def test_glt_ogbnproducts()

in benchmarks/api/bench_feature.py [0:0]

• 33 lines of code
• 2 McCabe index (conditional complexity)

def test_glt_ogbnproducts(split_ratio):
  root = osp.join(osp.dirname(osp.dirname(osp.realpath(__file__))),
                  '..', 'data', 'products')
  dataset = PygNodePropPredDataset('ogbn-products', root)
  train_idx = dataset.get_idx_split()["train"]
  train_loader = torch.utils.data.DataLoader(train_idx,
                                             batch_size=1024,
                                             pin_memory=True,
                                             shuffle=True)
  csr_topo = glt.data.Topology(dataset[0].edge_index)

  g = glt.data.Graph(csr_topo, 'CUDA', device=0)
  device = torch.device('cuda:0')
  sampler = glt.sampler.NeighborSampler(g, [15, 10, 5], device=device)

  cpu_tensor, id2index = glt.data.sort_by_in_degree(
      dataset[0].x, split_ratio, csr_topo)
  feature = glt.data.Feature(cpu_tensor,
                             id2index,
                             split_ratio,
                             device_group_list=[glt.data.DeviceGroup(0, [0])],
                             device=0)
  total_num = 0
  total_time = 0
  for seeds in train_loader:
    nid = sampler.sample_from_nodes(seeds).node
    torch.cuda.synchronize()
    start = time.time()
    res = feature[nid]
    torch.cuda.synchronize()
    total_time += time.time() - start
    total_num += res.numel()
  torch.cuda.synchronize()
  print('Lookup {} ids, takes {} secs, Throughput {} GB/s.'\
    .format(total_num, total_time, total_num * 4 / total_time/ (1024**3)))