Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Added LGBM_DatasetCreateFromFile #164

Open
wants to merge 10 commits into
base: all-params
Choose a base branch
from
Open

Added LGBM_DatasetCreateFromFile #164

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
753fc94
implemented LGBM_DatasetCreateFromFile
kainkad Oct 18, 2024
155de38
some unclear behaviour
kainkad Oct 18, 2024
c2340fb
further curiosities
kainkad Oct 22, 2024
617a788
updated comments
kainkad Oct 28, 2024
aa04d9e
:pencil2:
kainkad Oct 28, 2024
83a17ce
updated fitting method to include datasets from filepaths
kainkad Nov 5, 2024
02613ff
:pencil2: bye bye
kainkad Nov 7, 2024
8d5f3ee
updated with the target branch
kainkad Nov 11, 2024
d31e144
updated with all-params
kainkad Nov 26, 2024
2712934
updated params
kainkad Nov 27, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 16 additions & 0 deletions src/MLJInterface.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -80,6 +80,7 @@ MLJModelInterface.@mlj_model mutable struct LGBMRegressor <: MLJModelInterface.D
monotone_constraints_method::String = "basic"::(_ in ("basic", "intermediate", "advanced"))
monotone_penalty::Float64 = 0.0::(_ >= 0.0)
feature_contri::Vector{Float64} = Vector{Float64}()
forcedsplits_filename::String = ""
refit_decay_rate::Float64 = 0.9::(0.0 <= _ <= 1.0)
cegb_tradeoff::Float64 = 1.0::(_ >= 0.0)
cegb_penalty_split::Float64 = 0.0::(_ >= 0.0)
Expand All @@ -102,7 +103,14 @@ MLJModelInterface.@mlj_model mutable struct LGBMRegressor <: MLJModelInterface.D
zero_as_missing::Bool = false
feature_pre_filter::Bool = true
pre_partition::Bool = false
two_round::Bool = false
header::Bool = false
label_column::String = ""
weight_column::String = ""
ignore_column::String = ""
categorical_feature::Vector{Int} = Vector{Int}()
forcedbins_filename::String = ""
precise_float_parser::Bool = false

# Predict parameters
start_iteration_predict::Int = 0
Expand Down Expand Up @@ -203,6 +211,7 @@ MLJModelInterface.@mlj_model mutable struct LGBMClassifier <: MLJModelInterface.
monotone_constraints_method::String = "basic"::(_ in ("basic", "intermediate", "advanced"))
monotone_penalty::Float64 = 0.0::(_ >= 0.0)
feature_contri::Vector{Float64} = Vector{Float64}()
forcedsplits_filename::String = ""
refit_decay_rate::Float64 = 0.9::(0.0 <= _ <= 1.0)
cegb_tradeoff::Float64 = 1.0::(_ >= 0.0)
cegb_penalty_split::Float64 = 0.0::(_ >= 0.0)
Expand All @@ -225,7 +234,14 @@ MLJModelInterface.@mlj_model mutable struct LGBMClassifier <: MLJModelInterface.
zero_as_missing::Bool = false
feature_pre_filter::Bool = true
pre_partition::Bool = false
two_round::Bool = false
header::Bool = false
label_column::String = ""
weight_column::String = ""
ignore_column::String = ""
categorical_feature::Vector{Int} = Vector{Int}()
forcedbins_filename::String = ""
precise_float_parser::Bool = false

# Predict parameters
start_iteration_predict::Int = 0
Expand Down
116 changes: 97 additions & 19 deletions src/estimators.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -57,6 +57,7 @@ mutable struct LGBMRegression <: LGBMEstimator
monotone_constraints_method::String
monotone_penalty::Float64
feature_contri::Vector{Float64}
forcedsplits_filename::String
refit_decay_rate::Float64
cegb_tradeoff::Float64
cegb_penalty_split::Float64
Expand All @@ -79,7 +80,14 @@ mutable struct LGBMRegression <: LGBMEstimator
zero_as_missing::Bool
feature_pre_filter::Bool
pre_partition::Bool
two_round::Bool
header::Bool
label_column::String
weight_column::String
ignore_column::String
categorical_feature::Vector{Int}
forcedbins_filename::String
precise_float_parser::Bool

