USM-Lite: Quantization and Sparsity Aware Fine-Tuning for Speech Recognition with Universal Speech Models | IEEE Conference Publication | IEEE Xplore