schema_version: 1
id: EIDORA/CLIP_VITB32
name: CLIP_VITB32
version: 1.0.0
model_family: CLIP
backend: onnx
file_size: 352 MB
runtime:
  adapter: onnx_image
  execution_provider: CPUExecutionProvider
  model_path: model.onnx
  input_names:
  - pixel_values
  output_name: embedding
  onnxruntime:
    opset: 17
    tested_versions: '>=1.17,<2'
artifact:
  path: model.onnx
  sha256: 6e778c76fed3af2e98c837c304fa2f85f545b3e35d13854448c248812fcdf533
  package_size_bytes: 351584808
inputs:
- id: image
  modality: image
  label: Images
  required: true
  source_kind: media_source
  requirements:
    color_space: RGB
    layout: NCHW
    width: 224
    height: 224
preprocess:
  image:
    resize_mode:
      label: Image preprocessing
      kind: choice
      choices:
      - distort
      - crop_center
      - add_padding
      default: crop_center
      visible: true
    resize:
      mode: resize_shorter_edge_then_center_crop
      resize_size: 224
      crop_width: 224
      crop_height: 224
      interpolation: bicubic
      source: CLIPImageProcessor for openai/clip-vit-base-patch32
    rescale: 1/255
    normalize:
      mean:
      - 0.48145466
      - 0.4578275
      - 0.40821073
      std:
      - 0.26862954
      - 0.26130258
      - 0.27577711
      inside_onnx: true
embedding:
  dimensions: 512
  feature_type: embedding
  pooling: pooler
  normalized: true
  similarity: cosine
  output_name: embedding
  dtype: float32
  shape:
  - batch
  - 512
display:
  summary: 'Light: fast image embeddings for visual grouping and discovery on laptops.'
  compute_tier: light
  modality_labels:
  - image
  recommended_batch_size: 8
validation:
  fixtures:
  - id: image_tensor_001
    input_shape:
    - 1
    - 3
    - 224
    - 224
    expected_shape:
    - 1
    - 512
    seed: 29
  checks:
    load_with: onnxruntime
    execution_provider: CPUExecutionProvider
    output_dtype: float32
    finite: true
    normalized_l2_range:
    - 0.99
    - 1.01
provenance:
  base_model: openai/clip-vit-base-patch32
  source_repository: https://huggingface.co/openai/clip-vit-base-patch32
  original_model_name: CLIP ViT-B/32
  original_model_url: https://github.com/openai/CLIP
  authors:
  - Alec Radford
  - Jong Wook Kim
  - Chris Hallacy
  - Aditya Ramesh
  - Gabriel Goh
  - Sandhini Agarwal
  - Girish Sastry
  - Amanda Askell
  - Pamela Mishkin
  - Jack Clark
  - Gretchen Krueger
  - Ilya Sutskever
  paper_title: Learning Transferable Visual Models From Natural Language Supervision
  paper_url: https://arxiv.org/abs/2103.00020
  upstream_license: MIT
  training_data: WebImageText-style image/text pairs described by the upstream CLIP
    authors.
  citation: "@inproceedings{radford2021learning,\n  title={Learning Transferable Visual\
    \ Models From Natural Language Supervision},\n  author={Radford, Alec and Kim,\
    \ Jong Wook and Hallacy, Chris and Ramesh, Aditya and Goh, Gabriel and Agarwal,\
    \ Sandhini and Sastry, Girish and Askell, Amanda and Mishkin, Pamela and Clark,\
    \ Jack and Krueger, Gretchen and Sutskever, Ilya},\n  booktitle={International\
    \ Conference on Machine Learning},\n  year={2021}\n}\n"
  conversion_note: EIDORA produced this ONNX conversion and is not the original model
    creator.
  export_date: '2026-07-14'
  exporter_version: eidora-onnx-exporter 0.1.0
model_card:
  best_for:
  - Fast first-pass visual grouping.
  - Image collections where semantic similarity matters more than exact object classification.
  - A broadly useful starter model for EIDORA image workflows.
  not_ideal_for:
  - Fine-grained visual similarity where a heavier model is acceptable.
  - Text-only projects.
  - Specialized domains that need a domain-trained visual encoder.
  limitations: CLIP embeddings can reflect web-scale training data biases and may
    miss fine visual details. Similarity scores should not be used as sole evidence
    for identity, authorship, intent, or sensitive attributes.
  license:
    id: mit
    attribution: Converted to ONNX for EIDORA from the original OpenAI CLIP model.
  huggingface:
    org: eidora
    repo_name: CLIP_VITB32_512
    pipeline_tag: feature-extraction
    tags:
    - eidora
    - eidora-model-zoo
    - onnx
    - onnxruntime
    - embeddings
    - image
    - clip
    - compute:light
    - modality:image
    datasets:
    - openai/webimage-text
    metrics:
    - cosine-similarity