graphlearn_torch/python/data/vineyard_utils.py (1 line):
	- line 84: # TODO: optimise this query process if too slow


graphlearn_torch/python/distributed/dist_graph.py (1 line):
	- line 31: TODO: support graph operations.


graphlearn_torch/python/sampler/neighbor_sampler.py (1 line):
	- line 362: # TODO: make triplet negative sampling strict.


graphlearn_torch/python/distributed/dist_subgraph_loader.py (1 line):
	- line 85: # TODO: currently only support out-sample


graphlearn_torch/python/distributed/dist_feature.py (1 line):
	- line 156: # TODO: check performance with `return feat[ids].cpu()`