convert_any_model
Convert a HuggingFace model to AnyModel format.
Functions
Convert a HuggingFace model to AnyModel format. |
- convert_model(input_dir, output_dir, converter)
Convert a HuggingFace model to AnyModel format.
This function converts a HuggingFace checkpoint to the AnyModel format used for compression. The conversion process:
Copies non-weight files (config, tokenizer, etc.)
Creates block_configs for each layer
Reorganizes weights into subblock checkpoints
- Parameters:
input_dir (str) – Path to the input HuggingFace checkpoint directory.
output_dir (str) – Path to the output AnyModel checkpoint directory.
converter (Converter | str) – Either a converter name (e.g., “llama”) or a Converter class.
Example
>>> convert_model( ... input_dir="/path/to/Llama-3.1-8B-Instruct", ... output_dir="/path/to/output/ckpts/teacher", ... converter="llama", ... )