Skip to main content

/

Docs

Nightly

v24.6

Log in

Python module

cache_params

KVCacheParams​

class max.pipelines.kv_cache.cache_params.KVCacheParams(dtype: DType, n_kv_heads: int, head_dim: int, cache_strategy: KVCacheStrategy = KVCacheStrategy.CONTINUOUS, n_devices: int = 1)

dtype_shorthand​

property dtype_shorthand*: str*

The textual representation in shorthand of the dtype.

static_cache_shape​

property static_cache_shape*: tuple[str, str, str, str, str]*

KVCacheStrategy​

class max.pipelines.kv_cache.cache_params.KVCacheStrategy(value, names=None, *, module=None, qualname=None, type=None, start=1, boundary=None)

CONTINUOUS​

CONTINUOUS = 'continuous'

NAIVE​

NAIVE = 'naive'

PAGED​

PAGED = 'paged'

uses_opaque()​

uses_opaque() → bool