NOTICE

clip-models
Copyright 2026 Akoos <https://akoos.eu>

This repository redistributes a CLIP model as ONNX and adds its own packaging
and documentation. The model keeps its upstream license. The files are float16
ONNX exports produced from the upstream published weights; see SOURCE.md.

--------------------------------------------------------------------------------
clip_image.onnx, clip_text.onnx  (OpenCLIP ViT-B/16, DataComp-1B)
--------------------------------------------------------------------------------
Source:    laion/CLIP-ViT-B-16-DataComp.XL-s13B-b90K
           https://huggingface.co/laion/CLIP-ViT-B-16-DataComp.XL-s13B-b90K
           Trained with OpenCLIP (https://github.com/mlfoundations/open_clip)
           on DataComp-1B (https://github.com/mlfoundations/datacomp).
Copyright: the OpenCLIP authors (Ilharco, Wortsman, Wightman, et al.)
License:   MIT, full text in licenses/openclip-LICENSE.txt
Note:      exported to ONNX (float16) from the upstream weights; see SOURCE.md.
           The CLIP architecture and the byte-level BPE tokenizer vocabulary
           originate from OpenAI CLIP (https://github.com/openai/CLIP), MIT.

--------------------------------------------------------------------------------
Packaging and documentation (this repository's README, LICENSE, NOTICE and
SOURCE.md) are Copyright 2026 Akoos and licensed under the Apache License,
Version 2.0. See LICENSE.