About Heterogeneous Graph Sampler

[DuanhaoranCC]

Hello author!
I'm using the heterogeneous graph sampler, but have some questions.

from torch_geometric.datasets import OGB_MAG
from torch_geometric.loader import NeighborLoader

data = OGB_MAG('./')[0]
loader = NeighborLoader(
    data,
    num_neighbors={key: [30] * 2 for key in data.edge_types},
    batch_size=1,
    input_nodes=('paper', None),
)

batch = next(iter(loader))
print(batch)

  paper={
    x=[79, 128],
    year=[79],
    y=[79],
    train_mask=[79],
    val_mask=[79],
    test_mask=[79],
    input_id=[1],
    batch_size=1
  },
  author={
    num_nodes=45,
    x=[45, 1134649]
  },
  field_of_study={
    num_nodes=0,
    x=[0, 59965]
  },
  (author, writes, paper)={ edge_index=[2, 51] },
  (paper, cites, paper)={ edge_index=[2, 89] },
  (paper, has_topic, field_of_study)={ edge_index=[2, 0] },
  (paper, rev_cites, paper)={ edge_index=[2, 89] },
  (paper, rev_writes, author)={ edge_index=[2, 51] },
  (field_of_study, rev_has_topic, paper)={ edge_index=[2, 0] }
)

I set batch to 1, it should only sample one paper node, why there are 79 papers.
The same goes for author, field_of_study is a neighbor but none of them are sampled.
Why is this?
My understanding of neighbor sampling should be similar to homogeneous graphs, where a point is the center of a circle and its neighbors are sampled outward.
Finally for num_neighbors={key: [30] * 2 for key in hetero_data.edge_types}, does [30]*2 mean that 30 edges of each type are sampled, repeated twice.
I have also used num_neighbors=[10] * 2, does it represent 10 nodes of each type sampled, repeated twice.
It also has the above phenomenon when batch=1.

[DuanhaoranCC]

Sorry, I've solved my problem. Thank you all!