Mila
Deep Neural Network Library
Loading...
Searching...
No Matches
Quantization.ixx File Reference

Umbrella module for the Mila quantization subsystem. More...

import Dnn.Quantization.KvCache.QuantPolicy;
import Dnn.Quantization.KvCache.Policy;
import Dnn.Quantization.Weight.Policies;

Detailed Description

Umbrella module for the Mila quantization subsystem.

Re-exports the full quantization policy surface. Consumers that need all policy types in a single import (model builders, test harnesses) use this module. Components that need only one subsystem should import the targeted submodule directly to keep module dependency graphs tight:

// Linear component -- weight quantization only
import Dnn.Quantization.WeightQuant.Policies;
// GroupedQueryAttention -- KV cache policy only (Alpha.5)
// GroupedQueryAttention -- KV cache quantization (Alpha.6+)

Module dependency graph (no cycles):

Dnn.Quantization
Dnn.Quantization.WeightQuant.Policies
Dnn.Quantization.KvCache.Policy
Dnn.Quantization.KvCache.QuantPolicy