Add Time Series Block

FastTimeSeries/src/models.jl

@@ -7,7 +7,7 @@ Construct a model for time-series classification.
 function blockmodel(inblock::TimeSeriesRow,
                 outblock::OneHotTensor{0}, 
                 backbone)
-    data   = [rand(Float32, inblock.nfeatures, 256) for _ ∈ 1:inblock.obslength]
+    data   = [rand(Float32, inblock.nfeatures, 32) for _ ∈ 1:inblock.obslength]
     output = backbone(data)
     outs   = size(output)[1]
     return Models.RNNModel(backbone, outsize = length(outblock.classes), recout = outs)

FastTimeSeries/src/models/RNN.jl

@@ -1,3 +1,9 @@
+function tabular2rnn(X::AbstractArray{Float32, 3})
+    X = permutedims(X, (2, 1, 3))
+    X = [X[t, :, :] for t ∈ 1:size(X, 1)]
+    return X
+end
+
 """
     RNNModel(recbackbonem, outsize, recout[; kwargs...])
 
@@ -24,5 +30,5 @@ function RNNModel(recbackbone,
                   dropout_rate = 0.0)
 
     dropout = dropout_rate == 0 ? identity : Dropout(dropout_rate)
-    Chain(recbackbone, dropout, finalclassifier)
-end
+    Chain(tabular2rnn, recbackbone, dropout, finalclassifier)
+end

FastTimeSeries/src/models/StackedLSTM.jl

@@ -2,7 +2,7 @@
 #
 # Layers for stacked LSTM (referenced from https://github.com/sdobber/FluxArchitectures.jl)
 
-mutable struct StackedLSTMCell{A}
+struct StackedLSTMCell{A}
 	chain::A
 end
 
@@ -34,12 +34,13 @@ function StackedLSTM(in::Int, out::Integer, hiddensize::Integer, layers::Integer
 end
 
 function (m::StackedLSTMCell)(X)
+	Flux.reset!(m)
 	[m.chain(x) for x ∈ X[1:end-1]]
 	return m.chain(X[end])
 end
 
 Flux.@functor StackedLSTMCell 
-Flux.trainable(m::StackedLSTMCell) = (m.chain)
+# Flux.trainable(m::StackedLSTMCell) = (m.chain)
 
 # Initialize forget gate bias with 1
 function initialize_bias!(l::StackedLSTMCell)

FastTimeSeries/src/tasks/classification.jl

@@ -42,8 +42,7 @@ _tasks["tsclfsingle"] = (
         row = FastAI.mockblock(task.blocks[1])
 
         xtrain = FastAI.encodeinput(task, Training(), row);
-        @test length(xtrain[1,:]) == task.blocks.input.obslength
-        @test length(xtrain[:,1]) == task.blocks.input.nfeatures
+        @test size(xtrain) == (task.blocks.input.nfeatures, task.blocks.input.obslength)
 
         @test eltype(xtrain[1,:]) <: Number
     end