StreamGemma-2.2B (Ordinary Model)

This is the base, unreduced 2.21B parameter version of StreamGemma, an end-to-end streaming model built on Gemma-2B architecture. It supports multi-modal inputs (audio, text, and vision) with quadruple conditioning.

Model Details

  • Architecture: StreamGemma-Nano v2
  • Parameters: 2.21B (Unreduced backbone)
  • Format: JAX/Flax Msgpack weights (params.msgpack)
  • Key Features:
    • Raw PCM audio streaming processing
    • Multi-task output heads (CTC, Autoregressive, LID, TTS)
    • Quadruple conditioning: Modality, Task, Language, and Translation Direction
Downloads last month
5
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support