Description
Nemotron-Labs-Audex-2B is a sophisticated large language model (LLM) that integrates audio and text processing capabilities. It is designed to handle a variety of audio tasks, including audio understanding, speech recognition, speech translation, text-to-speech, audio generation, and speech-to-speech generation. The model operates in both thinking and instruct modes, allowing it to adapt to different task requirements.
Read More
The architecture of Audex-2B is compact and dense, which reduces the resource requirements for inference. It is trained using a multi-stage SFT recipe similar to its larger counterpart, Audex-30B-A3B, but without the additional cascaded reinforcement learning. This makes Audex-2B a more resource-efficient option while maintaining strong performance in audio tasks.
Audex-2B supports a context length of up to 128K tokens, enabling it to handle long-context tasks effectively. It uses the ChatML template for processing, with specific prefixes for thinking and instruct modes. The model is compatible with vLLM 0.20.0 and requires transformers version 4.53.0 or higher for inference.
For audio-related tasks, Audex-2B offers different recommended inference setups. It supports offline LLM generation and an OpenAI-compatible audio URL server for audio QA tasks. The model also provides options for text-to-speech and text-to-audio generation, with enhancements available for higher quality outputs.
Nemotron-Labs-Audex-2B is governed by the NVIDIA Oneway Noncommercial License, ensuring its use is restricted to non-commercial applications. This model is ideal for developers and researchers looking for a versatile tool to integrate audio and text processing capabilities into their applications.
- CategoryAI Models & LLMs
- Pricing ModelNoncommercial
- DeploymentCloud
- PlatformsWeb
Nemotron-Labs-Audex-2B Highlights
Unified audio-text processing
Supports thinking and instruct modes
Compact and dense architecture
128K-token context length
Compatible with vLLM 0.20.0
Requires transformers >= 4.53.0
Audio QA and generation capabilities
NVIDIA Oneway Noncommercial License
Getting Started with Nemotron-Labs-Audex-2B
Access model page: Visit the Hugging Face model page for Nemotron-Labs-Audex-2B
Authenticate: Ensure you have the necessary permissions to use the model
Configure: Set up the model with the appropriate inference environment
Send prompts: Use the ChatML template to send audio or text prompts
Fine-tune: Adjust model parameters for specific tasks if needed
Nemotron-Labs-Audex-2B's Use Cases
- Audio Understanding
- Speech Recognition
- Speech Translation
- Text-to-Speech
- Audio Generation
