mmBERT-base ONNX (int8 quantized)

ONNX int8 quantized version of jhu-clsp/mmBERT-base.

Specs

Param Value
Parameters 307M (110M non-embedding)
Hidden size 768
Layers 22
Attention heads 12
Context 8,192 tokens
Languages 1,800+
ONNX size (int8) 294MB
License MIT

Usage

Replaces MiniLM + ModernBERT in the samyx-extract pipeline for:

  • Semantic field name matching ("Invoice No:" → invoice_number)
  • Zero-shot text classification
  • Sentence embeddings (1,800 languages)

Export

Downloads last month
39
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for bluecopa/mmbert-base-onnx

Quantized
(11)
this model