cstr/siglip-base-GGUF

SigLIP Base GGUF

GGUF format of google/siglip-base-patch16-384 vision encoder for use with CrispEmbed.

768-dimensional image embeddings. Encodes 384x384 images via a 12-layer ViT with 16x16 patches.

Parity

Quant	Cosine vs HF	Size
F32	0.996 (mean pool)	355 MB

Quick Start

# Download
huggingface-cli download cstr/siglip-base-GGUF siglip-base.gguf --local-dir .

# Encode image
./crispembed -m siglip-base.gguf --image photo.jpg

# Print dimension
./crispembed -m siglip-base.gguf --dim  # → 768

Architecture

• Model: SigLIP ViT-B/16 (Google, Apache 2.0)
• Vision: 12 layers, 768-D, 12 heads, 3072 intermediate
• Image: 384×384, 16×16 patches → 576 tokens
• Pooling: Mean pool over patch tokens
• Normalization: L2 normalized output

Provenance and EU AI Act Art. 53 note

• Upstream model: google/siglip-base-patch16-384 — published by google.
• Upstream licence: apache-2.0. This repository redistributes under the same terms; it grants no rights the upstream licence does not.
• What was done here: format conversion and/or quantisation only (GGUF/GGML). No training, no fine-tuning, no merging, no distillation, no change to architecture, vocabulary or capability. Only the numeric representation of the upstream weights differs.
• Training data: documented — where it is documented at all — by the upstream provider; see the upstream model card. No training data was used, added or selected by this repository.
• Provider status: under Regulation (EU) 2024/1689 the upstream authors remain the provider of this model. Converting the serialisation format does not make this repository the provider of a new general-purpose AI model, and no such claim is made. Questions about training content, copyright policy or model capability belong upstream.