Retrain on microsoft/unixcoder-base

9e8f827 verified 5 days ago

972 Bytes

license: apache-2.0
base_model: microsoft/unixcoder-base
library_name: transformers.js
pipeline_tag: token-classification
tags:
  - onnx
  - token-classification
  - code

Code identifier classifier

Token classifier that tells, for every identifier in a code snippet, whether it is the project's own name (OWN) or a library/framework name (LIB) (BIO tags B-OWN I-OWN B-LIB I-LIB O). Used by the Redacto browser extension to rename a pasted snippet's own identifiers and keep library names readable.

Base model: microsoft/unixcoder-base (Apache-2.0)
Format: int8-quantized ONNX (onnx/model_quantized.onnx)
Training pipeline: https://github.com/BenKoncsik/pii-guardrail-browser-extension/tree/main/tools/identifier-classifier
Labels: derived automatically from open-source C# and TypeScript projects (Roslyn / TypeScript checker resolution), split by repository.