GitLab

Adding FilterBasis datastructure to contain information about the type of basis used in disco convolution

adding local SNO architecture and updating trraining script for SWE prediction with powerspectrum computation

This small fix removes the maximum version number requirement of torch harmonics

* safeguarding all custom CUDA and C++ routines via the _cuda_extension_available flag

* bumping up version number

* added analytic gradients to the gradient_analysis notebook

* fixing sht unittest to not check the roundtrip gradient but sht and isht individually

* Updated changelog

* changing distributeds SHT to use dim=-3 as the channel dimension for distributed transpose

* Fixed formatting in tests

* Adding bash script for running test suite

* Replaced asserts regarding number of dimensions in tensor with checks

* Adding spherical upsampling routine and example notebook for resampling

- doing manual type conversion in custom autograd
- fixing stride issues in backward pass by making some output tensors contiguous

* AMP hotfix

* Bumping up version to 0.7.1

* reworked the setup system to use pyproject.toml

* changed the way custom CUDA extension build flag is passed

* reworked the CI

* changed to source distribution only in the CI

* reverting deploy script

* moving setuptools install

* removing requirements

* bumping up version

* bumping up the torch-harmonics version to 0.7.0

* updating amp to new torch.amp
* using amp autocrats to FP32 for disco convolution kernels
* implemented reduce_scatter routines but disabled those because of memory fluctuations which can cause OOM on big networks

* missed one instance of python3

* fused multiplication of quadrature into multiplication of the Psi tensor

* Added quadrature change to changelog

* removing persistent quad weights tensor as it isn't needed anymore

* added pretty-prenting for convolution modules

* adjusting default value in convolution test

* removing unneccessary contiguous statements

* replacing reduce scatter with independent reduce and scatter

* adding cuda kernels for disco conv

* making psi_idx an attribute

* adding license headers

* adding author files

* reorganizing files

* draft implementation

* added conditional installation to setup.py

* formatting changes

* removing triton kernel in DISCO convolution

* updated github actions

* updated Readme and changelog

* adding another guard for the cuda installation

* renaming the  cuda extension

* simplifying setup.py

* minor bugfix

* Bbonev/cuda disco cleanup (#32)

* cleanup of disco convolutions based on CUDA extension

* fixing unittest

* changing version to experimental 0.7.0a

* initial rewrite of the distributed convolution with CUDA

* fixing streams

* need to fix install options

* fixing streams

* undid setup.py changes

* reset setup.py

* including CUDAStream

* adjusted the precomputation of theta_cutoff. If you rely on this, your models will not be backwards-compatible.

* adjusting theta_cutoff in the unittest

* adding newly refactored kernels for faster compile

* Tkurth/cuda disco distributed fix (#34)

* attempt to make disco distributed

* working distributed convolutions

* fixing distributed conv

* working distributed disco

* removing irrelevant extra argument

* using stream functions from at instead of c10

* using stream functions from at instead of c10, small fix

* Bbonev/disc even filters (#35)

* initial working commit with new convention of counting collocation points across the diameter instead of across the radius

* fixed a bug in the computation of the even kernels

* changing heuristic for computing theta_cutoff

* Fixing unittest

* Readability improvements

* reworked normalization of filter basis functions

* implemented discrete normalization of disco filters

* relaxing tolerances in convolution unit test

* bugfix to correctly support unequal scale factors in latitudes and longitudes

* hotfix to a bug in the imports

* Bbonev/distributed disco refactor (#37)

* cleaned up normalization code in convolution

* formatting changes in distributed convolution

* Fixing default theta_cutoff to be the same in distributed and local case

* fixed distributed convolution to support the same normalization as non-distributed one

* readability improvements

* fixed initial scale of convolution parameter weights and fixed naming of the normalization routine

* Updated Readme.md

* added comment in Dockerfile regarding older architectures

---------
Co-authored-by: Thorsten Kurth <tkurth@nvidia.com>
Co-authored-by: Boris Bonev <bbonev@nvidia.com>

* initial implementation of distributed DISCO layer

* working distributed convolution

* working refactored serial conv transpose with torch kernel

* working distributed conv and transposed conv when using the python kernel

* working distributed convolution with torch kernel

* fixed triton kernel tests

* adding print statement to debug CI

* adjusting tolerances in local convolution unittest

---------
Co-authored-by: Boris Bonev <bbonev@nvidia.com>

* Cleaned up DISCO convolutions

Changed the code to only implicitly use sparse tensors in the modules, in order to enable compatibility with DDP