[CUDA] "Warning: 'Using sparse features with CUDA is currently not supported' in LightGBM with GPU"

[sgapple]

Description

When attempting to use LightGBM with GPU support, I receive the warning message:

"Using sparse features with CUDA is currently not supported."

This occurs even though I am providing a dense dataset to the model. I believe this results in a fallback to CPU, which leads to slower training times.

Reproducible example

import numpy as np
from sklearn.model_selection import train_test_split
import lightgbm as lgb

# Set seed for reproducibility
np.random.seed(42)

n_samples = 500 * 10000
n_features = 51

X = np.random.rand(n_samples, n_features).astype(np.float32)
y = np.random.rand(n_samples).astype(np.float32)

X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42
)

X_train = np.ascontiguousarray(X_train, dtype=np.float32)
y_train = np.ascontiguousarray(y_train, dtype=np.float32)

new_lgb_train = lgb.Dataset(X_train, label=y_train)

cuda_params = {
    'objective': 'regression',
    'boosting_type': 'dart',
    'colsample_bytree': 0.7,
    'learning_rate': 0.01,
    'max_depth': 7,
    'subsample': 0.7,
    'n_jobs': 32,
    'num_leaves': 63,
    'verbose': 1,
    'device': 'cuda',
    'force_row_wise': True
}

gbm_cuda = lgb.train(cuda_params,
                new_lgb_train,
                num_boost_round=100)

gbm_cuda.predict(X_test)

Environment info

Docker:

• nvidia/cuda:12.4.1-cudnn-devel-ubuntu22.04
• CUDA version: 12.4

Host:

• ubuntu22.04
• GPU: NVIDIA GeForce RTX 3090
• CPU: intel i9
• Driver Version: 550.127.05

LightGBM version or commit hash: LightGBM version: 4.5.0

Command(s) you used to install LightGBM

FROM nvidia/cuda:12.4.1-cudnn-devel-ubuntu22.04

RUN apt-get update && \
apt-get install -y --no-install-recommends \
build-essential \
curl \
bzip2 \
ca-certificates \
libglib2.0-0 \
libxext6 \
libsm6 \
libxrender1 \
git \
gnupg \
swig \
vim \
mercurial \
subversion \
python3-dev \
python3-pip \
python3-setuptools \
ocl-icd-opencl-dev \
cmake \
libboost-dev \
libboost-system-dev \
libboost-filesystem-dev \
gcc \
g++ && \
# Install Node.js 18.x
curl -fsSL https://deb.nodesource.com/setup_18.x | bash - && \
apt-get update && \
apt-get install -y nodejs=18.20.4*


# Add OpenCL ICD files for LightGBM
RUN mkdir -p /etc/OpenCL/vendors && \
    echo "libnvidia-opencl.so.1" > /etc/OpenCL/vendors/nvidia.icd

RUN pip3 install --upgrade pip
RUN pip3 install --no-binary lightgbm --config-settings=cmake.define.USE_CUDA=ON lightgbm

Additional Comments

I've tried several installation methods, including cloning from source and using pip for the GPU and CUDA versions. One of the methods failed, while the other resulted in the same warning about sparse features, even with a dense dataset. I would appreciate your assistance, as I am looking to perform an accurate performance comparison between different boosters. Thank you

[tlongatalaya]

I am also having this issue.

[freshlu11]

me too

[pr1nce-harry]

i have the same issue

[shiyu1994]

The warning arises whenever is_enable_sparse=true, regardless of whether the actual features are dense or not. Thus, this does not result in a fallback to CPU. You may check your GPU utilization to ensure the training runs on GPU.

We may drop this warning soon since it is kind of misleading and annoying.

[github-actions]

This issue has been automatically closed because it has been awaiting a response for too long.
When you have time to to work with the maintainers to resolve this issue, please post a new comment and it will be re-opened. If the issue has been locked for editing by the time you return to it, please open a new issue and reference this one.
Thank you for taking the time to improve LightGBM!