thread throttling

Author: mayer79

R/plot_contribution.R

@@ -35,7 +35,8 @@
 #'    y = target,
 #'    objective = "reg:squarederror",
 #'    max_depth = 3,
-#'    nrounds = 20
+#'    nrounds = 20,
+#'    nthreads = 1
 #'   )
 #'   unified_model <- xgboost.unify(xgb_model, as.matrix(data))
 #'   x <- head(data, 1)

R/plot_feature_dependence.R

@@ -29,7 +29,8 @@
 #'    y = target,
 #'    objective = "reg:squarederror",
 #'    max_depth = 3,
-#'    nrounds = 20
+#'    nrounds = 20,
+#'    nthreads = 1
 #'   )
 #'   unified_model <- xgboost.unify(xgb_model, as.matrix(data))
 #'   x <- head(data, 100)

R/plot_feature_importance.R

@@ -32,7 +32,8 @@
 #'    y = target,
 #'    objective = "reg:squarederror",
 #'    max_depth = 3,
-#'    nrounds = 20
+#'    nrounds = 20,
+#'    nthreads = 1
 #'   )
 #'   unified_model <- xgboost.unify(xgb_model, as.matrix(data))
 #'   shaps <- treeshap(unified_model, as.matrix(head(data, 3)))

R/plot_interaction.R

@@ -30,7 +30,8 @@
 #'    y = target,
 #'    objective = "reg:squarederror",
 #'    max_depth = 5,
-#'    nrounds = 10
+#'    nrounds = 10,
+#'    nthreads = 1
 #'   )
 #'   unified_model2 <- xgboost.unify(xgb_model2, data)
 #'   inters <- treeshap(unified_model2, as.matrix(data[1:50, ]), interactions = TRUE)

R/treeshap.R

@@ -28,39 +28,41 @@
 #'   library(xgboost)
 #'   data <- fifa20$data[colnames(fifa20$data) != 'work_rate']
 #'   target <- fifa20$target
-#'  
+#'
 #'   # calculating simple SHAP values
 #'   xgb_model <- xgboost::xgboost(
 #'    x = as.matrix(data),
 #'    y = target,
 #'    objective = "reg:squarederror",
 #'    max_depth = 3,
-#'    nrounds = 20
+#'    nrounds = 20,
+#'    nthreads = 1
 #'   )
 #'   unified_model <- xgboost.unify(xgb_model, as.matrix(data))
 #'   treeshap1 <- treeshap(unified_model, head(data, 3))
 #'   plot_contribution(treeshap1, obs = 1)
 #'   treeshap1$shaps
-#'  
+#'
 #'   # It's possible to calcualte explanation over different part of the data set
-#'  
+#'
 #'   unified_model_rec <- set_reference_dataset(unified_model, data[1:1000, ])
 #'   treeshap_rec <- treeshap(unified_model, head(data, 3))
 #'   plot_contribution(treeshap_rec, obs = 1)
-#'  
+#'
 #'   # calculating SHAP interaction values
 #'   xgb_model2 <- xgboost::xgboost(
 #'    x = as.matrix(data),
 #'    y = target,
 #'    objective = "reg:squarederror",
 #'    max_depth = 7,
-#'    nrounds = 10
+#'    nrounds = 10,
+#'    nthreads = 1
 #'   )
 #'   unified_model2 <- xgboost.unify(xgb_model2, as.matrix(data))
 #'   treeshap2 <- treeshap(unified_model2, head(data, 3), interactions = TRUE)
 #'   treeshap2$interactions
 #' }}
-#' 
+#'
 treeshap <- function(unified_model, x, interactions = FALSE, verbose = TRUE) {
   UseMethod("treeshap", unified_model)
 }

