Pearu Peterson

Require nnz==0 in sparse meta tensors

1

As in the title and per discussion starting at https://github.com/pytorch/pytorch/pull/117907#issuecomment-2082426468 Stack from [ghstack](https://github.com/ezyang/ghstack) (oldest at bottom): * __->__ #125221

mp.asin returns incorrect results at complex infinities

8

As in the title. Reproducer (using 1.4.0a1.dev43+g0d287bf): ```python >>> z = complex(mpmath.inf, mpmath.inf); mpmath.mp.asin(z), mpmath.fp.asin(z) # mp.asin vs expected (mpc(real='nan', imag='+inf'), (0.7853981633974483+infj)) >>> z = complex(mpmath.inf, -mpmath.inf); mpmath.mp.asin(z), mpmath.fp.asin(z) #...

mp.asin is inaccurate for small complex inputs

3

As in the title. Reproducer: ```python >>> mpmath.mp.asin(mpmath.mpc(0, 1e-22)) mpc(real='0.0', imag='0.0') >>> mpmath.fp.asin(mpmath.mpc(0, 1e-22)) # expected result 1e-22j ``` A workaround is to increase precision: ```python >>> mpmath.mp.prec=80 >>> mpmath.mp.asin(mpmath.mpc(0,...

RFC-0004: Adding fill value property to PyTorch sparse tensors

14

This proposal introduces a fill value property to PyTorch sparse tensors that generalizes the current interpretation of unspecified elements from zero value to any value, including an indefinite fill value...

RFC-0003: Roadmap for PyTorch Sparse Tensors [WIP]

7

Update: this PR has grown very large, if you want to navigate it: - Look at the `.md` files at the end of the diff - Here is a rendered...

This rfc proposes a work plan for implementing Pytorch CSR storage format and introducing a new storage format Dimensions Mapping (DM) that allows extending CSR (and not only) to arbitrary...

Improve the accuracy of asin(x) and asinh(x) for complex x using modified Hull et al algorithm.

15

As in the title. ~Fixes https://github.com/openxla/xla/issues/8553~ - PR https://github.com/openxla/xla/pull/9802 disabled the fix. Update: the fix to https://github.com/openxla/xla/issues/8553 will be available via https://github.com/openxla/stablehlo/pull/2357

test_externals_cmath.py::test_external_cmath_omnisci tests are slow

5

As in title. The reason is that each subtest compiles its own function to an independent LLVM module. To speed up the tests, all functions should be compiled into a...

Support strings

6

Perhaps already have string dictionary support by inputting key_for_string, but could make nicer ux around this

Clean up when heavydb 6.0 is released

1

https://github.com/xnd-project/rbc/blob/1efb02691a751e5a69e3f256b29c6f51e93c40b8/rbc/omniscidb.py#L227-L230