Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Improved gpu_hist_experimental algorithm (#2866)

Browse Source 
- Implement colsampling, subsampling for gpu_hist_experimental

 - Optimised multi-GPU implementation for gpu_hist_experimental

 - Make nccl optional

 - Add Volta architecture flag

 - Optimise RegLossObj

 - Add timing utilities for debug verbose mode

 - Bump required cuda version to 8.0
This commit is contained in:
Rory Mitchell
2017-11-11 13:58:40 +13:00
committed by GitHub
parent 16c63f30d0
commit 40c6e2f0c8
14 changed files with 855 additions and 473 deletions
Download Patch File Download Diff File Expand all files Collapse all files

40
tests/python-gpu/test_gpu_updaters.py
View File

@@ -8,6 +8,7 @@ import numpy as np
import unittest
from nose.plugins.attrib import attr
from sklearn.datasets import load_digits, load_boston, load_breast_cancer, make_regression
import itertools as it
rng = np.random.RandomState(1994)
@@ -15,8 +16,9 @@ rng = np.random.RandomState(1994)
def non_increasing(L, tolerance):
return all((y - x) < tolerance for x, y in zip(L, L[1:]))
#Check result is always decreasing and final accuracy is within tolerance
def assert_accuracy(res, tree_method, comparison_tree_method, tolerance):
# Check result is always decreasing and final accuracy is within tolerance
def assert_accuracy(res, tree_method, comparison_tree_method, tolerance, param):
assert non_increasing(res[tree_method], tolerance)
assert np.allclose(res[tree_method][-1], res[comparison_tree_method][-1], 1e-3, 1e-2)
@@ -26,13 +28,14 @@ def train_boston(param_in, comparison_tree_method):
dtrain = xgb.DMatrix(data.data, label=data.target)
param = {}
param.update(param_in)
param['max_depth'] = 2
res_tmp = {}
res = {}
num_rounds = 10
xgb.train(param, dtrain, num_rounds, [(dtrain, 'train')], evals_result=res_tmp)
bst = xgb.train(param, dtrain, num_rounds, [(dtrain, 'train')], evals_result=res_tmp)
res[param['tree_method']] = res_tmp['train']['rmse']
param["tree_method"] = comparison_tree_method
xgb.train(param, dtrain, num_rounds, [(dtrain, 'train')], evals_result=res_tmp)
bst = xgb.train(param, dtrain, num_rounds, [(dtrain, 'train')], evals_result=res_tmp)
res[comparison_tree_method] = res_tmp['train']['rmse']
return res
@@ -92,17 +95,24 @@ def train_sparse(param_in, comparison_tree_method):
return res
# Enumerates all permutations of variable parameters
def assert_updater_accuracy(tree_method, comparison_tree_method, variable_param, tolerance):
param = {'tree_method': tree_method}
for k, set in variable_param.items():
for val in set:
param_tmp = param.copy()
param_tmp[k] = val
print(param_tmp, file=sys.stderr)
assert_accuracy(train_boston(param_tmp, comparison_tree_method), tree_method, comparison_tree_method, tolerance)
assert_accuracy(train_digits(param_tmp, comparison_tree_method), tree_method, comparison_tree_method, tolerance)
assert_accuracy(train_cancer(param_tmp, comparison_tree_method), tree_method, comparison_tree_method, tolerance)
assert_accuracy(train_sparse(param_tmp, comparison_tree_method), tree_method, comparison_tree_method, tolerance)
param = {'tree_method': tree_method }
names = sorted(variable_param)
combinations = it.product(*(variable_param[Name] for Name in names))
for set in combinations:
print(names, file=sys.stderr)
print(set, file=sys.stderr)
param_tmp = param.copy()
for i, name in enumerate(names):
param_tmp[name] = set[i]
print(param_tmp, file=sys.stderr)
assert_accuracy(train_boston(param_tmp, comparison_tree_method), tree_method, comparison_tree_method, tolerance, param_tmp)
assert_accuracy(train_digits(param_tmp, comparison_tree_method), tree_method, comparison_tree_method, tolerance, param_tmp)
assert_accuracy(train_cancer(param_tmp, comparison_tree_method), tree_method, comparison_tree_method, tolerance, param_tmp)
assert_accuracy(train_sparse(param_tmp, comparison_tree_method), tree_method, comparison_tree_method, tolerance, param_tmp)
@attr('gpu')
@@ -116,5 +126,5 @@ class TestGPU(unittest.TestCase):
assert_updater_accuracy('gpu_exact', 'exact', variable_param, 0.02)
def test_gpu_hist_experimental(self):
variable_param = {'max_depth': [2, 6], 'max_leaves': [255, 4], 'max_bin': [2, 16, 1024]}
variable_param = {'n_gpus': [1, -1], 'max_depth': [2, 6], 'max_leaves': [255, 4], 'max_bin': [2, 16, 1024]}
assert_updater_accuracy('gpu_hist_experimental', 'hist', variable_param, 0.01)