Functions
Merlin.LinearBase.:*Base.:+Base.:-Base.:/Base.:^Base.LinAlg.BLAS.gemmBase.LinAlg.BLAS.gemvBase.broadcastBase.broadcastBase.broadcastBase.expBase.getindexBase.logBase.reshapeBase.transposeMerlin.argmaxMerlin.concatMerlin.dropoutMerlin.gemm_batchMerlin.logsoftmaxMerlin.softmax
Activation
BLAS
Merlin.gemm_batch — Method.gemm_batch(tA::Char, tB::Char, alpha, A::Var, B::Var)Base.LinAlg.BLAS.gemm — Method.gemm(tA::Char, tB::Char, alpha, A::Var, B::Var)
gemm(A::Var, B::Var, [tA='N'], [tB='N'], [alpha=1])tA, tB: 'T' (transpose) or 'N' (not transpose)
\[C = \alpha \times \textrm{tA}(A) \times \textrm{tB}(B)\]
T = Float32
A = Var(rand(T,10,5))
B = Var(rand(T,10,7))
C = BLAS.gemm('T', 'N', 1, A, B)Base.LinAlg.BLAS.gemv — Method.BLAS.gemv(tA::Char, alpha, A::Var, x::Var)tA: 'T' (transpose) or 'N' (not transpose)
\[y = \alpha \times \textrm{tA}(A) \times x\]
T = Float32
A = Var(rand(T,10,5))
x = Var(rand(T,5))
B = BLAS.gemv('N', 1, A, x)Convolution
Embeddings
Loss
Math
Base.:* — Method.*(A::Var, B::Var)Base.:+ — Method.+(x1::Var, x2::Var)
+(a::Number, x::Var)
+(x::Var, a::Number)Base.:- — Method.-(x1, x2)Base.:/ — Method./(x1::Var, a)Base.:^ — Method.^(x::Var, a::Number)Base.broadcast — Method..*(x1::Var, x2::Var)Base.broadcast — Method..+(x1::Var, x2::Var)Base.broadcast — Method..-(x1::Var, x2::Var)Base.exp — Method.exp(x)Base.log — Method.log(x)Base.transpose — Method.transpose(x)Recurrent
Reduction
Misc
argmax
concat
dropout
getindex
Linear
lookup
reshape
softmax
logsoftmax
standardize