Pre-Trained Speech Models?
  di: Chen, Zih-Ching, et al.
  Pubblicazione: (2023)
• Text Injection for Neural Contextual Biasing
  di: Meng, Zhong, et al.
  Pubblicazione: (2024)
• SLiM: One-shot Quantization and Sparsity with Low-rank Approximation for LLM Weight Compression
  di: Mozaffari, Mohammad, et al.
  Pubblicazione: (2024)