Expand description
Hugging Face Hub integration and checkpoint packaging (quantization/hub.py parity).
Structs§
- Little
BitModel - LittleBit quantized model wrapper supporting local persistence and Hugging Face Hub upload/download.
- Load
Options - Options controlling how a checkpoint or dense model is loaded from disk or Hub.
- Push
Options - Options controlling Hub publication via
LittleBitModel::push_to_hub.
Functions§
- load_
dense_ model - Load a dense (unquantized) Hugging Face causal LM from a local directory or Hub repo ID.
- resolve_
model_ dir - Resolve a local directory or download a model repository snapshot from Hugging Face Hub.