For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /get-started.md).

Python function

lora_parameters

lora_parameters()

max.experimental.nn.lora_parameters(model)

source

Yields (qualified_name, slot) for every adapter stack in model.

Two slots per LoRA layer (lora_a then lora_b): a plain Linear’s A/B, or a fused qkv layer’s fused A/B (each stacking the three q/k/v adapters). Serving uses this both to append the adapter input types at compile() and to feed each call’s adapters in order.

Parameters:

model (Module[..., Any])

Return type:

Iterator[tuple[str, _LoRAWeight]]