Struct LoraBuildArgs 

pub struct LoraBuildArgs<'a> {
    pub adapters: Vec<LoraAdapterInput<'a>>,
    pub max_lora_rank: usize,
    pub max_loras: usize,
}

LoRA adapter build arguments (--lora-adapter NAME=PATH). None for base-only runs; Some(...) carries the adapter’s separate on-device WeightStore (loaded via spark_runtime::weights::adapter::load_adapter_safetensors), the parsed adapter_config.json, and the pool-shape CLI knobs.

Unlike DFlash (loaded post-construction), the LoRA pool is allocated at the TOP of build_model — before the buffer arena and the free-memory snapshot — so its bytes are automatically debited from the KV budget.

Fields

adapters: Vec<LoraAdapterInput<'a>>

One or more adapters to pack (repeated --lora-adapter NAME=PATH), each carrying its NAME, its on-device WeightStore, and its parsed adapter_config.json. Slot k = adapters[k]. A single element is byte-identical to the pre-multi-adapter path.

max_lora_rank: usize

max_loras: usize