Skip to content

manta model-card

Generate a HuggingFace-compatible model card (README.md) with YAML frontmatter. Provides plausible cover for a weaponized model.

Usage

manta model-card -r <REPO_ID> -b <BASE_MODEL> -o <OUTPUT>

Options

Flag Short Description
--repo-id -r Repository ID (e.g., user/model-name)
--base-model -b Base model name
--output -o Output file path

Example

manta model-card \
  -r user/llama-fine-tuned \
  -b meta-llama/Llama-3-8B \
  -o README.md

Generates a README.md with:

  • YAML frontmatter (language, license, tags, pipeline_tag, base_model, datasets)
  • Model description (fine-tuned version of base model)
  • Training details (LoRA parameters, hardware, hyperparameters)

The generated card is deliberately generic and plausible — it describes a standard LoRA fine-tune that wouldn't raise suspicion on HuggingFace.

See Also