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

Python class

LoRAInputs

`LoRAInputs`

class max.pipelines.lora.LoRAInputs(ids, ranks, grouped_offsets, num_active, end_idx, batch_seq_len, ids_kv, grouped_offsets_kv)

source

Bases: object

Per-batch LoRA buffers passed to a model graph.

Fields are ordered to match the graph’s LoRA input signature. Use buffers() to splice them into a model ABI call.

Parameters:

ids (Buffer)
ranks (Buffer)
grouped_offsets (Buffer)
num_active (Buffer)
end_idx (Buffer)
batch_seq_len (Buffer)
ids_kv (Buffer)
grouped_offsets_kv (Buffer)

`batch_seq_len`

batch_seq_len: Buffer

source

`buffers()`

buffers()

source

Returns the buffers in canonical (field-declaration) order.

Return type:: tuple[Buffer, …]

`end_idx`

end_idx: Buffer

source

`grouped_offsets`

grouped_offsets: Buffer

source

`grouped_offsets_kv`

grouped_offsets_kv: Buffer

source

`ids`

ids: Buffer

source

`ids_kv`

ids_kv: Buffer

source

`num_active`

num_active: Buffer

source

`ranks`

ranks: Buffer

source

LoRAInputs​

batch_seq_len​

buffers()​

end_idx​

grouped_offsets​

grouped_offsets_kv​

ids​

ids_kv​

num_active​

ranks​