MCPcopy
hub / github.com/dmlc/dgl / record_stream

Method record_stream

python/dgl/heterograph.py:5874–5897  ·  view source on GitHub ↗

Record the stream that is using this graph. This method only supports the PyTorch backend and requires graphs on the GPU. Parameters ---------- stream : torch.cuda.Stream The stream that is using this graph. Returns ------- DGLGra

(self, stream)

Source from the content-addressed store, hash-verified

5872 return self._graph.is_pinned()
5873
5874 def record_stream(self, stream):
5875 """Record the stream that is using this graph.
5876 This method only supports the PyTorch backend and requires graphs on the GPU.
5877
5878 Parameters
5879 ----------
5880 stream : torch.cuda.Stream
5881 The stream that is using this graph.
5882
5883 Returns
5884 -------
5885 DGLGraph
5886 self.
5887 """
5888 if F.get_preferred_backend() != "pytorch":
5889 raise DGLError("record_stream only support the PyTorch backend.")
5890 if F.device_type(self.device) != "cuda":
5891 raise DGLError("The graph must be on GPU to be recorded.")
5892 self._graph.record_stream(stream)
5893 for frame in itertools.chain(self._node_frames, self._edge_frames):
5894 for col in frame._columns.values():
5895 col.record_stream(stream)
5896
5897 return self
5898
5899 def clone(self):
5900 """Return a heterograph object that is a clone of current graph.

Callers

nothing calls this directly

Calls 2

DGLErrorClass · 0.85
valuesMethod · 0.45

Tested by

no test coverage detected