Merge pull request #24 from IBM/new_amlpbase

New amlpbase

Project.toml

@@ -1,7 +1,7 @@
 name = "Lale"
 uuid = "25676c37-aa2f-4f14-ad5b-b63670ababff"
 authors = ["Paulito Palmes <[email protected]>", "Martin Hirzel", "Avraham Shinnar", "Kiran Kate"]
-version = "0.1.2"
+version = "0.1.3"
 
 [deps]
 AMLPipelineBase = "e3c3008a-8869-4d53-9f34-c96f99c8a2b6"

src/Lale.jl

@@ -5,7 +5,7 @@ using PyCall
 # load abstract super-types and utils
 using AMLPipelineBase
 using AMLPipelineBase.AbsTypes
-export fit!, transform!,fit_transform!
+export fit, fit!, transform, transform!, fit_transform!
 
 using AMLPipelineBase
 using AMLPipelineBase: AbsTypes, Utils, BaselineModels, Pipelines
@@ -41,8 +41,9 @@ module LaleAbsTypes
    export LaleOperator, fit, transform, predict, lalepropertynames
 
    abstract type LaleOperator <: Learner end
-   fit(o::Machine, x::DataFrame, y::Vector=Vector()) = fit!(o,x,y)
-   transform(o::Machine, x::DataFrame) = transform!(o,x)
+
+   #fit(o::Machine, x::DataFrame, y::Vector=Vector()) = fit!(o,x,y)
+   #transform(o::Machine, x::DataFrame) = transform!(o,x)
    predict(o::Machine, x::DataFrame) = transform!(o,x)
 
    lalepropertynames(o::PyObject) = ispynull(o) ? Symbol[] : map(x->Symbol(first(x)), PyIterator{PyObject}(pycall(inspect."getmembers", PyObject, o)))

src/lalelibop.jl

@@ -9,11 +9,10 @@ using ..AbsTypes
 using ..Utils
 using ..LaleAbsTypes
 
-import ..AbsTypes: fit!, transform!
-import ..LaleAbsTypes: fit, transform, predict
+import ..AbsTypes: fit, fit!, transform, transform!
+import ..LaleAbsTypes: predict
 
-export fit!, transform!
-export fit, transform, predict
+export fit, fit!, transform, transform!, predict
 
 import Base: >>,|,+,|>,&
 export  >>,|,+,|>,&
@@ -107,28 +106,30 @@ function lalepipeoptimizers()
   println("Note: Consult Lale online help for more details about the auto_configure arguments.")
 end
 
-function fit!(lopt::LalePipeOptimizer, xx::DataFrame, y::Vector=Vector())
+function fit!(lopt::LalePipeOptimizer, xx::DataFrame, y::Vector=Vector())::Nothing
    Xpd = Pandas.DataFrame(xx).pyo
    Ypd = Pandas.DataFrame(y).pyo
    margs = lopt.model[:impl_args]
    pipe = lopt.model[:lalepipe]
    hyperopt = LALELIBS.Hyperopt(estimator=pipe; margs...)
    trained = hyperopt.fit(Xpd,Ypd)
    lopt.model[:trained] = trained
+   return nothing
 end
 
-function transform!(lopt::LalePipeOptimizer, xx::DataFrame)
-   Xpd = Pandas.DataFrame(xx).pyo
-   trainedmodel = lopt.model[:trained]
-   trainedmodel.predict(Xpd) |> Pandas.DataFrame |> DataFrame |> x -> x[:,1]
-end
-
-function fit(lopt::LalePipeOptimizer, xx::DataFrame, y::Vector=Vector()) 
+function fit(lopt::LalePipeOptimizer, xx::DataFrame, y::Vector=Vector())::LalePipeOptimizer
    fit!(lopt,xx,y)
    loptcopy = deepcopy(lopt)
    return loptcopy
 end
 
+function transform!(lopt::LalePipeOptimizer, xx::DataFrame)::Vector
+   Xpd = Pandas.DataFrame(xx).pyo
+   trainedmodel = lopt.model[:trained]
+   trainedmodel.predict(Xpd) |> Pandas.DataFrame |> DataFrame |> x -> x[:,1] |> collect
+end
+
+
 transform(lopt::LalePipeOptimizer, xx::DataFrame) = transform!(lopt,xx)
 predict(lopt::LalePipeOptimizer, xx::DataFrame) = transform!(lopt,xx)
 

src/laleop.jl

@@ -9,8 +9,8 @@ using ..AbsTypes
 using ..LaleAbsTypes
 using ..Utils
 
-import ..AbsTypes: fit!, transform!
-import ..LaleAbsTypes: fit, transform, predict
+import ..AbsTypes: fit, fit!, transform, transform!
+import ..LaleAbsTypes: predict
 
 export fit!, transform!, fit, transform, predict
 export LaleOp, skops, autogenops, lalelibops
@@ -168,7 +168,7 @@ function lalelibops()
   println("Note: Consult Scikitlearn's online help for more details about the learner's arguments.")
 end
 
-function fit!(lale::LaleOp, xx::DataFrame, y::Vector=Vector())
+function fit!(lale::LaleOp, xx::DataFrame, y::Vector=Vector())::Nothing
   x = xx |> Array
   impl_args = copy(lale.model[:impl_args])
   learner = lale.model[:learner]
@@ -195,9 +195,16 @@ function fit!(lale::LaleOp, xx::DataFrame, y::Vector=Vector())
   end
   lale.model[:laleobj]   = trained
   lale.model[:impl_args] = impl_args
+  return nothing
 end
 
-function transform!(lale::LaleOp, xx::DataFrame)
+function fit(lale::LaleOp, xx::DataFrame, y::Vector=Vector())::LaleOp 
+   fit!(lale,xx,y)
+   lcopy =deepcopy(lale)
+   return lcopy
+end
+
+function transform!(lale::LaleOp, xx::DataFrame)::Union{DataFrame, Vector}
    x = deepcopy(xx)|> Array
    laleobj = lale.model[:laleobj]
    # transform is predict for learners
@@ -210,14 +217,10 @@ function transform!(lale::LaleOp, xx::DataFrame)
    end
 end
 
-function fit(lale::LaleOp, xx::DataFrame, y::Vector=Vector()) 
-   fit!(lale,xx,y)
-   lcopy =deepcopy(lale)
-   return lcopy
+function transform(lale::LaleOp, xx::DataFrame)::Union{DataFrame, Vector}
+   return transform!(lale,xx)
 end
 
-transform(lale::LaleOp, xx::DataFrame) = transform!(lale,xx)
-predict(lale::LaleOp, xx::DataFrame)   = transform!(lale,xx)
+predict(lale::LaleOp, xx::DataFrame)  = transform!(lale,xx)
 
 end
-