Dnn.Quantization.Weight.Policies Module Reference

Concepts
concept	Mila::Dnn::Quant::Weight::HasCodebookTable
	The format decodes through a per-tensor table, which travels as its own tensor.
concept	Mila::Dnn::Quant::Weight::HasHighBitPlane
	The format spills one bit per element into a second byte-aligned plane, so the primary tensor's kStorageBitsPerElement does not account for the whole code.
concept	Mila::Dnn::Quant::Weight::WeightQuantPolicy

Classes
struct	Mila::Dnn::Quant::Weight::NoWeightQuant
struct	Mila::Dnn::Quant::Weight::PerChannelFp8< TStorage >
struct	Mila::Dnn::Quant::Weight::PerGroupCodebook2< kGroupSize >
struct	Mila::Dnn::Quant::Weight::PerGroupCodebook3< kGroupSize >
struct	Mila::Dnn::Quant::Weight::PerGroupFp4< kGroupSize >
struct	Mila::Dnn::Quant::Weight::PerGroupInt4< kGroupSize >

Files
file	Mila/Src/Dnn/Quantization/Weight/Policies.ixx