Issues with the creation of dataloader using NeighborLoader/ HGTLoader with Heterogenous Graph

Question

I am trying to create a dataloader based on a heterogenous graph built with PyG to be used in a model implemented in Pytorch Lightning. I tried to generate a dataloader with both HGTLoader and NeighborLoader but in both cases the output is an empty object.

To check if the dataloader creation process is working properly I created a toy graph as follows:

data= HeteroData(
  user={x=[5, 1],},
  poi={x=[10, 1] },
  (user, visits, poi)={
    edge_index=[2, 40],
    edge_attr=[40, 2],
  },
  (poi, connects, poi)={
    edge_index=[2, 100],
    edge_attr=[100, 2],
  }
)

with

poi_ids = (0,1,2,3,4,5,6,7,8,9)
user_ids = (10,11,12,13,14)

I have checked and all the edge_index elements are correctly indentified. I tried to generate a dataloader with both HGTLoader and NeighborLoader but in both cases the output is an empty object. Consider that I want the user nodes as input nodes for the sampling.
Below the code of the two samplers.

data_loader = HGTLoader(
    data, 
    batch_size=2, 
    input_nodes=('user', torch.tensor([10, 11, 12, 13,14], dtype=torch.long)),
    num_samples={key: [2]  for key in data.node_types}, 
    shuffle=True, 
)

data_loader= NeighborLoader(
          data, 
          batch_size=2, 
          input_nodes=('user', torch.tensor([10, 11, 12, 13,14], dtype=torch.long)), 
          num_neighbors={key: [2]  for key in data.edge_types}, 
          shuffle=True)

Can anybody help me identify the problem? Many thanks!