GPU offload (Intel Arc A750) of RandomForestRegressor (5000 x 50) yields RuntimeError: Provided range is out of integer limits.

[hansonmcoombs]

GPU offload of RandomForestRegressor fails at a relativly low sample and dimensional level

Error

RuntimeError: Provided range is out of integer limits. Pass `-fno-sycl-id-queries-fit-in-int' to disable range check.

Minimum working example:

from sklearnex import patch_sklearn, config_context
patch_sklearn()
from sklearn.ensemble import RandomForestRegressor
import numpy as np


def rastrigin(x):
    x = np.atleast_2d(x).astype(float)
    return np.squeeze(np.sum(x * x - 10 * np.cos(2 * np.pi * x), axis=1) + 10 * np.size(x, axis=1))

def main():
    nd = 50
    nsamp = 5000
    X = np.random.uniform(-5.12, 5.12, size=(nsamp, nd))
    X = X.astype(np.float32)
    y = rastrigin(X)
    rforrest = RandomForestRegressor(n_jobs=-1, n_estimators=100, random_state=42)
    rforrest.fit(X, y)


if __name__ == '__main__':
    main()  # without gpu offload runs fine

    with config_context(target_offload="gpu"):
        main()  # fails: RuntimeError: Provided range is out of integer limits. Pass `-fno-sycl-id-queries-fit-in-int' to disable range check.

System specs

• AMD Ryzen 5 7600X
• Intel Arc A750
• Ubuntu 24.04.3 LTS

Conda environment (yml)

name: sklex
channels:
  - https://software.repos.intel.com/python/conda
  - conda-forge
  - defaults
dependencies:
  - _openmp_mutex=4.5=6_kmp_llvm
  - bzip2=1.0.8=hda65f42_8
  - ca-certificates=2025.11.12=hbd8a1cb_0
  - dal=2025.9.0=h9289deb_52
  - dpcpp-cpp-rt=2025.3.1=intel_760
  - dpcpp_cpp_rt=2025.3.1=intel_760
  - icu=75.1=he02047a_0
  - impi_rt=2021.16.0=h4715778_443
  - intel-cmplr-lib-rt=2025.3.1=intel_760
  - intel-cmplr-lib-ur=2025.3.1=intel_760
  - intel-cmplr-lic-rt=2025.3.1=intel_760
  - intel-gpu-ocl-icd-system=1.0.0=h6b64f20_2
  - intel-opencl-rt=2025.3.1=intel_760
  - intel-openmp=2025.3.1=intel_760
  - intel-sycl-rt=2025.3.1=intel_760
  - joblib=1.5.2=pyhd8ed1ab_0
  - ld_impl_linux-64=2.45=bootstrap_ha15bf96_3
  - libblas=3.11.0=2_h4a7cf45_openblas
  - libcblas=3.11.0=2_h0358290_openblas
  - libexpat=2.7.3=hecca717_0
  - libffi=3.5.2=h9ec8514_0
  - libgcc=15.2.0=h767d61c_7
  - libgcc-ng=15.2.0=h69a702a_7
  - libgfortran=15.2.0=h69a702a_7
  - libgfortran5=15.2.0=hcd61629_7
  - libhwloc=2.12.1=default_h7f8ec31_1002
  - libiconv=1.18=h3b78370_2
  - liblapack=3.11.0=2_h47877c9_openblas
  - liblzma=5.8.1=hb9d3cd8_2
  - libmpdec=4.0.0=hb9d3cd8_0
  - libopenblas=0.3.30=pthreads_h94d23a6_4
  - libsqlite=3.51.0=hee844dc_0
  - libstdcxx=15.2.0=h8f9b012_7
  - libstdcxx-ng=15.2.0=h4852527_7
  - libuuid=2.41.2=he9a06e4_0
  - libxml2=2.15.1=h26afc86_0
  - libxml2-16=2.15.1=ha9997c6_0
  - libzlib=1.3.1=hb9d3cd8_2
  - llvm-openmp=20.1.8=hc2b8361_0
  - mpi=1.0=impi
  - ncurses=6.5=h2d0b736_3
  - numpy=2.3.5=py313hf6604e3_0
  - openssl=3.6.0=h26f9b46_0
  - pip=25.3=pyh145f28c_0
  - python=3.13.9=hc97d973_101_cp313
  - python_abi=3.13=8_cp313
  - readline=8.2=h8c095d6_2
  - scikit-learn=1.7.2=py313h06d4379_0
  - scikit-learn-intelex=2025.9.0=py313hfd9b6fe_0
  - scipy=1.16.3=py313h11c21cd_1
  - setuptools=80.9.0=pyhff2d567_0
  - tbb=2022.3.0=h8d10470_1
  - tcm=1.4.1=intel_445
  - threadpoolctl=3.6.0=pyhecae5ae_0
  - tk=8.6.13=noxft_ha0e22de_103
  - tzdata=2025b=h78e105d_0
  - umf=1.0.2=intel_81