Skip to content
View as Markdown

tinker_cookbook.weights.ModelCardConfig

Generated from tinker-cookbook at commit 4a3a282. Source links point at that snapshot.

class tinker_cookbook.weights.ModelCardConfig()

Configuration for auto-generating a HuggingFace model card.

Contains only standard HuggingFace metadata fields. List fields default to empty (omitted from the card when empty). Additional tags are appended to the auto-generated defaults (tinker, tinker-cookbook, and peft/lora for adapters).

config = ModelCardConfig(
base_model="Qwen/Qwen3-8B",
datasets=["my-org/my-dataset"],
tags=["sft"],
license="apache-2.0",
)

Fields:

  • base_model (str) – HuggingFace model ID used as the base (e.g. "Qwen/Qwen3-8B").
  • datasets (list[str], default: field(default_factory=list)) – HuggingFace dataset IDs used for training.
  • tags (list[str], default: field(default_factory=list)) – Additional tags to include (appended to auto-generated defaults).
  • license (str | None, default: None) – SPDX license identifier (e.g. "apache-2.0").
  • language (list[str], default: field(default_factory=list)) – Language codes (e.g. ["en"]).

Referenced by