LightGBMでGPUを使用するための環境構築

(.venv)> pip install --no-binary lightgbm lightgbm --config-settings=cmake.define.USE_GPU=ON --config-settings=cmake.define.BOOST_ROOT="C:/local/boost_1_85_0" --config-settings=cmake.define.BOOST_LIBRARYDIR="C:/local/boost_1_85_0/lib64-msvc-14.3"

(.venv)> pip install scikit-learn

import lightgbm as lgb
from sklearn.datasets import load_breast_cancer
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score

# データの読み込み
data = load_breast_cancer()
X_train, X_test, y_train, y_test = train_test_split(data.data, data.target, test_size=0.2, random_state=42)

# データセットの作成
train_data = lgb.Dataset(X_train, label=y_train)
test_data = lgb.Dataset(X_test, label=y_test, reference=train_data)

# パラメータ設定
params = {
    'boosting_type': 'gbdt',
    'objective': 'binary',
    'metric': 'binary_error',
    #'device': 'cpu',
    'device': 'gpu',
    'gpu_platform_id': 0,
    'gpu_device_id': 0,
}

# モデルのトレーニング
print("Training started...")
bst = lgb.train(params, train_data, valid_sets=[test_data], num_boost_round=100, callbacks=[lgb.early_stopping(10)])

# 予測
y_pred = bst.predict(X_test, num_iteration=bst.best_iteration)
y_pred = (y_pred > 0.5).astype(int)

# 精度の評価
accuracy = accuracy_score(y_test, y_pred)
print(f"Accuracy: {accuracy * 100:.2f}%")

(.venv)> python check_gpu.py
Training started...
[LightGBM] [Info] Number of positive: 286, number of negative: 169
[LightGBM] [Info] This is the GPU trainer!!
[LightGBM] [Info] Total Bins 4548
[LightGBM] [Info] Number of data points in the train set: 455, number of used features: 30
[LightGBM] [Info] Using requested OpenCL platform 0 device 0
[LightGBM] [Info] Using GPU Device: NVIDIA GeForce RTX 3060, Vendor: NVIDIA Corporation
[LightGBM] [Info] Compiling OpenCL Kernel with 256 bins...
[LightGBM] [Info] GPU programs have been built
[LightGBM] [Info] Size of histogram bin entry: 8
[LightGBM] [Info] 30 dense feature groups (0.01 MB) transferred to GPU in 0.003200 secs. 0 sparse feature groups
[LightGBM] [Info] [binary:BoostFromScore]: pavg=0.628571 -> initscore=0.526093
[LightGBM] [Info] Start training from score 0.526093
[LightGBM] [Warning] No further splits with positive gain, best gain: -inf
Training until validation scores don't improve for 10 rounds
[LightGBM] [Warning] No further splits with positive gain, best gain: -inf
[LightGBM] [Warning] No further splits with positive gain, best gain: -inf
[LightGBM] [Warning] No further splits with positive gain, best gain: -inf
[LightGBM] [Warning] No further splits with positive gain, best gain: -inf
[LightGBM] [Warning] No further splits with positive gain, best gain: -inf
[LightGBM] [Warning] No further splits with positive gain, best gain: -inf
[LightGBM] [Warning] No further splits with positive gain, best gain: -inf
[LightGBM] [Warning] No further splits with positive gain, best gain: -inf
[LightGBM] [Warning] No further splits with positive gain, best gain: -inf
[LightGBM] [Warning] No further splits with positive gain, best gain: -inf
[LightGBM] [Warning] No further splits with positive gain, best gain: -inf
[LightGBM] [Warning] No further splits with positive gain, best gain: -inf
[LightGBM] [Warning] No further splits with positive gain, best gain: -inf
[LightGBM] [Warning] No further splits with positive gain, best gain: -inf
[LightGBM] [Warning] No further splits with positive gain, best gain: -inf
[LightGBM] [Warning] No further splits with positive gain, best gain: -inf
[LightGBM] [Warning] No further splits with positive gain, best gain: -inf
[LightGBM] [Warning] No further splits with positive gain, best gain: -inf
Early stopping, best iteration is:
[9]     valid_0's binary_error: 0.0350877
Accuracy: 96.49%

[LightGBM] [Info] Using requested OpenCL platform 0 device 0
[LightGBM] [Info] Using GPU Device: NVIDIA GeForce RTX 3060, Vendor: NVIDIA Corporation
[LightGBM] [Info] Compiling OpenCL Kernel with 256 bins...
[LightGBM] [Info] GPU programs have been built

# パラメータ設定
params = {
    'boosting_type': 'gbdt',
    'objective': 'binary',
    'metric': 'binary_error',
    # 'device': 'cpu',
    'device': 'gpu',
    'gpu_platform_id': 0,
    'gpu_device_id': 0,
}

(.venv)> pip install pyopencl

import pyopencl as cl

# 全てのプラットフォームを取得
platforms = cl.get_platforms()

# 各プラットフォームの詳細を表示
for platform_id, platform in enumerate(platforms):
    print(f"Platform ID: {platform_id}, Name: {platform.name}")
    # 各プラットフォーム内のデバイスを取得
    devices = platform.get_devices()
    for device_id, device in enumerate(devices):
        print(f"  Device ID: {device_id}, Name: {device.name}, Type: {cl.device_type.to_string(device.type)}")

Platform ID: 0, Name: NVIDIA CUDA
  Device ID: 0, Name: NVIDIA GeForce RTX 3060, Type: ALL | GPU