# Predict parameters
start_iteration_predict::Int
Expand Down Expand Up @@ -170,6 +178,7 @@ end
monotone_constraints_method = "basic",
monotone_penalty = 0.,
feature_contri = Float64[],
forcedsplits_filename = "",
refit_decay_rate = 0.9,
cegb_tradeoff = 1.0,
cegb_penalty_split = 0.,
Expand All @@ -190,7 +199,14 @@ end
zero_as_missing = false,
feature_pre_filter = true,
pre_partition = false,
two_round = false,
header = false,
label_column = "",
weight_column = "",
ignore_column = "",
categorical_feature = Int[],
forcedbins_filename = "",
precise_float_parser = false,
start_iteration_predict = 0,
num_iteration_predict = -1,
predict_raw_score = false,
Expand Down Expand Up @@ -271,6 +287,7 @@ function LGBMRegression(;
monotone_constraints_method = "basic",
monotone_penalty = 0.,
feature_contri = Float64[],
forcedsplits_filename = "",
refit_decay_rate = 0.9,
cegb_tradeoff = 1.0,
cegb_penalty_split = 0.,
Expand All @@ -291,7 +308,14 @@ function LGBMRegression(;
zero_as_missing = false,
feature_pre_filter = true,
pre_partition = false,
two_round = false,
header = false,
label_column = "",
weight_column = "",
ignore_column = "",
categorical_feature = Int[],
forcedbins_filename = "",
precise_float_parser = false,
start_iteration_predict = 0,
num_iteration_predict = -1,
predict_raw_score = false,
Expand Down Expand Up @@ -327,14 +351,17 @@ function LGBMRegression(;
extra_trees, extra_seed, early_stopping_round, first_metric_only, max_delta_step,
lambda_l1, lambda_l2, linear_lambda, min_gain_to_split, drop_rate, max_drop, skip_drop,
xgboost_dart_mode, uniform_drop, drop_seed, top_rate, other_rate, min_data_per_group, max_cat_threshold,
cat_l2, cat_smooth, max_cat_to_onehot, top_k, monotone_constraints, monotone_constraints_method, monotone_penalty, feature_contri, refit_decay_rate,
cat_l2, cat_smooth, max_cat_to_onehot, top_k, monotone_constraints, monotone_constraints_method, monotone_penalty,
feature_contri, forcedsplits_filename, refit_decay_rate,
cegb_tradeoff, cegb_penalty_split, cegb_penalty_feature_lazy, cegb_penalty_feature_coupled, path_smooth, interaction_constraints, verbosity,
linear_tree, max_bin, max_bin_by_feature, min_data_in_bin, bin_construct_sample_cnt, data_random_seed,
is_enable_sparse, enable_bundle, use_missing, zero_as_missing, feature_pre_filter, pre_partition, categorical_feature,
is_enable_sparse, enable_bundle, use_missing, zero_as_missing, feature_pre_filter, pre_partition,
two_round, header, label_column, weight_column, ignore_column, categorical_feature, forcedbins_filename, precise_float_parser,
start_iteration_predict, num_iteration_predict, predict_raw_score, predict_leaf_index, predict_contrib, predict_disable_shape_check,
1, is_unbalance, boost_from_average, reg_sqrt, alpha, fair_c, poisson_max_delta_step, tweedie_variance_power,
metric, metric_freq, is_provide_training_metric, eval_at,
num_machines, local_listen_port, time_out, machine_list_filename, machines, gpu_platform_id, gpu_device_id, gpu_use_dp, num_gpu,
num_machines, local_listen_port, time_out, machine_list_filename, machines,
gpu_platform_id, gpu_device_id, gpu_use_dp, num_gpu,
)
end

Expand Down Expand Up @@ -397,6 +424,7 @@ mutable struct LGBMClassification <: LGBMEstimator
monotone_constraints_method::String
monotone_penalty::Float64
feature_contri::Vector{Float64}
forcedsplits_filename::String
refit_decay_rate::Float64
cegb_tradeoff::Float64
cegb_penalty_split::Float64
Expand All @@ -419,7 +447,14 @@ mutable struct LGBMClassification <: LGBMEstimator
zero_as_missing::Bool
feature_pre_filter::Bool
pre_partition::Bool
two_round::Bool
header::Bool
label_column::String
weight_column::String
ignore_column::String
categorical_feature::Vector{Int}
forcedbins_filename::String
precise_float_parser::Bool

