[Pickle] Fix HeteroGraphConv pickle problem (dmlc#2761)

* fix pickle problem

* lint

* add pickle tests

* fix

* fix

* fix

* fix

* fix for windows

python/dgl/nn/pytorch/hetero.py

@@ -1,6 +1,8 @@
 """Heterograph NN modules"""
+from functools import partial
 import torch as th
 import torch.nn as nn
+from ...base import DGLError
 
 __all__ = ['HeteroGraphConv']
 
@@ -196,6 +198,29 @@ def forward(self, g, inputs, mod_args=None, mod_kwargs=None):
                 rsts[nty] = self.agg_fn(alist, nty)
         return rsts
 
+def _max_reduce_func(inputs, dim):
+    return th.max(inputs, dim=dim)[0]
+
+def _min_reduce_func(inputs, dim):
+    return th.min(inputs, dim=dim)[0]
+
+def _sum_reduce_func(inputs, dim):
+    return th.sum(inputs, dim=dim)
+
+def _mean_reduce_func(inputs, dim):
+    return th.mean(inputs, dim=dim)
+
+def _stack_agg_func(inputs, dsttype): # pylint: disable=unused-argument
+    if len(inputs) == 0:
+        return None
+    return th.stack(inputs, dim=1)
+
+def _agg_func(inputs, dsttype, fn): # pylint: disable=unused-argument
+    if len(inputs) == 0:
+        return None
+    stacked = th.stack(inputs, dim=0)
+    return fn(stacked, dim=0)
+
 def get_aggregate_fn(agg):
     """Internal function to get the aggregation function for node data
     generated from different relations.
@@ -213,28 +238,19 @@ def get_aggregate_fn(agg):
         and returns one aggregated tensor.
     """
     if agg == 'sum':
-        fn = th.sum
+        fn = _sum_reduce_func
     elif agg == 'max':
-        fn = lambda inputs, dim: th.max(inputs, dim=dim)[0]
+        fn = _max_reduce_func
     elif agg == 'min':
-        fn = lambda inputs, dim: th.min(inputs, dim=dim)[0]
+        fn = _min_reduce_func
     elif agg == 'mean':
-        fn = th.mean
+        fn = _mean_reduce_func
     elif agg == 'stack':
         fn = None  # will not be called
     else:
         raise DGLError('Invalid cross type aggregator. Must be one of '
                        '"sum", "max", "min", "mean" or "stack". But got "%s"' % agg)
     if agg == 'stack':
-        def stack_agg(inputs, dsttype):  # pylint: disable=unused-argument
-            if len(inputs) == 0:
-                return None
-            return th.stack(inputs, dim=1)
-        return stack_agg
+        return _stack_agg_func
     else:
-        def aggfn(inputs, dsttype):  # pylint: disable=unused-argument
-            if len(inputs) == 0:
-                return None
-            stacked = th.stack(inputs, dim=0)
-            return fn(stacked, dim=0)
-        return aggfn
+        return partial(_agg_func, fn=fn)

tests/pytorch/test_nn.py

@@ -1,3 +1,4 @@
+import io
 import torch as th
 import networkx as nx
 import dgl
@@ -8,9 +9,12 @@
 from test_utils.graph_cases import get_cases, random_graph, random_bipartite, random_dglgraph
 from test_utils import parametrize_dtype
 from copy import deepcopy
+import pickle
 
 import scipy as sp
 
+tmp_buffer = io.BytesIO()
+
 def _AXWb(A, X, W, b):
     X = th.matmul(X, W)
     Y = th.matmul(A, X.view(X.shape[0], -1)).view_as(X)
@@ -25,6 +29,11 @@ def test_graph_conv0(out_dim):
     conv = nn.GraphConv(5, out_dim, norm='none', bias=True)
     conv = conv.to(ctx)
     print(conv)
+
+    # test pickle
+    th.save(conv, tmp_buffer)
+
+
     # test#1: basic
     h0 = F.ones((3, 5))
     h1 = conv(g, h0)
@@ -119,6 +128,10 @@ def test_graph_conv_e_weight(idtype, g, norm, weight, bias, out_dim):
 def test_graph_conv_e_weight_norm(idtype, g, norm, weight, bias, out_dim):
     g = g.astype(idtype).to(F.ctx())
     conv = nn.GraphConv(5, out_dim, norm=norm, weight=weight, bias=bias).to(F.ctx())
+
+    # test pickle
+    th.save(conv, tmp_buffer)
+
     ext_w = F.randn((5, out_dim)).to(F.ctx())
     nsrc = g.number_of_src_nodes()
     ndst = g.number_of_dst_nodes()
@@ -141,6 +154,10 @@ def test_graph_conv_bi(idtype, g, norm, weight, bias, out_dim):
     # Test a pair of tensor inputs
     g = g.astype(idtype).to(F.ctx())
     conv = nn.GraphConv(5, out_dim, norm=norm, weight=weight, bias=bias).to(F.ctx())
+
+    # test pickle
+    th.save(conv, tmp_buffer)
+
     ext_w = F.randn((5, out_dim)).to(F.ctx())
     nsrc = g.number_of_src_nodes()
     ndst = g.number_of_dst_nodes()
@@ -175,6 +192,9 @@ def test_tagconv(out_dim):
     conv = nn.TAGConv(5, out_dim, bias=True)
     conv = conv.to(ctx)
     print(conv)
+
+    # test pickle
+    th.save(conv, tmp_buffer)
 
