gemma3

Model type

gemma3

Class

Gemma3MultiModalModel

Task

gemma3-vision-language

Source

models/gemma3.py

Description

Gemma 3 vision-language model (3-model split).

Builds three separate ONNX models:

  • decoder: Gemma3 text decoder taking inputs_embeds

  • vision_encoder: SigLIP + Gemma3MultiModalProjector

  • embedding: token embedding + image feature fusion

Usage

mobius build --model <MODEL_ID> output_dir/

from mobius import build

model = build("<MODEL_ID>")