mlboydaisuke/SigLIP-base-patch16-224-CoreML

SigLIP — Core ML

Zero-Shot Classification, 2023

Zero-shot image classification. Dual encoder (image + text). 224×224 input.

Core ML conversion of google-research/big_vision for on-device inference on iPhone, iPad and Mac. Converted with coremltools; the packages are stateless, so all sequencing and buffering lives in your Swift code.

Task	zero shot image classification
Upstream	google-research/big_vision
Packages	2
Download size	358 MB
Minimum iOS	17.0
Peak RAM	~800 MB

Files

File	Size	Compute units	SHA-256
SigLIP_ImageEncoder.mlpackage.zip	162 MB	cpuOnly	98f6abf5f4aa1451…
SigLIP_TextEncoder.mlpackage.zip	195 MB	cpuOnly	9dead2d58705838a…
siglip_vocab.json	658 KB	-	b94b3a58e04f6199…
Total	358 MB

compute_units is not a suggestion -- it is the configuration the conversion was verified against. Moving a package to a different compute unit can silently change the numerics (FP16 attention overflow) or crash on the GPU.

Download

hf download mlboydaisuke/coreml-zoo --include "siglip/*" --local-dir ./siglip
unzip './siglip/siglip/*.zip' -d ./siglip

Use in Swift

import CoreML

let config = MLModelConfiguration()
config.computeUnits = .cpuOnly   // as converted — see the table above

// Unzip the .mlpackage, drop it into your Xcode target and Xcode compiles it
// at build time:
let model = try SigLIP_ImageEncoder(configuration: config)

// ...or compile a downloaded .mlpackage at runtime:
let compiled = try await MLModel.compileModel(at: mlpackageURL)
let model = try MLModel(contentsOf: compiled, configuration: config)

This model is split into 2 Core ML packages that are driven in sequence from Swift. Load them one at a time, copy the outputs out of the MLMultiArray buffers and release each model before loading the next — two large Core ML models resident at once will OOM on an iPhone.

Demo

• Sample app — sample_apps/SigLIPDemo, a standalone SwiftUI project.
• Models Zoo — this model is downloadable and runnable inside the Models Zoo app on the App Store, no build required.

Conversion

• Script: convert_siglip.py
• Pitfalls hit during conversion (FP16 overflow, ANE buffer limits, stride handling): docs/coreml_conversion_notes.md
• Model index: CoreML-Models

License

The conversion inherits the upstream license: Apache-2.0.

Credits

• Upstream authors: google-research/big_vision, 2023
• Core ML conversion: john-rocky (Daisuke Majima)