heat

heat copied to clipboard

Refactoring of QR: try out pure stabilized Gram-Schmidt (split=1) and TS-QR (split=0)

3

This issue is to suggest refactoring of the code for QR decomposition. Currently, a quite elaborate algorithm is used for QR. The following suggestions might turn out to be actually...

mrfh92

[Bug]: New Pytorch Release workflow fails

4

### What happened? Manually triggered GH Action [Get latest PyTorch version main branch](https://github.com/helmholtz-analytics/heat/actions/workflows/pytorch-latest-main.yml) fails. ### Code snippet triggering the error _No response_ ### Error message or erroneous outcome ```shell /usr/bin/git...

ClaudiaComito

Refactoring of `communication`: separate `mpi4py`-wrappers from `DNDarrays`

2

Currently, the code wrapping `mpi4py` for `torch.Tensor`'s is contained in `heat/core/communication.py` and not fully separated from constructions related to the `DNDarray`-class (e.g., computation of chunk sizes is also introduced there)....

mrfh92

Implement distributed Fast Fourier Transform of 1-D DNDarray

1

**Related** #1097 #1218 **Feature functionality** The current (#1218) implementation of `ht.fft.fft()` is only distributed if the input DNDarray is n-dimensional (n>1). Research suitable algorithm and implement distributed 1-D FFT of...

ClaudiaComito

Convert and Load Image Data via Memory Mapped Files

2

**Related** This issue is complements Heat's Dataloader (implemented in datatools.py) and can be used alongside Heat's Dataset **Feature functionality** Enables the conversion of images stored in the Pytorch ImageFolder format...

krajsek

[Bug]: CI matrix fails with new `setup-mpi` 1.2.1

10

### What happened? The Python version vs. PyTorch version cross-checks we run after a PR is approved are failing. Something has changed in the jobs set up that we couldn't...

ClaudiaComito

Distributed convolution in arbitrary dimension and with arbitrary kernel size

2

It is well-known that convolution is (up to constants and conventions) the same as applying Fourier transform, a pointwise multiplication, and inverse Fourier transform. Since we will soon have a...

mrfh92

See #1104 for the general problem/the idea of batched linear algebra (focussing on the case of matmul for batches of matrices). This issue is intended to do the same for...

mrfh92

Implement consistent linear algebra for arrays with dimension > 2, in particular `matmul`

3

In Pytorch, most linear algebra routines can be applied to batches of matrices. In Heat it would be nice to have a similar rule how to "define" linear algebra (`matmul`,`qr`,`cholesky`,`svd`...

mrfh92

Module `manipulations`: replace `factories.array()` with `DNDarray` construct where possible

3

**Feature functionality** Will result in significant performance enhancement. Probably requires expanding the tests.

ClaudiaComito