Pearu Peterson

"@coroutine" decorator is deprecated since Python 3.8, use "async def" instead

10

When importing `thriftpy2.contrib.aio` modules with Python 3.8 or newer, a large number of deprecation warnings is trigger. Reproducer: ``` Python 3.8.6 | packaged by conda-forge | (default, Oct 7 2020,...

When in conda environment with all pymapd dependencies installed, ``` python setup.py develop ``` unnecessarily triggers installing the dependencies via pip. As a general but possibly complicated fix is to...

Remove bitarray, bittools and lzw.py

2

Bittools should be moved to a new project. Bitarray can be just deleted as it is maintained elsewhere.

Add sparse COO tensor support to torch.sum(dim=..., keepdim=...)

3

Fixes #92757, #86232 Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * __->__ #92979 cc @alexsamardzic @nikitaved @cpuhrsch @amjames @bhosmer

from_dlpack on an object not implementing DLPack protocol throws irrelevant error message

6

### Description As in the title. ### To Reproduce ```py >>> import cupy >>> cupy.from_dlpack(object()) Traceback (most recent call last): File "", line 1, in File "cupy/_core/dlpack.pyx", line 344, in...

Fix printing sparse COO meta tensors

1

As in the title. Also, add tests for printing sparse COO/CSR/CSC/BSR/BSC meta tensors. The PR fixes the following failure: ``` >>> import torch >>> indices = torch.empty((2, 1), device='meta', dtype=torch.int64)...

Add meta device support to sparse compressed tensors

2

As in the title. Unblocks https://github.com/pytorch/pytorch/pull/117907#discussion_r1499251745 Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * #120562 * __->__ #120498 cc @alexsamardzic @nikitaved @cpuhrsch @amjames @bhosmer @jcaip

Fix closeness tests in testOnComplexPlane for small inputs. Update expm1 inaccuracy regions.

5

As in the title. https://github.com/openxla/xla/pull/10376 fixes all inaccuracy issues in jax.numpy.expm1: - [before this PR](https://github.com/pearu/complex_function_validation/blob/5756d0e5b303bb7b084f18f612f69ed3bf5547a2/mpmath_jax_results/data/expm1_MPMath_complex128_cpu_versus_JAX_complex64_cpu.txt) the inaccuracies and mismatch rate was 17.7% - [after this PR](https://github.com/pearu/complex_function_validation/blob/5756d0e5b303bb7b084f18f612f69ed3bf5547a2/mpmath_jax_results_dev/data/expm1_MPMath_complex128_cpu_versus_JAX_complex64_cpu.txt) the inaccuracies rate is...

As in the title, follow-up to https://github.com/google/jax/pull/20896. Required by https://github.com/openxla/xla/pull/8589