README.md

metadata

license: apache-2.0
tags:
  - speech
  - audio
  - mos
  - non-verbal-vocalization
  - speech-quality-assessment

NVMOS SPEAR-L9 Scorer

This repository hosts the released NVMOS downstream scorer checkpoint for non-verbal vocalization quality assessment.

Files:

• nvmos_spear_l9.pt: PyTorch state dict for the text-query cross-attention scorer.
• config.json: inference configuration, including upstream encoder model IDs and scorer dimensions.
• training_run_config.json: training-time configuration record.

The full inference code is available at https://github.com/yongaifadian1/NVMOS.