| Predicate — Criterion-Conditional Classifier |
| Copyright 2026 NopeNet LLC |
|
|
| This product is licensed under the Apache License, Version 2.0 (see |
| LICENSE in this repository). |
|
|
| The public OSS distribution ships as the Docker image |
| `ghcr.io/nope-net/predicate-oss:<ver>` and as standalone adapter + |
| head weights on the Hugging Face Hub. Both are governed solely by the |
| Apache License, Version 2.0 (above); the image ships without a |
| license-key gate. |
|
|
| For commercial / on-prem inquiries: sales@nope.net. |
|
|
| |
| Third-party components |
| |
|
|
| This product includes the IBM Granite 4.0 Micro model weights |
| (ibm-granite/granite-4.0-micro, Apache License 2.0, Copyright IBM |
| Corporation), pulled at runtime and adapted by NopeNet-trained LoRA |
| adapters. The LoRA adapter, the MLP classification head, and the |
| isotonic calibration shipped in this Work are licensed under the |
| Apache License, Version 2.0, as part of this Work. |
|
|
| |
| About Predicate's role |
| |
|
|
| Predicate is a criterion-conditional content classifier. Given a |
| natural-language criterion and a piece of content, it returns a |
| calibrated [0, 1] score for how the content relates to that criterion. |
| It is not a predictive, diagnostic, or therapeutic system, and is not a |
| substitute for human judgment. Scores are signals for routing, ranking, |
| and human review; the operator sets thresholds and response policies |
| appropriate to their application. |
|
|
