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

- b -

• b : Mila::Dnn::Visualization::RGB
• B_ : Mila::Dnn::Compute::CpuAttentionOp, Mila::Dnn::Compute::Cuda::Gqa::CudaGqaOp< TPrecision >, Mila::Dnn::Compute::Cuda::MultiHeadAttention::CudaMultiHeadAttentionOp< TPrecision >
• back() : Mila::Dnn::TensorShape
• back_buffer_ready_ : Mila::Data::TokenSequenceLoader< TMemoryResource >
• back_input_tensor_ : Mila::Data::TokenSequenceLoader< TMemoryResource >
• back_target_tensor_ : Mila::Data::TokenSequenceLoader< TMemoryResource >
• backward() : Mila::Dnn::Compute::BinaryOperation< TDeviceType, TPrecision, TInputA, TInputB >, Mila::Dnn::Compute::CpuAttentionOp, Mila::Dnn::Compute::CpuCrossEntropyOp, Mila::Dnn::Compute::CpuEncoderOp, Mila::Dnn::Compute::CpuGeluOp, Mila::Dnn::Compute::CpuLayerNormOp, Mila::Dnn::Compute::CpuLinearOp, Mila::Dnn::Compute::CpuResidualOp, Mila::Dnn::Compute::CpuSoftmaxCrossEntropyOp< TPrecision, TLogits, TTargets >, Mila::Dnn::Compute::CpuSoftmaxOp, Mila::Dnn::Compute::Cuda::Gelu::CudaGeluOp< TPrecision >, Mila::Dnn::Compute::Cuda::Gelu::Detail::cuda_gelu_impl< float >, Mila::Dnn::Compute::Cuda::Gelu::Detail::cuda_gelu_impl< half >, Mila::Dnn::Compute::Cuda::Gqa::CudaGqaOp< TPrecision >, Mila::Dnn::Compute::Cuda::LayerNorm::CudaLayerNormOp< TPrecision >, Mila::Dnn::Compute::Cuda::LayerNorm::Detail::cuda_layernorm_impl< float >, Mila::Dnn::Compute::Cuda::LayerNorm::Detail::cuda_layernorm_impl< half >, Mila::Dnn::Compute::Cuda::Linear::CudaLinearOp< TComputePrecision, TWeightQuant >, Mila::Dnn::Compute::Cuda::Linear::Detail::cuda_matmul_impl< float >, Mila::Dnn::Compute::Cuda::Lpe::CudaLpeOp< TInput, TPrecision >, Mila::Dnn::Compute::Cuda::Lpe::Detail::cuda_lpe_impl< float >, Mila::Dnn::Compute::Cuda::Lpe::Detail::cuda_lpe_impl< half >, Mila::Dnn::Compute::Cuda::MatMulBiasGelu::CudaMatMulBiasGeluOp< TInput, TOutput >, Mila::Dnn::Compute::Cuda::MultiHeadAttention::CudaMultiHeadAttentionOp< TPrecision >, Mila::Dnn::Compute::Cuda::Residual::CudaResidualOp< TInputA, TInputB, TPrecision >, Mila::Dnn::Compute::Cuda::Residual::Detail::cuda_residual_impl< float >, Mila::Dnn::Compute::Cuda::Residual::Detail::cuda_residual_impl< nv_bfloat16 >, Mila::Dnn::Compute::Cuda::RmsNorm::CudaRmsNormOp< TPrecision >, Mila::Dnn::Compute::Cuda::RmsNorm::Detail::cuda_rmsnorm_impl< float >, Mila::Dnn::Compute::Cuda::RmsNorm::Detail::cuda_rmsnorm_impl< nv_bfloat16 >, Mila::Dnn::Compute::Cuda::Rope::CudaRopeOp< TComputePrecision >, Mila::Dnn::Compute::Cuda::Rope::Detail::cuda_rope_impl< __nv_bfloat16 >, Mila::Dnn::Compute::Cuda::Rope::Detail::cuda_rope_impl< float >, Mila::Dnn::Compute::Cuda::Softmax::CudaSoftmaxOp< TPrecision >, Mila::Dnn::Compute::Cuda::Softmax::Detail::cuda_softmax_impl< float >, Mila::Dnn::Compute::Cuda::Softmax::Detail::cuda_softmax_impl< half >, Mila::Dnn::Compute::Cuda::SoftmaxCrossEntropy::CudaSoftmaxCrossEntropyOp< TPrecision, TLogits, TTargets >, Mila::Dnn::Compute::Cuda::SoftmaxCrossEntropy::Detail::cuda_softmax_crossentropy_impl< float >, Mila::Dnn::Compute::Cuda::SoftmaxCrossEntropy::Detail::cuda_softmax_crossentropy_impl< half >, Mila::Dnn::Compute::Cuda::Swiglu::CudaSwigluOp< TPrecision >, Mila::Dnn::Compute::Cuda::Swiglu::Detail::cuda_swiglu_impl< __nv_bfloat16 >, Mila::Dnn::Compute::Cuda::Swiglu::Detail::cuda_swiglu_impl< float >, Mila::Dnn::Compute::Cuda::TokenEmbedding::CudaTokenEmbeddingOp< TInput, TPrecision >, Mila::Dnn::Compute::Cuda::TokenEmbedding::Detail::cuda_token_embedding_impl< __nv_bfloat16 >, Mila::Dnn::Compute::Cuda::TokenEmbedding::Detail::cuda_token_embedding_impl< float >, Mila::Dnn::Compute::PairedOperation< TDeviceType, TPrecision, TInputA, TInputB >, Mila::Dnn::Compute::UnaryOperation< TDeviceType, TInput, TPrecision >, Mila::Dnn::Dropout< TDeviceType, TInput, TOutput >, Mila::Dnn::FusedComponent< TDeviceType, TPrecision >, Mila::Dnn::Gelu< TDeviceType, TPrecision >, Mila::Dnn::GptBlock< TDeviceType, TPrecision >, Mila::Dnn::GptTransformer< TDeviceType, TPrecision >, Mila::Dnn::GroupedQueryAttention< TDeviceType, TComputePrecision, TKvPolicy >, Mila::Dnn::LanguageNetwork< TDeviceType, TPrecision >, Mila::Dnn::LayerNorm< TDeviceType, TPrecision >, Mila::Dnn::Linear< TDeviceType, TComputePrecision, TWeightQuant >, Mila::Dnn::LlamaBlock< TDeviceType, TPrecision, TWeightQuant, TKvPolicy >, Mila::Dnn::LlamaTransformer< TDeviceType, TPrecision, TWeightQuantization, TKvCachePolicy >, Mila::Dnn::Lpe< TDeviceType, TIndex, TPrecision >, Mila::Dnn::MLP< TDeviceType, TPrecision >, Mila::Dnn::MultiHeadAttention< TDeviceType, TPrecision >, Mila::Dnn::Residual< TDeviceType, TPrecision >, Mila::Dnn::RmsNorm< TDeviceType, TPrecision >, Mila::Dnn::Rope< TDeviceType, TPrecision >, Mila::Dnn::Softmax< TDeviceType, TPrecision >, Mila::Dnn::SoftmaxCrossEntropy< TDeviceType, TLogits, TTargets, TPrecision >, Mila::Dnn::Swiglu< TDeviceType, TPrecision >, Mila::Dnn::TokenEmbedding< TDeviceType, TIndex, TPrecision >
• backward_att_plan_ : Mila::Dnn::Compute::Cuda::Gqa::CudaGqaOp< TPrecision >, Mila::Dnn::Compute::Cuda::MultiHeadAttention::CudaMultiHeadAttentionOp< TPrecision >
• backward_fast() : Mila::Dnn::Compute::Cuda::LayerNorm::Detail::cuda_layernorm_impl< float >, Mila::Dnn::Compute::Cuda::LayerNorm::Detail::cuda_layernorm_impl< half >
• backward_func : Mila::Dnn::Compute::Cuda::Gelu::Detail::cuda_gelu_impl< float >
• backward_impl() : Mila::Dnn::Compute::CpuCrossEntropyOp
• backward_input_plan_cache_ : Mila::Dnn::Compute::Cuda::Linear::CudaLinearOp< TComputePrecision, TWeightQuant >
• backward_k_plan_ : Mila::Dnn::Compute::Cuda::Gqa::CudaGqaOp< TPrecision >, Mila::Dnn::Compute::Cuda::MultiHeadAttention::CudaMultiHeadAttentionOp< TPrecision >
• backward_q_plan_ : Mila::Dnn::Compute::Cuda::Gqa::CudaGqaOp< TPrecision >, Mila::Dnn::Compute::Cuda::MultiHeadAttention::CudaMultiHeadAttentionOp< TPrecision >
• backward_v_plan_ : Mila::Dnn::Compute::Cuda::Gqa::CudaGqaOp< TPrecision >, Mila::Dnn::Compute::Cuda::MultiHeadAttention::CudaMultiHeadAttentionOp< TPrecision >
• backward_weight_plan_ : Mila::Dnn::Compute::Cuda::Linear::CudaLinearOp< TComputePrecision, TWeightQuant >