# Predict parameters
start_iteration_predict::Int
Expand Down Expand Up @@ -514,6 +549,7 @@ end
monotone_constraints_method = "basic",
monotone_penalty = 0.,
feature_contri = Float64[],
forcedsplits_filename = "",
refit_decay_rate = 0.9,
cegb_tradeoff = 1.0,
cegb_penalty_split = 0.,
Expand All @@ -534,7 +570,14 @@ end
zero_as_missing = false,
feature_pre_filter = true,
pre_partition = false,
two_round = false,
header = false,
label_column = "",
weight_column = "",
ignore_column = "",
categorical_feature = Int[],
forcedbins_filename = "",
precise_float_parser = false,
start_iteration_predict = 0,
num_iteration_predict = -1,
predict_raw_score = false,
Expand Down Expand Up @@ -620,6 +663,7 @@ function LGBMClassification(;
monotone_constraints_method = "basic",
monotone_penalty = 0.,
feature_contri = Float64[],
forcedsplits_filename = "",
refit_decay_rate = 0.9,
cegb_tradeoff = 1.0,
cegb_penalty_split = 0.,
Expand All @@ -640,7 +684,14 @@ function LGBMClassification(;
zero_as_missing = false,
feature_pre_filter = true,
pre_partition = false,
two_round = false,
header = false,
label_column = "",
weight_column = "",
ignore_column = "",
categorical_feature = Int[],
forcedbins_filename = "",
precise_float_parser = false,
start_iteration_predict = 0,
num_iteration_predict = -1,
predict_raw_score = false,
Expand Down Expand Up @@ -675,19 +726,20 @@ function LGBMClassification(;
return LGBMClassification(
Booster(), "", objective, boosting, num_iterations, learning_rate, num_leaves, tree_learner, num_threads, device_type, seed, deterministic,
force_col_wise, force_row_wise, histogram_pool_size, max_depth, min_data_in_leaf, min_sum_hessian_in_leaf,
bagging_fraction, pos_bagging_fraction, neg_bagging_fraction,bagging_freq, bagging_seed,
feature_fraction, feature_fraction_bynode, feature_fraction_seed, extra_trees, extra_seed, early_stopping_round, first_metric_only, max_delta_step,
lambda_l1, lambda_l2, linear_lambda, min_gain_to_split, drop_rate, max_drop, skip_drop,
xgboost_dart_mode, uniform_drop, drop_seed, top_rate, other_rate, min_data_per_group, max_cat_threshold,
cat_l2, cat_smooth, max_cat_to_onehot, top_k, monotone_constraints, monotone_constraints_method, monotone_penalty,
feature_contri, refit_decay_rate, cegb_tradeoff, cegb_penalty_split, cegb_penalty_feature_lazy, cegb_penalty_feature_coupled, path_smooth, interaction_constraints, verbosity,
bagging_fraction, pos_bagging_fraction, neg_bagging_fraction, bagging_freq, bagging_seed,
feature_fraction, feature_fraction_bynode, feature_fraction_seed, extra_trees, extra_seed, early_stopping_round, first_metric_only, max_delta_step, lambda_l1, lambda_l2, linear_lambda,
min_gain_to_split, drop_rate, max_drop, skip_drop, xgboost_dart_mode, uniform_drop, drop_seed, top_rate, other_rate,
min_data_per_group, max_cat_threshold, cat_l2, cat_smooth, max_cat_to_onehot, top_k, monotone_constraints, monotone_constraints_method, monotone_penalty,
feature_contri, forcedsplits_filename, refit_decay_rate, cegb_tradeoff, cegb_penalty_split, cegb_penalty_feature_lazy, cegb_penalty_feature_coupled,
path_smooth, interaction_constraints, verbosity,
linear_tree, max_bin, max_bin_by_feature, min_data_in_bin, bin_construct_sample_cnt, data_random_seed,
is_enable_sparse, enable_bundle, use_missing, zero_as_missing, feature_pre_filter, pre_partition, categorical_feature,
is_enable_sparse, enable_bundle, use_missing, zero_as_missing, feature_pre_filter, pre_partition,
two_round, header, label_column, weight_column, ignore_column, categorical_feature, forcedbins_filename, precise_float_parser,
start_iteration_predict, num_iteration_predict, predict_raw_score, predict_leaf_index, predict_contrib, predict_disable_shape_check, pred_early_stop, pred_early_stop_freq, pred_early_stop_margin,
num_class, is_unbalance, scale_pos_weight, sigmoid, boost_from_average,
metric, metric_freq, is_provide_training_metric, eval_at, multi_error_top_k, auc_mu_weights,
num_machines, local_listen_port, time_out, machine_list_filename,
machines, gpu_platform_id, gpu_device_id, gpu_use_dp, num_gpu,
num_machines, local_listen_port, time_out, machine_list_filename, machines,
gpu_platform_id, gpu_device_id, gpu_use_dp, num_gpu,
)
end

