Here is a list of all variables with links to the classes they belong to:

- k -

• k_ : Mila::Dnn::Compute::CpuAttentionOp, Mila::Dnn::Compute::Cuda::Gqa::CudaGqaOp< TPrecision >, Mila::Dnn::Compute::Cuda::MultiHeadAttention::CudaMultiHeadAttentionOp< TPrecision >, Mila::Dnn::LlamaBlock< TDeviceType, TPrecision, TWeightQuant, TKvPolicy >
• k_exp_ : Mila::Dnn::Compute::Cuda::Gqa::CudaGqaOp< TPrecision >
• k_exp_tensor_ : Mila::Dnn::Compute::Cuda::Gqa::CudaGqaOp< TPrecision >
• k_opt_ : Mila::Dnn::Compute::Cuda::Gqa::CudaGqaOp< TPrecision >
• k_prefill_shape_ : Mila::Dnn::LlamaBlock< TDeviceType, TPrecision, TWeightQuant, TKvPolicy >
• k_proj : Mila::Dnn::Visualization::VisualizerContext
• k_shape_ : Mila::Dnn::LlamaBlock< TDeviceType, TPrecision, TWeightQuant, TKvPolicy >, Mila::Dnn::Rope< TDeviceType, TPrecision >
• k_tensor_ : Mila::Dnn::Compute::CpuAttentionOp, Mila::Dnn::Compute::Cuda::Gqa::CudaGqaOp< TPrecision >, Mila::Dnn::Compute::Cuda::MultiHeadAttention::CudaMultiHeadAttentionOp< TPrecision >
• kAbsoluteRoots : Mila::Dnn::Serialization::ModelArchive
• kCacheDtype : Mila::Dnn::GroupedQueryAttention< TDeviceType, TComputePrecision, TKvPolicy >
• kCublasLtWorkspaceSize : Mila::Dnn::Compute::ExecutionContext< DeviceType::Cuda >
• kIsActive : Mila::Dnn::Quant::KvCache::NoKvCompression, Mila::Dnn::Quant::KvCache::PerChannelKvFp8< TStorage >
• kIsFp4E2M1 : Mila::Dnn::Quant::Weight::PerGroupFp4< kGroupSize >, Mila::Dnn::Quant::Weight::PerGroupInt4< kGroupSize >
• kIsPerChannelQuantized : Mila::Dnn::Compute::Cuda::Linear::CudaLinearOp< TComputePrecision, TWeightQuant >
• kIsPerGroupQuantized : Mila::Dnn::Compute::Cuda::Linear::CudaLinearOp< TComputePrecision, TWeightQuant >
• kIsQuantized : Mila::Dnn::Compute::Cuda::CublasLtLinearPlan< TComputePrecision, TParameterPrecision >, Mila::Dnn::Compute::Cuda::Linear::CudaLinearOp< TComputePrecision, TWeightQuant >, Mila::Dnn::Linear< TDeviceType, TComputePrecision, TWeightQuant >, Mila::Dnn::Quant::Weight::NoWeightQuant, Mila::Dnn::Quant::Weight::PerChannelFp8< TStorage >, Mila::Dnn::Quant::Weight::PerGroupFp4< kGroupSize >, Mila::Dnn::Quant::Weight::PerGroupInt4< kGroupSize >
• kKvCompressed : Mila::Dnn::GroupedQueryAttention< TDeviceType, TComputePrecision, TKvPolicy >
• kPerChannel : Mila::Dnn::Quant::Weight::NoWeightQuant, Mila::Dnn::Quant::Weight::PerChannelFp8< TStorage >, Mila::Dnn::Quant::Weight::PerGroupFp4< kGroupSize >, Mila::Dnn::Quant::Weight::PerGroupInt4< kGroupSize >
• kPerHeadPerToken : Mila::Dnn::Quant::KvCache::PerChannelKvFp8< TStorage >
• kQuantizationGroupSize : Mila::Dnn::Quant::Weight::PerGroupFp4< kGroupSize >, Mila::Dnn::Quant::Weight::PerGroupInt4< kGroupSize >
• kScaleDtype : Mila::Dnn::Quant::KvCache::PerChannelKvFp8< TStorage >, Mila::Dnn::Quant::Weight::NoWeightQuant, Mila::Dnn::Quant::Weight::PerChannelFp8< TStorage >, Mila::Dnn::Quant::Weight::PerGroupFp4< kGroupSize >, Mila::Dnn::Quant::Weight::PerGroupInt4< kGroupSize >
• kStorageDtype : Mila::Dnn::Quant::KvCache::PerChannelKvFp8< TStorage >, Mila::Dnn::Quant::Weight::NoWeightQuant, Mila::Dnn::Quant::Weight::PerChannelFp8< TStorage >, Mila::Dnn::Quant::Weight::PerGroupFp4< kGroupSize >, Mila::Dnn::Quant::Weight::PerGroupInt4< kGroupSize >
• kSymmetric : Mila::Dnn::Quant::KvCache::PerChannelKvFp8< TStorage >
• kUseW8A16Gemm : Mila::Dnn::Compute::Cuda::Linear::CudaLinearOp< TComputePrecision, TWeightQuant >
• kv_cache_compression_ : Mila::Dnn::LanguageModelConfig< TDerived >
• kv_cache_enabled_ : Mila::Dnn::Compute::Cuda::Gqa::CudaGqaOp< TPrecision >, Mila::Dnn::Compute::Cuda::MultiHeadAttention::CudaMultiHeadAttentionOp< TPrecision >
• kv_cache_op_ : Mila::Dnn::GroupedQueryAttention< TDeviceType, TComputePrecision, TKvPolicy >, Mila::Dnn::MultiHeadAttention< TDeviceType, TPrecision >
• kWeightDtype : Mila::Dnn::Compute::Cuda::Linear::CudaLinearOp< TComputePrecision, TWeightQuant >, Mila::Dnn::Linear< TDeviceType, TComputePrecision, TWeightQuant >