Rishi Puri comments

Results 44 comments of


                                            Rishi Puri

Improvements for Papers100m single gpu and single node multi gpu examples (Cugraph, GATConv, better default hyperparams, eval on all ranks)

we get up to 5x speed ups with cugraph depending on the system used

Improvements for Papers100m single gpu and single node multi gpu examples (Cugraph, GATConv, better default hyperparams, eval on all ranks)

in general GCN does better than GAT w/ about 45% test acc for GCN and slightly less for GAT. before the PR GCN was getting closer to 40% test acc

Improvements for Papers100m single gpu and single node multi gpu examples (Cugraph, GATConv, better default hyperparams, eval on all ranks)

need @alexbarghi-nv to mirror my work on eval/testing on all ranks: https://github.com/pyg-team/pytorch_geometric/pull/8173/commits/c86f852644b3455bdaea6ee8cf8264eed6c5465a for cugraph then we will be ready to merge this

Improvements for Papers100m single gpu and single node multi gpu examples (Cugraph, GATConv, better default hyperparams, eval on all ranks)

@akihironitta @rusty1s ready to merge

Improvements for Papers100m single gpu and single node multi gpu examples (Cugraph, GATConv, better default hyperparams, eval on all ranks)

needs a fix for cugraph multigpu from @alexbarghi-nv

Improvements for Papers100m single gpu and single node multi gpu examples (Cugraph, GATConv, better default hyperparams, eval on all ranks)

okay @akihironitta @rusty1s, the current code in this PR only works with in 24.02 or newer cugraph so i will circle back for merging this once that is released publicly

Improvements for Papers100m single gpu and single node multi gpu examples (Cugraph, GATConv, better default hyperparams, eval on all ranks)

@akihironitta @rusty1s anything needed to get this merged?

Improvements for Papers100m single gpu and single node multi gpu examples (Cugraph, GATConv, better default hyperparams, eval on all ranks)

@rusty1s lmk if anything needed to merge

CI failing for `test_dist_neighbor_sampler_hetero[False]`

same thing for this test: ``` test/distributed/test_dist_neighbor_sampler.py::test_dist_neighbor_sampler_hetero[True] Computing METIS partitioning... python: malloc.c:4302: _int_malloc: Assertion `(unsigned long) (size) >= (unsigned long) (nb)' failed. Session terminated, killing shell... ...killed. ``` i produced...

CI failing for `test_dist_neighbor_sampler_hetero[False]`

test/distributed/test_dist_neighbor_sampler.py::test_dist_neighbor_sampler_temporal_hetero[uniform-None] Computing METIS partitioning... double free or corruption (!prev) Session terminated, killing shell...******TIMEOUT EXPIRED******