llama_converter

Llama converter for AnyModel compression.

Classes

LlamaConverter

Converter for Llama models to AnyModel format.

class LlamaConverter

Bases: Converter

Converter for Llama models to AnyModel format.

static create_block_configs_from_main_config(config)

Create uniform block configs for all Llama layers.

Llama models have uniform architecture across all layers, so we create the same BlockConfig for each layer.

Parameters:

config (LlamaConfig)

Return type:

List[BlockConfig]