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

DenoisingCacheState

`DenoisingCacheState`

class max.pipelines.diffusion.DenoisingCacheState(prev_residual=None, prev_output=None, taylor_factor_0=None, taylor_factor_1=None, taylor_factor_2=None, taylor_last_compute_step=None)

source

Bases: object

Per-request mutable cache state for a single denoising stream.

One instance per stream. Created fresh per execute() call.

Parameters:

prev_residual (Tensor | None)
prev_output (Tensor | None)
taylor_factor_0 (Tensor | None)
taylor_factor_1 (Tensor | None)
taylor_factor_2 (Tensor | None)
taylor_last_compute_step (int | None)

`prev_output`

prev_output: Tensor | None = None

source

`prev_residual`

prev_residual: Tensor | None = None

source

`taylor_factor_0`

taylor_factor_0: Tensor | None = None

source

`taylor_factor_1`

taylor_factor_1: Tensor | None = None

source

`taylor_factor_2`

taylor_factor_2: Tensor | None = None

source

`taylor_last_compute_step`

taylor_last_compute_step: int | None = None

source

DenoisingCacheState​

prev_output​

prev_residual​

taylor_factor_0​

taylor_factor_1​

taylor_factor_2​

taylor_last_compute_step​

`DenoisingCacheState`

`prev_output`

`prev_residual`

`taylor_factor_0`

`taylor_factor_1`

`taylor_factor_2`

`taylor_last_compute_step`