Repository History
3 repositories tagged with deeplearning
NVIDIA NeMo Speech: Scalable Generative AI for Speech Models
NVIDIA NeMo Speech is a powerful, scalable generative AI framework designed for researchers and developers focused on Large Language Models, Multimodal, and Speech AI. It provides tools for Automatic Speech Recognition (ASR) and Text-to-Speech (TTS), enabling efficient creation, customization, and deployment of new AI models using existing code and pre-trained checkpoints. This framework supports a wide range of applications, from real-time streaming ASR to high-quality multilingual TTS.
VoxCPM: Tokenizer-Free TTS for Multilingual Speech, Voice Design, and Cloning
VoxCPM2 is a groundbreaking tokenizer-free Text-to-Speech system, offering highly natural and expressive synthesis across 30 languages. It enables creative voice design from natural language descriptions and provides advanced controllable voice cloning capabilities. With its 2B parameter model, VoxCPM2 delivers 48kHz studio-quality audio, making it a powerful tool for diverse speech generation needs.
PETSA: Parameter-Efficient Test-Time Adaptation for Time Series Forecasting
PETSA offers a parameter-efficient solution for Test-Time Adaptation (TTA) in time series forecasting, addressing the performance degradation caused by non-stationary data. It adapts pre-trained models during inference by updating small calibration modules, reducing memory and compute costs. This method, which includes low-rank adapters, dynamic gating, and a specialized loss, improves forecasting accuracy across diverse backbones and datasets.