dpnp

dpnp copied to clipboard

Set minimum supported NumPy version to `2.0.0`

2

- [ ] Have you provided a meaningful PR description? - [ ] Have you added a test, reproducer or referred to an issue with a reproducer? - [ ]...

antonwolfy

This throws an error: ```python import numpy as np import dpnp import pickle x = dpnp.array( np.sqrt(np.arange(10)).astype(np.float32), device="gpu", ) pickle.dumps(x) ``` ``` File :2, in dpctl.tensor._usmarray.usm_ndarray.__reduce_cython__() TypeError: no default __reduce__...

david-cortes-intel

Median 1d via sycl kernel

3

Draft PR for median implementation. * sycl kernels added, but cleanup is needed * kernels are not used in python yet * tests need to be added - [ ]...

AlexanderKalistratov

Cross-linking with the [issue](https://github.com/IntelPython/dpctl/issues/2138#issuecomment-3222990884) from dpctl for supporting `order` arg in `dpt.asnumpy` Reproducer: ```python import numpy as np import dpnp as dp print("dpnp version:", getattr(dp, "__version__", "unknown")) # Build the...

abagusetty

Follow ```__sycl_usm_array_interface__``` standard in ```asarray```

1

As of now, arrays which present the ```__sycl_usm_array_interface__``` analogous to the numpy ```__array__``` interface are copied, which is not the case in dpctl. This is problematic for scikit-learn-intelex, and ideally...

icfaust

Requesting a feature development for fusing the function to reuse cache and reduce kernel launch latency similar to [cupy.fuse](https://docs.cupy.dev/en/stable/reference/generated/cupy.fuse.html) decorator

abagusetty

The docs mention that device drivers will come preinstalled on windows, and gives a link to the windows drivers:  But it does not cover other topics - e.g.: *...

david-cortes-intel

numpy-2.3 added [`sort` parameter to](https://numpy.org/doc/stable/reference/generated/numpy.unique.html#:~:text=sortedbool%2C%20optional) `numpy.unique`, `dpnp` should align with it. In addition, the [behavior](https://numpy.org/doc/stable/reference/generated/numpy.unique_values.html#:~:text=Changed%20in%20version%202.3%3A%20The%20algorithm%20was%20changed%20to%20a%20faster%20one%20that%20does%20not%20rely%20on%20sorting%2C%20and%20hence%20the%20results%20are%20no%20longer%20implicitly%20sorted.) of `numpy.unique_value` is modified in numpy-2.3, `dpnp` should align here as well.

vtavana

[Question] dpnp functionality with switching device and context

1

I am working with a cupy project for Aurora and was wondering how to port with switching devices such as the following. Appreciate any pointers ```python with cupy.cuda.Device(device_id), _streams[device_id]: ```

abagusetty