• BackwardFn : Mila::Dnn::Detail::mlp_activation_impl< ActivationType::Gelu, TDeviceType, TPrecision >, Mila::Dnn::Detail::mlp_activation_impl< ActivationType::Swiglu, TDeviceType, TPrecision >
• Base : Mila::Dnn::LanguageModel< TDeviceType, TPrecision >
• base : Mila::Dnn::Compute::Cuda::Rope::RopeCacheRegistry::CacheKey
• base_ : Mila::Dnn::RopeConfig
• BaseLoader : Mila::Data::TokenSequenceLoader< TMemoryResource >
• baseToString() : Mila::Dnn::LanguageModelConfig< TDerived >, Mila::Dnn::ModelConfig
• batch_size_ : Mila::Data::DataLoader< TInputDataType, TTargetDataType, TMemoryResource >, Mila::Dnn::Compute::CpuEncoderOp, Mila::Dnn::Compute::CpuLinearOp, Mila::Dnn::Compute::Cuda::Lpe::CudaLpeOp< TInput, TPrecision >, Mila::Dnn::Compute::Cuda::Rope::CudaRopeOp< TComputePrecision >, Mila::Dnn::Compute::Cuda::TokenEmbedding::CudaTokenEmbeddingOp< TInput, TPrecision >, Mila::Dnn::GptTransformer< TDeviceType, TPrecision >, Mila::Dnn::LlamaTransformer< TDeviceType, TPrecision, TWeightQuantization, TKvCachePolicy >
• batch_timeout_ms : Mila::Data::TokenSequenceLoaderConfig
• batches_per_window_ : Mila::Data::TokenSequenceLoader< TMemoryResource >
• batchSize() : Mila::Data::DataLoader< TInputDataType, TTargetDataType, TMemoryResource >
• begin() : Mila::Dnn::TensorShape
• beta1_ : Mila::Dnn::Optimizers::AdamWConfig
• beta2_ : Mila::Dnn::Optimizers::AdamWConfig
• bias_ : Mila::Dnn::Compute::CpuLayerNormOp, Mila::Dnn::Compute::CpuLinearOp, Mila::Dnn::Compute::Cuda::LayerNorm::CudaLayerNormOp< TPrecision >, Mila::Dnn::Compute::Cuda::Linear::CudaLinearOp< TComputePrecision, TWeightQuant >, Mila::Dnn::Compute::Cuda::RmsNorm::CudaRmsNormOp< TPrecision >, Mila::Dnn::LayerNorm< TDeviceType, TPrecision >, Mila::Dnn::Linear< TDeviceType, TComputePrecision, TWeightQuant >, Mila::Dnn::RmsNorm< TDeviceType, TPrecision >
• bias_grad_ : Mila::Dnn::Compute::CpuLayerNormOp, Mila::Dnn::Compute::CpuLinearOp, Mila::Dnn::Compute::Cuda::LayerNorm::CudaLayerNormOp< TPrecision >, Mila::Dnn::Compute::Cuda::Linear::CudaLinearOp< TComputePrecision, TWeightQuant >, Mila::Dnn::Compute::Cuda::RmsNorm::CudaRmsNormOp< TPrecision >, Mila::Dnn::LayerNorm< TDeviceType, TPrecision >, Mila::Dnn::Linear< TDeviceType, TComputePrecision, TWeightQuant >, Mila::Dnn::RmsNorm< TDeviceType, TPrecision >
• binary_registry_ : Mila::Dnn::Compute::OperationRegistry
• BinaryOperationBase : Mila::Dnn::Compute::CpuSoftmaxCrossEntropyOp< TPrecision, TLogits, TTargets >, Mila::Dnn::Compute::Cuda::SoftmaxCrossEntropy::CudaSoftmaxCrossEntropyOp< TPrecision, TLogits, TTargets >
• bind() : Mila::Dnn::Detail::mlp_activation_impl< ActivationType::Gelu, TDeviceType, TPrecision >, Mila::Dnn::Detail::mlp_activation_impl< ActivationType::Swiglu, TDeviceType, TPrecision >
• bits_per_element : Mila::Dnn::TensorDataTypeTraits< TensorDataType::FP4_E2M1 >, Mila::Dnn::TensorDataTypeTraits< TensorDataType::FP4_E3M0 >
• block_input_ptrs_ : Mila::Dnn::GptTransformer< TDeviceType, TPrecision >, Mila::Dnn::LlamaTransformer< TDeviceType, TPrecision, TWeightQuantization, TKvCachePolicy >
