phi4-siglip

Model type

phi4-siglip

Class

Phi4SigLIPModel

Task

vision-language

Source

models/phi4_siglip.py

Description

Phi-4-Reasoning-Vision (phi4-siglip) model (3-model split).

Builds three separate ONNX models:

  • decoder: Phi4 text decoder taking inputs_embeds

  • vision_encoder: SigLIP-2 + mlp2x_gelu projector

  • embedding: token embedding + image feature fusion

Supports microsoft/Phi-4-reasoning-vision-15B.

Usage

mobius build --model <MODEL_ID> output_dir/

from mobius import build

model = build("<MODEL_ID>")