SabiQG/siglip-go-onnx

SigLIP ONNX (Go)

Pre-exported ONNX models for siglip-go — zero-shot image classification in Go.

Based on google/siglip-base-patch16-256-multilingual.

Files

File	Size	Description
vision_model.onnx + .data	~356MB	Image encoder
text_model.onnx + .data	~1.0GB	Text encoder
vocab.json	~18MB	SentencePiece vocabulary (250K tokens)

Usage

These models are auto-downloaded by the Go package:

go get github.com/SabiQG/siglip-go

import siglip "github.com/SabiQG/siglip-go"

c, _ := siglip.New()

// Zero-shot classification
results, _ := c.Classify("photo.jpg", []string{"a dog", "a cat"})

// Get raw image embedding (768-dim)
embedding, _ := c.ImageEmbedding("photo.jpg")

Export

Generated with:

python export_siglip.py   # ONNX models (opset 18)
python export_vocab.py    # vocabulary