R/unify_gpboost.R

@@ -19,7 +19,7 @@
 #' \code{\link{gbm.unify}} for \code{\link[gbm:gbm]{GBM models}}
 #'
 #' \code{\link{xgboost.unify}} for \code{\link[xgboost:xgboost]{XGBoost models}}
-#' 
+#'
 #' \code{\link{lightgbm.unify}} for \code{\link[lightgbm:lightgbm]{LightGBM models}}
 #'
 #' \code{\link{ranger.unify}} for \code{\link[ranger:ranger]{ranger models}}
@@ -40,7 +40,7 @@
 #'   x <- gpboost::gpb.Dataset(sparse_data, label = as.matrix(data[,ncol(data)]))
 #'   gpb_data <- gpboost::gpb.Dataset.construct(x)
 #'   gpb_model <- gpboost::gpboost(data = gpb_data, params = param_gpb,
-#'                                   verbose = -1, num_threads = 0)
+#'                                   verbose = -1, num_threads = 1)
 #'   unified_model <- gpboost.unify(gpb_model, sparse_data)
 #'   shaps <- treeshap(unified_model, data[1:2, ])
 #'   plot_contribution(shaps, obs = 1)
@@ -90,26 +90,26 @@ gpboost.unify <- function(gpb_model, data, recalculate = FALSE) {
   df <- df[, c("tree_index", "split_index", "split_feature", "Decision.type", "threshold", "Yes", "No", "Missing", "split_gain", "internal_count")]
   colnames(df) <- c("Tree", "Node", "Feature", "Decision.type", "Split", "Yes", "No", "Missing", "Prediction", "Cover")
   attr(df, "sorted") <- NULL
-  
+
   ID <- paste0(df$Node, "-", df$Tree)
   df$Yes <- match(paste0(df$Yes, "-", df$Tree), ID)
   df$No <- match(paste0(df$No, "-", df$Tree), ID)
   df$Missing <- match(paste0(df$Missing, "-", df$Tree), ID)
-  
+
   # Here we lose "Quality" information
   df$Prediction[!is.na(df$Feature)] <- NA
-  
+
   feature_names <- jsonlite::fromJSON(gpb_model$dump_model())$feature_names
   data <- data[,colnames(data) %in% feature_names]
-  
+
   ret <- list(model = as.data.frame(df), data = as.data.frame(data), feature_names = feature_names)
   class(ret) <- "model_unified"
   attr(ret, "missing_support") <- TRUE
   attr(ret, "model") <- "gpboost"
-  
+
   if (recalculate) {
     ret <- set_reference_dataset(ret, as.data.frame(data))
   }
-  
+
   return(ret)
 }

R/unify_lightgbm.R

@@ -37,7 +37,7 @@
 #'   x <- lightgbm::lgb.Dataset(sparse_data, label = as.matrix(data[,ncol(data)]))
 #'   lgb_data <- lightgbm::lgb.Dataset.construct(x)
 #'   lgb_model <- lightgbm::lightgbm(data = lgb_data, params = param_lgbm,
-#'                                   verbose = -1, num_threads = 0)
+#'                                   verbose = -1, num_threads = 1)
 #'   unified_model <- lightgbm.unify(lgb_model, sparse_data)
 #'   shaps <- treeshap(unified_model, data[1:2, ])
 #'   plot_contribution(shaps, obs = 1)

R/unify_xgboost.R

@@ -31,7 +31,8 @@
 #'    y = target,
 #'    objective = "reg:squarederror",
 #'    max_depth = 3,
-#'    nrounds = 20
+#'    nrounds = 20,
+#'    nthreads = 1
 #'   )
 #'   unified_model <- xgboost.unify(xgb_model, as.matrix(data))
 #'   shaps <- treeshap(unified_model, data[1:2,])

cran-comments.md

@@ -1,4 +1,8 @@
-# Resubmission
+# Resubmission 2
+
+Resubmission after fixing warnings on too high thread count in "dont test" examples and unit tests.
+
+# Resubmission 1
 
 Resubmission after fixing a warning in a "don't" test example.