Expand Down Expand Up @@ -749,6 +801,7 @@ mutable struct LGBMRanking <: LGBMEstimator
monotone_constraints_method::String
monotone_penalty::Float64
feature_contri::Vector{Float64}
forcedsplits_filename::String
refit_decay_rate::Float64
cegb_tradeoff::Float64
cegb_penalty_split::Float64
Expand All @@ -771,8 +824,15 @@ mutable struct LGBMRanking <: LGBMEstimator
zero_as_missing::Bool
feature_pre_filter::Bool
pre_partition::Bool
two_round::Bool
header::Bool
label_column::String
weight_column::String
group_column::String
ignore_column::String
categorical_feature::Vector{Int}
forcedbins_filename::String
precise_float_parser::Bool

# Predict parameters
start_iteration_predict::Int
Expand Down Expand Up @@ -869,6 +929,7 @@ end
monotone_constraints_method = "basic",
monotone_penalty = 0.,
feature_contri = Float64[],
forcedsplits_filename = "",
refit_decay_rate = 0.9,
cegb_tradeoff = 1.0,
cegb_penalty_split = 0.,
Expand All @@ -889,8 +950,16 @@ end
zero_as_missing = false,
feature_pre_filter = true,
pre_partition = false,
two_round = false,
header = false,
label_column = "",
weight_column = "",
group_column = "",
ignore_column = "",
group_column = "",
categorical_feature = Int[],
forcedbins_filename = "",
precise_float_parser = false,
start_iteration_predict = 0,
num_iteration_predict = -1,
predict_raw_score = false,
Expand Down Expand Up @@ -978,6 +1047,7 @@ function LGBMRanking(;
monotone_constraints_method = "basic",
monotone_penalty = 0.,
feature_contri = Float64[],
forcedsplits_filename = "",
refit_decay_rate = 0.9,
cegb_tradeoff = 1.0,
cegb_penalty_split = 0.,
Expand All @@ -998,8 +1068,15 @@ function LGBMRanking(;
zero_as_missing = false,
feature_pre_filter = true,
pre_partition = false,
two_round = false,
header = false,
label_column = "",
weight_column = "",
group_column = "",
ignore_column = "",
categorical_feature = Int[],
forcedbins_filename = "",
precise_float_parser = false,
start_iteration_predict = 0,
num_iteration_predict = -1,
predict_raw_score = false,
Expand Down Expand Up @@ -1036,14 +1113,15 @@ function LGBMRanking(;
return LGBMRanking(
Booster(), "", objective, boosting, num_iterations, learning_rate, num_leaves, tree_learner, num_threads, device_type, seed, deterministic,
force_col_wise, force_row_wise, histogram_pool_size, max_depth, min_data_in_leaf, min_sum_hessian_in_leaf,
bagging_fraction, pos_bagging_fraction, neg_bagging_fraction, bagging_freq,
bagging_seed, feature_fraction, feature_fraction_bynode, feature_fraction_seed, extra_trees, extra_seed, early_stopping_round, first_metric_only, max_delta_step, lambda_l1, lambda_l2, linear_lambda,
min_gain_to_split, drop_rate, max_drop, skip_drop,
xgboost_dart_mode, uniform_drop, drop_seed, top_rate, other_rate, min_data_per_group, max_cat_threshold,
cat_l2, cat_smooth, max_cat_to_onehot, top_k, monotone_constraints, monotone_constraints_method, monotone_penalty,
feature_contri, refit_decay_rate, cegb_tradeoff, cegb_penalty_split, cegb_penalty_feature_lazy, cegb_penalty_feature_coupled, path_smooth, interaction_constraints, verbosity,
bagging_fraction, pos_bagging_fraction, neg_bagging_fraction, bagging_freq, bagging_seed,
feature_fraction, feature_fraction_bynode, feature_fraction_seed, extra_trees, extra_seed, early_stopping_round, first_metric_only, max_delta_step, lambda_l1, lambda_l2, linear_lambda,
min_gain_to_split, drop_rate, max_drop, skip_drop, xgboost_dart_mode, uniform_drop, drop_seed, top_rate, other_rate,
min_data_per_group, max_cat_threshold, cat_l2, cat_smooth, max_cat_to_onehot, top_k, monotone_constraints, monotone_constraints_method, monotone_penalty,
feature_contri, forcedsplits_filename, refit_decay_rate, cegb_tradeoff, cegb_penalty_split, cegb_penalty_feature_lazy, cegb_penalty_feature_coupled,
path_smooth, interaction_constraints, verbosity,
linear_tree, max_bin, max_bin_by_feature, min_data_in_bin, bin_construct_sample_cnt, data_random_seed,
is_enable_sparse, enable_bundle, use_missing, zero_as_missing, feature_pre_filter, pre_partition, group_column, categorical_feature,
is_enable_sparse, enable_bundle, use_missing, zero_as_missing, feature_pre_filter, pre_partition,
two_round, header, label_column, weight_column, group_column, ignore_column, categorical_feature, forcedbins_filename, precise_float_parser,
start_iteration_predict, num_iteration_predict, predict_raw_score, predict_leaf_index, predict_contrib, predict_disable_shape_check, pred_early_stop, pred_early_stop_freq, pred_early_stop_margin,
objective_seed, num_class, is_unbalance, scale_pos_weight, sigmoid, boost_from_average, lambdarank_truncation_level, lambdarank_norm, label_gain,
metric, metric_freq, is_provide_training_metric, eval_at,
Expand Down
30 changes: 29 additions & 1 deletion src/fit.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,6 +31,9 @@ array that holds the validation metric's value at each iteration.
* or
* `train_dataset::Dataset`: prepared train_dataset
* `test_datasets::Vector{Dataset}`: (optional) prepared test_datasets
* or
* `train_filepath::String`: path to the training data file.
* `test_filepath::String`: (optional) path to the test data file.
## Keyword Arguments
* `verbosity::Integer`: keyword argument that controls LightGBM's verbosity. `< 0` for fatal logs
only, `0` includes warning logs, `1` includes info logs, and `> 1` includes debug logs.
Expand Down Expand Up @@ -108,13 +111,38 @@ function fit!(
end


