herinsoni/gemma3n-e2b

A community-hosted Hugging Face model repository forking or fine-tuning Google's Gemma 3n E2B, an open-weight multimodal model optimized for on-device inference with text, image, audio, and video inputs. The underlying Gemma 3n E2B base model uses selective parameter activation to run efficiently within approximately 2 GB of GPU RAM. Source: supabase.