Hi,
I am now trying to implement the GraphSAGE in a giant graph.
At first, my plan is trying to use NodeFlow and SAGEconv, but due to SAGEConv only accept DGLGraph and NodeFlow could not be transferred into DGLGraph for now, then I give up this idea.
Then I tried original DGLGraph + SAGEConv ( I know this could be very low efficient), but got this complicated error information:
---------------------------------------------------------------------------
DGLError Traceback (most recent call last)
<ipython-input-7-7b1b03d1411a> in <module>
35 for epoch in range(num_epochs):
36 epoch_loss = 0
---> 37 prediction = model(g, cora_nodes_features[:34])
38 print(prediction.shape)
39 # loss = loss_func(prediction, ground_truth)
/opt/miniconda3/envs/PyTorch/lib/python3.7/site-packages/torch/nn/modules/module.py in __call__(self, *input, **kwargs)
539 result = self._slow_forward(*input, **kwargs)
540 else:
--> 541 result = self.forward(*input, **kwargs)
542 for hook in self._forward_hooks.values():
543 hook_result = hook(self, input, result)
<ipython-input-6-6679e06a574f> in forward(self, graph, features)
15 h = features
16 for layer in self.layers:
---> 17 h = layer(g, h)
18 return h
/opt/miniconda3/envs/PyTorch/lib/python3.7/site-packages/torch/nn/modules/module.py in __call__(self, *input, **kwargs)
539 result = self._slow_forward(*input, **kwargs)
540 else:
--> 541 result = self.forward(*input, **kwargs)
542 for hook in self._forward_hooks.values():
543 hook_result = hook(self, input, result)
/opt/miniconda3/envs/PyTorch/lib/python3.7/site-packages/dgl/nn/pytorch/conv/sageconv.py in forward(self, graph, feat)
108 if self._aggre_type == 'mean':
109 graph.ndata['h'] = feat
--> 110 graph.update_all(fn.copy_src('h', 'm'), fn.mean('m', 'neigh'))
111 h_neigh = graph.ndata['neigh']
112 elif self._aggre_type == 'gcn':
/opt/miniconda3/envs/PyTorch/lib/python3.7/site-packages/dgl/graph.py in update_all(self, message_func, reduce_func, apply_node_func)
2745 reduce_func=reduce_func,
2746 apply_func=apply_node_func)
-> 2747 Runtime.run(prog)
2748
2749 def prop_nodes(self,
/opt/miniconda3/envs/PyTorch/lib/python3.7/site-packages/dgl/runtime/runtime.py in run(prog)
9 for exe in prog.execs:
10 # prog.pprint_exe(exe)
---> 11 exe.run()
/opt/miniconda3/envs/PyTorch/lib/python3.7/site-packages/dgl/runtime/ir/executor.py in run(self)
1198 self.ret.data = F.copy_reduce(
1199 self.reducer, graph, self.target, in_data, self.out_size, in_map,
-> 1200 out_map)
1201
1202
/opt/miniconda3/envs/PyTorch/lib/python3.7/site-packages/dgl/backend/pytorch/tensor.py in forward(ctx, reducer, graph, target, in_data, out_size, in_map, out_map)
371 K.copy_reduce(
372 reducer if reducer != 'mean' else 'sum',
--> 373 graph, target, in_data_nd, out_data_nd, in_map[0], out_map[0])
374 # normalize if mean reducer
375 # NOTE(zihao): this is a temporary hack and we should have better solution in the future.
/opt/miniconda3/envs/PyTorch/lib/python3.7/site-packages/dgl/kernel.py in copy_reduce(reducer, G, target, X, out, X_rows, out_rows)
370 _CAPI_DGLKernelCopyReduce(
371 reducer, G, int(target),
--> 372 X, out, X_rows, out_rows)
373
374 # pylint: disable=invalid-name
dgl/_ffi/_cython/./function.pxi in dgl._ffi._cy3.core.FunctionBase.__call__()
dgl/_ffi/_cython/./function.pxi in dgl._ffi._cy3.core.FuncCall()
dgl/_ffi/_cython/./base.pxi in dgl._ffi._cy3.core.CALL()
DGLError: [14:24:40] /Users/xiangsx/work/dgl/dgl/src/kernel/cpu/../binary_reduce_impl.h:112: Unsupported dtype: _@
Stack trace:
[bt] (0) 1 libdgl.dylib 0x0000000120bb1309 dmlc::LogMessageFatal::~LogMessageFatal() + 57
[bt] (1) 2 libdgl.dylib 0x000000012113dc9d void dgl::kernel::BinaryReduceImpl<1>(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, dgl::kernel::CSRWrapper const&, dgl::kernel::binary_op::Target, dgl::kernel::binary_op::Target, dgl::runtime::NDArray, dgl::runtime::NDArray, dgl::runtime::NDArray, dgl::runtime::NDArray, dgl::runtime::NDArray, dgl::runtime::NDArray) + 1037
[bt] (2) 3 libdgl.dylib 0x0000000120bd4ce7 dgl::kernel::CopyReduce(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, dgl::kernel::CSRWrapper const&, dgl::kernel::binary_op::Target, dgl::runtime::NDArray, dgl::runtime::NDArray, dgl::runtime::NDArray, dgl::runtime::NDArray) + 2151
[bt] (3) 4 libdgl.dylib 0x0000000120bde991 std::__1::__function::__func<dgl::kernel::$_4, std::__1::allocator<dgl::kernel::$_4>, void (dgl::runtime::DGLArgs, dgl::runtime::DGLRetValue*)>::operator()(dgl::runtime::DGLArgs&&, dgl::runtime::DGLRetValue*&&) + 1473
[bt] (4) 5 libdgl.dylib 0x0000000121368de6 DGLFuncCall + 70
[bt] (5) 6 core.cpython-37m-darwin.so 0x000000012197f86c __pyx_f_3dgl_4_ffi_4_cy3_4core_FuncCall(void*, _object*, DGLValue*, int*) + 924
[bt] (6) 7 core.cpython-37m-darwin.so 0x0000000121983c27 __pyx_pw_3dgl_4_ffi_4_cy3_4core_12FunctionBase_5__call__(_object*, _object*, _object*) + 55
[bt] (7) 8 python 0x0000000103a82e03 _PyObject_FastCallKeywords + 179
[bt] (8) 9 python 0x0000000103bbfd75 call_function + 453
It seems caused by SAGEConv…Any help?
Thank you~
. Have you found solution?