DitCacheSearchConfig

Function

DiT cache search parameter configuration class, which stores configuration parameters for cache search.

Prototype

class DitCacheSearchConfig(cache_ratio=1.3, dit_block_num=None, num_sampling_steps=None)

Parameters

Parameter	Input/Return	Description	Constraints	Instructions
cache_ratio	Input	Speedup ratio	Optional. Data type: float. Default: 1.3 Value range: (1.0, 2.0)	Controls the speedup ratio for cache application. Larger values indicate a more significant expected speedup effect.
dit_block_num	Input	Number of DiT blocks	Optional. Data type: int. Default: None	Usually set automatically by the system; manual specification is not required.
num_sampling_steps	Input	Number of sampling steps	Required. Data type: int. Must be a positive integer	Should match the number of sampling steps during actual inference.

Sample

from msmodelslim.pytorch.multi_modal.dit_cache import DitCacheSearchConfig

# Set the search configuration.
config = DitCacheSearchConfig(
    cache_ratio=1.3,
    num_sampling_steps=100
)

Notes

1. cache_ratio should be set within a reasonable range. Excessively large values may lead to quality degradation.

2. The time complexity of the search process is proportional to num_sampling_steps.