E5_SMALL_384

E5_SMALL_384 is a light text embedding model for fast semantic grouping and search in EIDORA. Choose it when you want a compact text model that runs comfortably on ordinary laptops.

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.

Compute Tier

Light: small download, low memory, faster CPU runtime. Intended for laptop CPU use and large first-pass projects.

Inputs

text: required text input from text_source.

Output

The primary output is embedding, a float32 tensor shaped [batch, 384]. Embeddings are already normalized and are intended for cosine similarity.

Usage In EIDORA

EIDORA shows this package as a light text embedding model in the Model Zoo. Use it for discovery maps, grouping, retrieval, and related embedding workflows.

Preprocessing

text: .

Authorship And Citation

This ONNX package was produced by EIDORA from the original E5 Small v2 model. EIDORA converted the model to ONNX and is not the original model creator. Please cite Text Embeddings by Weakly-Supervised Contrastive Pre-training and the original model repository when using this converted model.

Original model: https://github.com/microsoft/unilm/tree/master/e5

Original paper: https://arxiv.org/abs/2212.03533

Authors: Liang Wang, Nan Yang, Xiaolong Huang, Binxing Jiao, Linjun Yang, Daxin Jiang, Rangan Majumder, Furu Wei

@article{wang2022text,
  title={Text Embeddings by Weakly-Supervised Contrastive Pre-training},
  author={Wang, Liang and Yang, Nan and Huang, Xiaolong and Jiao, Binxing and Yang, Linjun and Jiang, Daxin and Majumder, Rangan and Wei, Furu},
  journal={arXiv preprint arXiv:2212.03533},
  year={2022}
}

Training Data And Provenance

Base model: intfloat/e5-small-v2. Source repository: https://huggingface.co/intfloat/e5-small-v2. Known training data: Weakly supervised text pairs from the E5 training recipe, including CCPairs and supervised fine-tuning data described by the upstream authors. Package payload size: 133782386 bytes.

Evaluation And Validation

The package validation checks that the ONNX graph loads with ONNX Runtime CPU execution, runs the declared fixtures, returns finite float32 embeddings with the declared shape, and matches the artifact hash recorded in config.yaml.

Limitations And Safety

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.