Intel® Distribution for Python* Known Issues

Published: 06/02/2016   Last Updated: 09/29/2021

Intel Distribution for Python 2021.4 Release

No known issues found at this time.

Intel Distribution for Python 2021.3 Release

cpython socket does not have SO_REUSEPORT attribute

On some operating systems, cpython's socket module does not have SO_REUSEPORT attribute. A workaround is to instead use SO_BROADCAST to achieve the same effect.

Intel Distribution for Python 2021.2 Release

Conda installing may display clobber warnings

The files in the ca-certificates conda package used to be in the openssl conda package. Not having removed these files in the openssl package results in potential harmless clobber warnings.

Tampering with PS1 can corrupt command prompt

The conda backend relies on modifying the PS1 environment variable to properly display the activated conda environment in the command line. Any tampering with the variable can result in unexpected behavior when activating conda environments.

Intel Distribution for Python 2021.1 Release

DPNP may experience the occasional crash

The DPNP module can crash when an underlying library call is not implemented for a given device (e.g. sampling from gamma distribution is not implemented for GPU).
It can also crash with certain combinations of the DPCTL module.
Other DPNP issues can be found at https://github.com/IntelPython/dpnp/issues.

Numba’s automatic GPU offloading fails in certain corner cases

For certain corner cases, automatic GPU offloading fails and the code silently falls back to the CPU (refer to https://github.com/IntelPython/numba/issues/77). The issue has been fixed in current trunk of IntelPython/numba.

DPCTL copy implementation as blocking calls

DPCTL’s data copy operations are currently implemented as blocking calls.

For DPCTL, some functionality is not yet supported

For SYCL DPC++ CUDA and host device queues are not yet supported. The module is also not supported on macOS.

Scikit-learn may produce incorrect shuffling on Windows

The train_test_split in the daal4py backend of scikit-learn can potentially produce incorrect shuffling on the Windows platform

Some daal4py examples do not work on Intel® Iris Xe MAX with float64 compute mode 

The following daal4py examples do not work on Intel® Iris Xe MAX with float64 compute mode:

  • gradient_boosted_regression_batch
  • decision_forest_classification_batch
  • decision_forest_regression_batch
  • bf_knn_classification_batch
  • dbscan_batch
  • svm_batch
  • sklearn_sycl.py
  • kmeans_batch

Run daal4py examples using float32 compute mode instead:

  1. Use np.float32 data type for input data. To do this, add parameter t=np.float32 to the readcsv function used in the examples.
  2. Set the parameter fptype to float in the algorithm object constructor: fptype='float'.
  3. Switch on float64 software emulation on Intel® Iris Xe MAX

K-Means example in daal4py (examples/sycl/kmeans_batch.py) produces different results on GPU and CPU.

To avoid failures, comment assert statements that compare GPU results and classic results in the example.

DBSCAN example in daal4py (examples/sycl/dbscan_batch.py) hangs when it is running on CPU with data wrapped in sycl_buffer.

To avoid hangs, do not pass sycl_buffer objects to DBSCAN on CPU.

Product and Performance Information

1

Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.