• block_output_ptrs_ : Mila::Dnn::GptTransformer< TDeviceType, TPrecision >, Mila::Dnn::LlamaTransformer< TDeviceType, TPrecision, TWeightQuantization, TKvCachePolicy >
• BlockVisualizer() : Mila::Dnn::Visualization::BlockVisualizer
• bos_token : Mila::Data::SpecialTokens
• BpeTokenizer() : Mila::Data::BpeTokenizer
• BpeTrainer() : Mila::Data::BpeTrainer
• BpeVocabulary() : Mila::Data::BpeVocabulary
• BpeVocabularyConfig() : Mila::Data::BpeVocabularyConfig
• buckets() : Mila::Dnn::Compute::Cuda::CublasLtPlanCache< TPlan >
• buckets_ : Mila::Dnn::Compute::Cuda::CublasLtPlanCache< TPlan >
• buf_ : Mila::Dnn::BufferedTokenStreamer< Sink, BufSize >
• buffer_ : Mila::Dnn::Tensor< TDataType, TMemoryResource >
• BufferedTokenStreamer() : Mila::Dnn::BufferedTokenStreamer< Sink, BufSize >
• build() : Mila::Dnn::Component< TDeviceType, TPrecision >, Mila::Dnn::Compute::CpuAttentionOp, Mila::Dnn::Compute::CpuEncoderOp, Mila::Dnn::Compute::CpuGeluOp, Mila::Dnn::Compute::CpuLayerNormOp, Mila::Dnn::Compute::CpuLinearOp, Mila::Dnn::Compute::CpuSoftmaxCrossEntropyOp< TPrecision, TLogits, TTargets >, Mila::Dnn::Compute::CpuSoftmaxOp, Mila::Dnn::Compute::Cuda::Gqa::CudaGqaOp< TPrecision >, Mila::Dnn::Compute::Cuda::LayerNorm::CudaLayerNormOp< TPrecision >, Mila::Dnn::Compute::Cuda::Linear::CudaLinearOp< TComputePrecision, TWeightQuant >, Mila::Dnn::Compute::Cuda::Lpe::CudaLpeOp< TInput, TPrecision >, Mila::Dnn::Compute::Cuda::MultiHeadAttention::CudaMultiHeadAttentionOp< TPrecision >, Mila::Dnn::Compute::Cuda::Residual::CudaResidualOp< TInputA, TInputB, TPrecision >, Mila::Dnn::Compute::Cuda::RmsNorm::CudaRmsNormOp< TPrecision >, Mila::Dnn::Compute::Cuda::Rope::CudaRopeOp< TComputePrecision >, Mila::Dnn::Compute::Cuda::Softmax::CudaSoftmaxOp< TPrecision >, Mila::Dnn::Compute::Cuda::SoftmaxCrossEntropy::CudaSoftmaxCrossEntropyOp< TPrecision, TLogits, TTargets >, Mila::Dnn::Compute::Cuda::TokenEmbedding::CudaTokenEmbeddingOp< TInput, TPrecision >, Mila::Dnn::Compute::Operation< TDeviceType, TComputePrecision >
• build_cache() : Mila::Dnn::Compute::Cuda::Rope::Detail::cuda_rope_impl< __nv_bfloat16 >, Mila::Dnn::Compute::Cuda::Rope::Detail::cuda_rope_impl< float >
• build_context_ : Mila::Dnn::Component< TDeviceType, TPrecision >
• BuildContext() : Mila::Dnn::BuildContext
• buildCublasLtPlans() : Mila::Dnn::Compute::Cuda::Gqa::CudaGqaOp< TPrecision >, Mila::Dnn::Compute::Cuda::Linear::CudaLinearOp< TComputePrecision, TWeightQuant >, Mila::Dnn::Compute::Cuda::MultiHeadAttention::CudaMultiHeadAttentionOp< TPrecision >
• buildCublasLtPlans_optimized() : Mila::Dnn::Compute::Cuda::Gqa::CudaGqaOp< TPrecision >
• buildFromText() : Mila::Data::BpeVocabulary, Mila::Data::CharVocabulary
• buildMergeMap() : Mila::Data::BpeVocabulary
• buildMessage() : CudaException
• buildSpecialTokenList() : Mila::Data::BpeVocabulary
• built_ : Mila::Dnn::Component< TDeviceType, TPrecision >
• byte_level_ : Mila::Data::BpeVocabularyConfig, Mila::Data::CharVocabularyConfig