     # test#1: basic
     h0 = F.ones((3, 5))
@@ -231,6 +251,9 @@ def test_glob_att_pool():
     gap = gap.to(ctx)
     print(gap)
 
+    # test pickle
+    th.save(gap, tmp_buffer)
+
     # test#1: basic
     h0 = F.randn((g.number_of_nodes(), 5))
     h1 = gap(g, h0)
@@ -347,6 +370,10 @@ def test_rgcn(O):
     I = 10
 
     rgc_basis = nn.RelGraphConv(I, O, R, "basis", B).to(ctx)
+
+    # test pickle
+    th.save(rgc_basis, tmp_buffer)
+
     rgc_basis_low = nn.RelGraphConv(I, O, R, "basis", B, low_mem=True).to(ctx)
     rgc_basis_low.weight = rgc_basis.weight
     rgc_basis_low.w_comp = rgc_basis.w_comp
@@ -509,6 +536,10 @@ def test_gat_conv(g, idtype, out_dim, num_heads):
     feat = F.randn((g.number_of_nodes(), 5))
     gat = gat.to(ctx)
     h = gat(g, feat)
+
+    # test pickle
+    th.save(gat, tmp_buffer)
+
     assert h.shape == (g.number_of_nodes(), num_heads, out_dim)
     _, a = gat(g, feat, get_attention=True)
     assert a.shape == (g.number_of_edges(), num_heads, 1)
@@ -536,6 +567,8 @@ def test_sage_conv(idtype, g, aggre_type):
     sage = nn.SAGEConv(5, 10, aggre_type)
     feat = F.randn((g.number_of_nodes(), 5))
     sage = sage.to(F.ctx())
+    # test pickle
+    th.save(sage, tmp_buffer)
     h = sage(g, feat)
     assert h.shape[-1] == 10
 
@@ -583,6 +616,10 @@ def test_sgc_conv(g, idtype, out_dim):
     g = g.astype(idtype).to(ctx)
     # not cached
     sgc = nn.SGConv(5, out_dim, 3)
+
+    # test pickle
+    th.save(sgc, tmp_buffer)
+
     feat = F.randn((g.number_of_nodes(), 5))
     sgc = sgc.to(ctx)
 
@@ -605,6 +642,9 @@ def test_appnp_conv(g, idtype):
     appnp = nn.APPNPConv(10, 0.1)
     feat = F.randn((g.number_of_nodes(), 5))
     appnp = appnp.to(ctx)
+
+    # test pickle
+    th.save(appnp, tmp_buffer)
 
     h = appnp(g, feat)
     assert h.shape[-1] == 5
@@ -622,6 +662,10 @@ def test_gin_conv(g, idtype, aggregator_type):
     feat = F.randn((g.number_of_nodes(), 5))
     gin = gin.to(ctx)
     h = gin(g, feat)
+
+    # test pickle
+    th.save(h, tmp_buffer)
+
     assert h.shape == (g.number_of_nodes(), 12)
 
 @parametrize_dtype
@@ -784,6 +828,10 @@ def test_edge_conv(g, idtype, out_dim):
     ctx = F.ctx()
     edge_conv = nn.EdgeConv(5, out_dim).to(ctx)
     print(edge_conv)
+
+    # test pickle
+    th.save(edge_conv, tmp_buffer)
+
     h0 = F.randn((g.number_of_nodes(), 5))
     h1 = edge_conv(g, h0)
     assert h1.shape == (g.number_of_nodes(), out_dim)
@@ -811,6 +859,10 @@ def test_dotgat_conv(g, idtype, out_dim, num_heads):
     dotgat = nn.DotGatConv(5, out_dim, num_heads)
     feat = F.randn((g.number_of_nodes(), 5))
     dotgat = dotgat.to(ctx)
+
+    # test pickle
+    th.save(dotgat, tmp_buffer)
+
     h = dotgat(g, feat)
     assert h.shape == (g.number_of_nodes(), num_heads, out_dim)
     _, a = dotgat(g, feat, get_attention=True)
@@ -919,6 +971,7 @@ def test_atomic_conv(g, idtype):
     dist = F.randn((g.number_of_edges(), 1))
 
     h = aconv(g, feat, dist)
+
     # current we only do shape check
     assert h.shape[-1] == 4
 
@@ -968,6 +1021,10 @@ def test_hetero_conv(agg, idtype):
         'sells': nn.GraphConv(3, 4, allow_zero_in_degree=True)},
         agg)
     conv = conv.to(F.ctx())
+
+    # test pickle
+    th.save(conv, tmp_buffer)
+
     uf = F.randn((4, 2))
     gf = F.randn((4, 4))
     sf = F.randn((2, 3))