cstr/auraface-v1-GGUF

AuraFace-v1 GGUF

GGUF format of fal/AuraFace-v1 (Apache 2.0) for CrispEmbed.

512-D ArcFace-compatible face embeddings from ResNet-100. Drop-in replacement for InsightFace buffalo_l — same 512-D embedding dimension, commercially permissive.

Parity

Metric	Value
Cosine vs ONNX Runtime	0.999997
Architecture	ResNet-100 (103 Conv, 49 residual Add, 50 PReLU)
Input	Aligned 112x112 face crop
Output	512-D L2-normalized embedding

Quick Start

License

Apache 2.0 (fal.ai). No non-commercial restrictions. Schema-compatible with InsightFace buffalo_l (same 512-D ArcFace space).

Provenance and EU AI Act Art. 53 note

• Upstream model: fal/AuraFace-v1 — published by fal.
• 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.