# Pass filepaths and set Dataset parameters including label, group and weights via the estimator.
function fit!(
estimator::LGBMEstimator, train_filepath::String;
test_filepath::String = "",
verbosity::Integer = nothing,
truncate_booster::Bool=true,
)

verbosity = isnothing(verbosity) ? estimator.verbosity : verbosity
log_debug(verbosity, "Started creating LGBM training dataset\n")
ds_parameters = stringifyparams(estimator)
train_ds = dataset_constructor(train_filepath, ds_parameters)

test_dss = []

if test_filepath != ""
test_ds = dataset_constructor(test_filepath, ds_parameters, train_ds)
push!(test_dss, test_ds)
end

return fit!(estimator, train_ds, test_dss..., verbosity=verbosity, truncate_booster=truncate_booster)
end


dataset_constructor(mat::Matrix, params::String, rm::Bool, ds::Dataset) = LGBM_DatasetCreateFromMat(mat, params, ds, rm)
dataset_constructor(mat::Matrix, params::String, rm::Bool) = LGBM_DatasetCreateFromMat(mat, params, rm)
dataset_constructor(mat::SparseArrays.SparseMatrixCSC, params::String, rm::Bool) = LGBM_DatasetCreateFromCSC(mat, params)
dataset_constructor(mat::SparseArrays.SparseMatrixCSC, params::String, rm::Bool, ds::Dataset) = LGBM_DatasetCreateFromCSC(mat, params, ds)
dataset_constructor(mat::AbstractMatrix, p::String, r::Bool, d::Dataset) = throw(TypeError(:fit!, Union{SparseArrays.SparseMatrixCSC, Matrix}, mat))
dataset_constructor(mat::AbstractMatrix, p::String, r::Bool) = throw(TypeError(:fit!, Union{SparseArrays.SparseMatrixCSC, Matrix}, mat))

dataset_constructor(filepath::String, params::String, ds::Dataset) = LGBM_DatasetCreateFromFile(filepath, params, ds)
dataset_constructor(filepath::String, params::String) = LGBM_DatasetCreateFromFile(filepath, params)

function train!(
estimator::LGBMEstimator,
Expand Down
Loading
Loading