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

- f -

• factories_ : Mila::Dnn::Compute::DeviceRegistry
• fast_host_device_transfer : Mila::Dnn::Compute::MemoryResourceTraits< CudaPinnedMemoryResource >
• fb : Mila::Dnn::Visualization::BlockVisualizer
• fc1_ : Mila::Dnn::MLP< TDeviceType, TPrecision >
• fc2_ : Mila::Dnn::MLP< TDeviceType, TPrecision >
• fc_down_ : Mila::Dnn::LlamaBlock< TDeviceType, TPrecision, TWeightQuant, TKvPolicy >
• fc_gate_up_ : Mila::Dnn::LlamaBlock< TDeviceType, TPrecision, TWeightQuant, TKvPolicy >
• ffn_ : Mila::Dnn::GptBlock< TDeviceType, TPrecision >
• file_ : Mila::Dnn::Compute::CublasLtError, Mila::Dnn::Compute::CudaError, Mila::Dnn::Serialization::PretrainedModelReader
• file_size_ : Mila::Data::TokenSequenceLoader< TMemoryResource >
• file_type_ : Mila::Data::MilaFileHeader
• filename_ : Mila::Dnn::Serialization::PretrainedModelReader, Mila::Dnn::Serialization::ZipSerializer
• filepath_ : Mila::Dnn::Serialization::ModelArchive, Mila::Dnn::Serialization::PretrainedModelReader
• FileSink() : Mila::Logging::FileSink
• fileStream_ : Mila::Logging::FileSink
• fill() : Mila::Dnn::Compute::Cpu::FillOps, Mila::Dnn::Compute::Cuda::FillOps
• fill_normal() : Mila::Dnn::Compute::Cuda::RandomOps
• fill_uniform() : Mila::Dnn::Compute::Cuda::RandomOps
• fillBatch() : Mila::Data::TokenSequenceLoader< TMemoryResource >
• final_layernorm_ : Mila::Dnn::GptTransformer< TDeviceType, TPrecision >
• final_lr_ : Mila::Dnn::CosineLRScheduler, Mila::Dnn::LinearLRScheduler
• final_rmsnorm_ : Mila::Dnn::LlamaTransformer< TDeviceType, TPrecision, TWeightQuantization, TKvCachePolicy >
• findComponent() : Mila::Dnn::CompositeComponent< TDeviceType, TPrecision >
• findPlugin() : Mila::Dnn::Extensibility::PluginManager
• finish() : Mila::Dnn::BufferedTokenStreamer< Sink, BufSize >
• flatten() : Mila::Dnn::Tensor< TDataType, TMemoryResource >
• flush() : Mila::Dnn::BufferedTokenStreamer< Sink, BufSize >
• flush_every_ : Mila::Utils::StepLogger
• forClassification() : Mila::Data::SpecialTokens
• forMLM() : Mila::Data::SpecialTokens
• forward() : 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::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 >
• forward_executed_ : Mila::Dnn::GptBlock< TDeviceType, TPrecision >, Mila::Dnn::LlamaBlock< TDeviceType, TPrecision, TWeightQuant, TKvPolicy >
• forward_fast() : Mila::Dnn::Compute::Cuda::LayerNorm::Detail::cuda_layernorm_impl< float >, Mila::Dnn::Compute::Cuda::LayerNorm::Detail::cuda_layernorm_impl< half >
• forward_func : Mila::Dnn::Compute::Cuda::Gelu::Detail::cuda_gelu_impl< float >
• forward_general() : Mila::Dnn::Compute::Cuda::Softmax::Detail::cuda_softmax_impl< float >, Mila::Dnn::Compute::Cuda::Softmax::Detail::cuda_softmax_impl< half >
• forward_optimized() : Mila::Dnn::Compute::Cuda::Softmax::Detail::cuda_softmax_impl< float >, Mila::Dnn::Compute::Cuda::Softmax::Detail::cuda_softmax_impl< half >
• forward_plan_cache_ : Mila::Dnn::Compute::Cuda::Linear::CudaLinearOp< TComputePrecision, TWeightQuant >
• ForwardFn : Mila::Dnn::Detail::mlp_activation_impl< ActivationType::Gelu, TDeviceType, TPrecision >, Mila::Dnn::Detail::mlp_activation_impl< ActivationType::Swiglu, TDeviceType, TPrecision >
• forwardNaive() : Mila::Dnn::Compute::CpuLinearOp
• forwardUnrolled() : Mila::Dnn::Compute::CpuLinearOp
• fp32_compute_type : Mila::Dnn::Compute::CudaDataTypeMap< __nv_bfloat16 >, Mila::Dnn::Compute::CudaDataTypeMap< half >
• Framebuffer() : Mila::Dnn::Visualization::Framebuffer
• fromArchive_() : Mila::Dnn::Gelu< TDeviceType, TPrecision >, Mila::Dnn::Swiglu< TDeviceType, TPrecision >
• fromCheckpoint() : Mila::Dnn::GptModel< TDeviceType, TPrecision >
• fromJson() : Mila::Data::SerializationMetadata, Mila::Dnn::Optimizers::SerializationMetadata, Mila::Dnn::Serialization::SerializationMetadata, Mila::Dnn::SerializationMetadata
• fromMetadata() : Mila::Data::BpeVocabularyConfig, Mila::Data::CharVocabularyConfig, Mila::Dnn::ComponentConfig, Mila::Dnn::CrossEntropyConfig, Mila::Dnn::GeluConfig, Mila::Dnn::GptBlockConfig, Mila::Dnn::GptConfig, Mila::Dnn::GqaConfig, Mila::Dnn::LayerNormConfig, Mila::Dnn::LinearConfig, Mila::Dnn::LlamaConfig, Mila::Dnn::LpeConfig, Mila::Dnn::MLPConfig, Mila::Dnn::MultiHeadAttentionConfig, Mila::Dnn::Optimizers::AdamWConfig, Mila::Dnn::ResidualConfig, Mila::Dnn::RmsNormConfig, Mila::Dnn::RopeConfig, Mila::Dnn::SoftmaxConfig, Mila::Dnn::SwigluConfig, Mila::Dnn::TokenEmbeddingConfig
• fromPretrained() : Mila::Dnn::GptModel< TDeviceType, TPrecision >, Mila::Dnn::GptTransformer< TDeviceType, TPrecision >, Mila::Dnn::LlamaModel< TDeviceType, TPrecision >
• fromPretrainedImpl() : Mila::Dnn::LlamaModel< TDeviceType, TPrecision >
• front_buffer_ready_ : Mila::Data::TokenSequenceLoader< TMemoryResource >
• front_input_tensor_ : Mila::Data::TokenSequenceLoader< TMemoryResource >
• front_target_tensor_ : Mila::Data::TokenSequenceLoader< TMemoryResource >
• function_ : Mila::Dnn::Compute::CublasLtError, Mila::Dnn::Compute::CudaError
• FusedComponent() : Mila::Dnn::FusedComponent< TDeviceType, TPrecision >