f-dangel / curvlinops / 8317419815

Builds Branch Commit Type Ran Committer Via Coverage
8317419815 kfac-damping Merge nested if-statements Co-authored-by: sourcery-ai[bot] <58596630+sourcery-ai[bot]@users.noreply.github.com> push web-flow github
89.43
8317405133 kfac-damping Simplify if-statement Co-authored-by: sourcery-ai[bot] <58596630+sourcery-ai[bot]@users.noreply.github.com> push web-flow github
89.44
8309881347 kfac-damping Add heuristic and exact damping to KFACInverseLinearOperator push runame github
89.44
8301095449 kfac-torch-only Increase test tolerance Pull #91 runame github
89.1
8298085305 kfac-torch-only Nicer if-statement Co-authored-by: sourcery-ai[bot] <58596630+sourcery-ai[bot]@users.noreply.github.com> Pull #91 web-flow github
88.94
8297888358 kfac-torch-only Add torch_matmat and torch_matvec methods to KFACLinearOperator push runame github
88.94
8255504330 no-cpu [REF] Parallelize `sample_grad_output` in MC Fisher (#89) push web-flow github
89.12
8250604944 main [REF] Parallelize `sample_grad_output` in MC Fisher (#89) push web-flow github
89.12
8250294523 main [REF] Parallelize `grad_output` computation in empirical Fisher (#88) push web-flow github
89.12
8250276308 parallelize-sample-grad-output [REF] Parallelize `sample_grad_output` in MC Fisher Pull #89 f-dangel github
89.15
8249825057 parallelize-grad-output [REF] Parallelize `grad_output` computation in empirical Fisher Pull #88 f-dangel github
89.12
8248126639 main [REF] Remove for-loop over samples in MC Fisher (#85) push web-flow github
89.15
8248124531 main [REF] Improve efficiency of empirical Fisher (#84) Remove for-loop over data and use GGN-vector products of a pseudo-loss instead which parallelizes the loop. push web-flow github
89.06
8195475763 faster-fisher-mc [REF] Remove for-loop over samples in MC Fisher push f-dangel github
89.09
8193765593 faster-ef [REF] Improve efficiency of empirical Fisher Remove for-loop over data and use GGN-vector products of a pseudo-loss instead which parallelizes the loop. Pull #84 f-dangel github
89.06