test-profil-it-cuda fails w/ latest

[puririshi98]

🐛 Describe the bug

_____________________________ test_profileit_cuda ______________________________
get_dataset = functools.partial(<function load_dataset at 0xffff974227a0>, '/tmp/pyg_test_datasets')
    @onlyCUDA
    @onlyOnline
    @withPackage('pytorch_memlab')
    def test_profileit_cuda(get_dataset):
        warnings.filterwarnings('ignore', '.*arguments of DataFrame.drop.*')

        dataset = get_dataset(name='Cora')
        data = dataset[0].cuda()
        model = GraphSAGE(dataset.num_features, hidden_channels=64, num_layers=3,
                          out_channels=dataset.num_classes).cuda()
        optimizer = torch.optim.Adam(model.parameters(), lr=0.01)

        @profileit('cuda')
        def train(model, x, edge_index, y):
            model.train()
            optimizer.zero_grad()
            out = model(x, edge_index)
            loss = F.cross_entropy(out, y)
            loss.backward()
            return float(loss)

        stats_list = []
        for epoch in range(5):
            _, stats = train(model, data.x, data.edge_index, data.y)
            assert stats.time > 0
            assert stats.max_allocated_gpu > 0
            assert stats.max_reserved_gpu > 0
            assert stats.max_active_gpu > 0
            assert stats.nvidia_smi_free_cuda > 0
            assert stats.nvidia_smi_used_cuda > 0

            if epoch >= 2:  # Warm-up
                stats_list.append(stats)

        stats_summary = get_stats_summary(stats_list)
        assert stats_summary.time_mean > 0
        assert stats_summary.time_std > 0
        assert stats_summary.max_allocated_gpu > 0
        assert stats_summary.max_reserved_gpu > 0
        assert stats_summary.max_active_gpu > 0
>       assert stats_summary.min_nvidia_smi_free_cuda > 0
E       AttributeError: 'GPUStatsSummary' object has no attribute 'min_nvidia_smi_free_cuda'
test/profile/test_profile.py:86: AttributeError

Versions

latest stack as of 12 hours ago

[puririshi98]

https://github.com/pyg-team/pytorch_geometric/blob/master/torch_geometric/profile/profile.py#L34-L39 https://github.com/pyg-team/pytorch_geometric/commit/0e7458d21433f0191f50951dd2ff56e6aefa492d it seems the profiler definition was updated but the tests weren't. @DamianSzwichtenberg @rusty1s

[DamianSzwichtenberg]

Thanks for reporting 😉 Fix pushed in #8609. Actually, tests were good, the issue comes from get_stats_summary, where I should check first if the incoming data is of the lowest type in the hierarchy...