Download-metrics QA toy classifier
This is a deliberately tiny synthetic logistic classifier, published solely as an isolated artifact for a bounded download-statistics QA check. It is not a useful pretrained product, a security detector, or an adoption claim.
What it does
The model was trained with full-batch gradient descent on eight one-dimensional synthetic examples to distinguish negative from positive numbers. Four held-out synthetic examples are included in training.json. This toy check does not establish real-world model quality.
Files
- config.json: custom model metadata.
- model.safetensors: two trained float32 parameters, weight and bias.
- inference.mjs: local Node.js inference for this custom model.
- training.json: synthetic training and held-out examples.
This custom format is not compatible with Transformers AutoModel. To run it locally, place model.safetensors next to inference.mjs and run node inference.mjs 0.75.
Analytics caveat
The owner authorized at most two deliberate test requests as part of the initial QA check. Public download totals may include those test requests, platform processing, and other automation. They must not be described as unique users or organic adoption.
No third-party training data, personal information, credentials, or private project notes are included.