Dnn.Components.GemmaBlock Module Reference

Exported Modules
module	Serialization.Mode
module	Dnn.Components.RmsNorm
module	Compute.CpuMemoryResource
module	Compute.IExecutionContext
module	Compute.DeviceId
module	Compute.ExecutionContext
module	Compute.DeviceType
module	Dnn.Components.Gqa
module	Dnn.Components.Residual
module	Compute.ExecutionContextFactory
module	Dnn.Components.Rope
module	Dnn.Tensor
module	Dnn.Components.Linear
module	Dnn.Quantization.Weight.Policies
module	Serialization.ModelArchive
module	Compute.Device
module	Dnn.Components.GemmaConfig
module	Dnn.Component
module	Dnn.TensorTypes
module	Dnn.TensorDataTypeTraits
module	Compute.GqaState
module	Dnn.TensorDataType
module	Serialization.Tensor
module	Dnn.CompositeComponent
module	Dnn.Quantization.KvCache.Policy
module	Dnn.ComponentType
module	Serialization.SafeTensors
module	Dnn.TensorOps
module	Dnn.ITensor
module	Dnn.Components.Swiglu
module	Dnn.ActivationType
module	Dnn.Components.IDecoderLayer
module	Compute.DeviceTypeTraits

Classes
struct	Mila::Dnn::GemmaBlock< TDeviceType, TPrecision, kGlobal, TWeightQuant, TKvPolicy >::BlockBuildContexts
	The per-child build contexts and split-scratch geometry this block implies. More...
class	Mila::Dnn::GemmaBlock< TDeviceType, TPrecision, kGlobal, TWeightQuant, TKvPolicy >
	One Gemma 4 decoder block; kGlobal selects the global (full-attention) geometry. More...
struct	Mila::Dnn::GemmaBlockWorkspace< TDeviceType, TPrecision >
	Transformer-owned shared activation workspace for GemmaBlock (pooling). More...

Files
file	Mila/Src/Dnn/Components/Transformers/Gemma/Gemma.Block.ixx
	Gemma 4 decoder block (inference: prefill + decode), one type per attention kind.