config.yaml

schema_version: 1
id: EIDORA/E5_small
name: E5_small
version: 1.0.0
model_family: E5
backend: onnx
file_size: 134 MB
runtime:
  adapter: onnx_text
  execution_provider: CPUExecutionProvider
  model_path: model.onnx
  input_names:
  - input_ids
  - attention_mask
  output_name: embedding
  onnxruntime:
    opset: 17
    tested_versions: '>=1.17,<2'
artifact:
  path: model.onnx
  sha256: 67c0e5500336f67e22d59f942604308aca6af009ebc91ece04c649f86e1d2961
  package_size_bytes: 133782386
inputs:
- id: text
  modality: text
  label: Text
  required: true
  source_kind: text_source
  requirements:
    max_tokens: 512
    recommended_prefixes:
    - 'query: '
    - 'passage: '
preprocess:
  text:
    tokenizer:
      path: tokenizer
      max_length: 512
      truncation: true
      padding: max_length
    prefix_policy:
      query: 'query: '
      document: 'passage: '
embedding:
  dimensions: 384
  feature_type: embedding
  pooling: mean
  normalized: true
  similarity: cosine
  output_name: embedding
  dtype: float32
  shape:
  - batch
  - 384
display:
  summary: 'Light: compact text embeddings for fast search, grouping, and discovery
    on laptops.'
  compute_tier: light
  modality_labels:
  - text
  recommended_batch_size: 16
validation:
  fixtures:
  - id: text_tokens_001
    input_shapes:
      input_ids:
      - 1
      - 16
      attention_mask:
      - 1
      - 16
    input_dtypes:
      input_ids: int64
      attention_mask: int64
    expected_shape:
    - 1
    - 384
    seed: 23
  checks:
    load_with: onnxruntime
    execution_provider: CPUExecutionProvider
    output_dtype: float32
    finite: true
    normalized_l2_range:
    - 0.99
    - 1.01
provenance:
  base_model: intfloat/e5-small-v2
  source_repository: https://huggingface.co/intfloat/e5-small-v2
  original_model_name: E5 Small v2
  original_model_url: https://github.com/microsoft/unilm/tree/master/e5
  authors:
  - Liang Wang
  - Nan Yang
  - Xiaolong Huang
  - Binxing Jiao
  - Linjun Yang
  - Daxin Jiang
  - Rangan Majumder
  - Furu Wei
  paper_title: Text Embeddings by Weakly-Supervised Contrastive Pre-training
  paper_url: https://arxiv.org/abs/2212.03533
  upstream_license: MIT
  training_data: Weakly supervised text pairs from the E5 training recipe, including
    CCPairs and supervised fine-tuning data described by the upstream authors.
  citation: "@article{wang2022text,\n  title={Text Embeddings by Weakly-Supervised\
    \ Contrastive Pre-training},\n  author={Wang, Liang and Yang, Nan and Huang, Xiaolong\
    \ and Jiao, Binxing and Yang, Linjun and Jiang, Daxin and Majumder, Rangan and\
    \ Wei, Furu},\n  journal={arXiv preprint arXiv:2212.03533},\n  year={2022}\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 grouping of text notes, captions, and metadata.
  - Semantic search over medium and large text projects on laptops.
  - A compact starter model for EIDORA text embedding workflows.
  not_ideal_for:
  - Long-document reasoning or generation.
  - Fine-grained domain retrieval where a larger text embedding model is acceptable.
  - Image, video, or audio inputs.
  limitations: E5 embeddings can reflect the biases and language coverage of the upstream
    training data. The upstream recipe recommends query and passage prefixes; quality
    may drop if text is passed without the expected prefix style.
  license:
    id: mit
    attribution: Converted to ONNX for EIDORA from the upstream intfloat E5 small
      v2 model.
  huggingface:
    org: eidora
    repo_name: E5_SMALL_384
    pipeline_tag: feature-extraction
    tags:
    - eidora
    - eidora-model-zoo
    - onnx
    - onnxruntime
    - embeddings
    - text
    - e5
    - compute:light
    - modality:text
    datasets:
    - intfloat/e5
    metrics:
    - cosine-similarity
tokenizer:
  path: tokenizer
  source: intfloat/e5-small-v2