Jdb/api (#279)

* minor fit init simplification

* fix tutorials reproducibility

Author: jeremiedb

docs/src/tutorials/classification-iris.md

@@ -50,7 +50,7 @@ config = EvoTreeClassifier(
     rowsample=0.8, 
     colsample=0.8)
 
-model = fit(config, dtrain;
+model = EvoTrees.fit(config, dtrain;
     target_name,
     feature_names,
     deval,

docs/src/tutorials/examples-API.md

@@ -10,6 +10,7 @@ Minimal example to fit a noisy sinus wave.
 
 ```julia
 using EvoTrees
+using EvoTrees: fit
 using EvoTrees: sigmoid, logit
 using StatsBase: sample
 
@@ -48,7 +49,7 @@ config = EvoTreeRegressor(
     max_depth=6, min_weight=1.0,
     rowsample=0.5, colsample=1.0)
 
-model = fit_evotree(config; x_train, y_train, x_eval, y_eval, print_every_n=25)
+model = fit(config; x_train, y_train, x_eval, y_eval, print_every_n=25)
 pred_eval_logistic = model(x_eval)
 
 # L1

docs/src/tutorials/logistic-regression-titanic.md

@@ -10,7 +10,7 @@ To begin, we will load the required packages and the dataset:
 using EvoTrees
 using MLDatasets
 using DataFrames
-using Statistics: mean
+using Statistics: mean, median
 using CategoricalArrays
 using Random
 
@@ -71,7 +71,7 @@ config = EvoTreeRegressor(
   rowsample=0.5, 
   colsample=0.9)
 
-model = fit(
+model = EvoTrees.fit(
     config, dtrain; 
     deval,
     target_name,

docs/src/tutorials/ranking-LTRC.md

@@ -76,7 +76,7 @@ config = EvoTreeRegressor(
     colsample=0.9,
 )
 
-m_mse, logger_mse = fit(
+m_mse, logger_mse = EvoTrees.fit(
     config;
     x_train=x_train,
     y_train=y_train,
@@ -146,7 +146,7 @@ config = EvoTreeRegressor(
     colsample=0.9,
 )
 
-m_logloss, logger_logloss = fit_evotree(
+m_logloss, logger_logloss = EvoTrees.fit(
     config;
     x_train=x_train,
     y_train=y_train,

docs/src/tutorials/regression-boston.md

@@ -49,7 +49,7 @@ config = EvoTreeRegressor(
     rowsample=0.9, 
     colsample=0.9)
 
-model = fit(config;
+model = EvoTrees.fit(config;
     x_train, y_train,
     x_eval, y_eval,
     print_every_n=10)

src/fit.jl

@@ -358,14 +358,8 @@ function fit(
     _device = params.device == :gpu ? GPU : CPU
     m, cache = init(params, dtrain, _device; target_name, feature_names, weight_name, offset_name)
 
-    # initialize callback and logger if tracking eval data
-    metric = params.metric
-    logging_flag = !isnothing(deval)
-    any_flag = !isnothing(deval)
-    if !logging_flag && any_flag
-        @warn "To track eval metric in logger, `deval` must be provided."
-    end
-    if logging_flag
+    # initialize callback and logger if deval is provided
+    if !isnothing(deval)
         deval = Tables.columntable(deval)
         cb = CallBack(params, m, deval, _device; target_name, weight_name, offset_name)
         logger = init_logger(; metric=params.metric, maximise=is_maximise(cb.feval), params.early_stopping_rounds)