| ---
|
| language: en
|
| library_name: transformers
|
| pipeline_tag: text-classification
|
| tags:
|
| - bert
|
| - privacy
|
| - text-classification
|
| - gradio
|
| model-index:
|
| - name: JayNightmare/PrERT-CNM-v4-privacybert
|
| results: []
|
| ---
|
|
|
| # JayNightmare/PrERT-CNM-v4-privacybert
|
|
|
| PrERT-CNM v4 PrivacyBERT is a Transformers sequence-classification model prepared from the local checkpoint at `artifacts\phase-3-privacybert\classifier_checkpoint\privacybert`.
|
|
|
| ## Intended Use
|
|
|
| Use this model for text classification in the privacy/CNM workflow it was trained for. It is intended for research and application prototyping unless your own validation shows it is suitable for production use.
|
|
|
| ## Labels
|
|
|
| - `user`
|
| - `system`
|
| - `organization`
|
|
|
| ## Usage
|
|
|
| ```python
|
| from transformers import pipeline
|
|
|
| classifier = pipeline("text-classification", model="JayNightmare/PrERT-CNM-v4-privacybert", top_k=None)
|
| scores = classifier("Paste text to classify.")
|
| print(scores)
|
| ```
|
|
|
| ## Training Details
|
|
|
| - Base architecture: BERT-compatible sequence classifier
|
| - Source checkpoint: `artifacts\phase-3-privacybert\classifier_checkpoint\privacybert`
|
| - Training metadata: included when available in the checkpoint folder
|
|
|
| ## Evaluation
|
|
|
| Add the final held-out metrics before publishing if they are available. Include dataset split details, label distribution, and any thresholding used by downstream consumers.
|
|
|
| ## Limitations
|
|
|
| The model can be sensitive to domain shift, ambiguous language, long inputs, and label definitions that differ from the training data. Review outputs before using them in automated decisions.
|
|
|
| ## Gradio Demo
|
|
|
| The companion Space can be prepared from `huggingface/space` and pointed at this model with the `MODEL_ID` environment variable.
|
|
|
