multidistilbert

Embedded multilingual DistilBERT NER model package for gonnx/models.

Model

• Open: multidistilbert.Open(gonnx.WithThreads(1))
• Model adapter: multidistilbert.NER()
• Source/export: https://huggingface.co/Xenova/distilbert-base-multilingual-cased-ner-hrl
• Source model: https://huggingface.co/Davlan/distilbert-base-multilingual-cased-ner-hrl
• Upstream ONNX revision packaged: c2a4dbf593c57f47004c5bc2d3770d311aee9c43
• Source model revision observed: d421f57d5b1d36b375408588669e9340f9b11a89
• License: AFL-3.0 per source model card at packaging time
• Labels: O, B-DATE, I-DATE, B-PER, I-PER, B-ORG, I-ORG, B-LOC, I-LOC
• Tokenizer: multilingual cased WordPiece (vocab.txt), lower-casing disabled
• Max sequence length: 512 tokens including special tokens
• Embedded ONNX asset: quantized ONNX (assets/model_quantized.onnx)

Checksums

24a0b98f4dd4cd92842f5a541272f86f760225a64a29928eddef14bdb2edb986  assets/model_quantized.onnx
fe0fda7c425b48c516fc8f160d594c8022a0808447475c1a7c6d6479763f310c  assets/vocab.txt
38847be4dc6699b1218a749ed69f888c2ccc7b4deba98e3c4a1cac8cb34d54c8  assets/config.json

Reproduce assets

./prepare_assets.sh

When to use

Use this model when you need multilingual cased NER with generic DATE, PER, ORG, and LOC labels.

Avoid it when you need a small binary, English-only best accuracy, PII-specific labels, or domain-specific labels.