RPTools Privacy Filter β€” Multilingual v2 (MLX, 8-bit)

On-device PII / privacy token classifier packaged for the RPTools / RedactFlow Apple-Silicon MLX runtime (rptools-mlx manifest format, openai-privacy-filter architecture family, BIOES-Viterbi decode). Ships the rptools-mlx.json manifest the RPTools Swift loader expects.

Provenance (honest): the weights, tokenizer, config, and label map are a byte-for-byte copy of OpenMed/privacy-filter-multilingual-v2-mlx-8bit (revision df97c293841c59d253541960908e23b2d3a68212). Only the MLX manifest is re-emitted under the RPTools rptools-mlx format string (and source_model_id rebranded); the load-bearing architecture identifier openai-privacy-filter is preserved. Redistributed under the upstream Apache-2.0 license.

Files

  • weights.safetensors β€” 8-bit affine-quantized weights (group size 64)
  • tokenizer.json, tokenizer_config.json β€” tiktoken-style tokenizer
  • config.json, id2label.json β€” model config + 217-label map
  • rptools-mlx.json β€” RPTools MLX artifact manifest
Downloads last month
51
MLX
Hardware compatibility
Log In to add your hardware

Quantized

Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support