Python class
BlockParameters
BlockParameters
class max.graph.quantization.BlockParameters(elements_per_block, block_size)
Bases: object
Parameters describing the structure of a quantization block.
Block-based quantization stores elements in fixed-size blocks. Each block contains a specific number of elements in a compressed format.
block_size
block_size: int
The size in bytes of the encoded representation of one quantization block.
elements_per_block
elements_per_block: int
The number of original tensor elements grouped into one quantization block.
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!