super(GCN, self).__init__() torch.manual_seed(12345) self.conv1 = GCNConv(dataset.num_features, hidden_channels) self.conv2 = GCNConv(hidden_channels, dataset.num_classes) def forward(self, x, edge_index): x = self.conv1(x, edge_index) x = x.relu() x = F.dropout(x, p=0....
(input, weight, bias) torch.manual_seed(124) batch_input = torch.randn(4, 2, 2, dtype=torch.double, requires_grad=True) batch_weight = torch.randn(4, 3, 2, dtype=torch.double, requires_grad=True) batch_bias = torch.randn(4, 3, dtype=torch.double, requires_grad=True) output = ...
As explained in this forum thread torch.normal only gives an error for negative sigma when there is no size argument given in. Code examples are below.To ReproduceSteps to reproduce the behavior:This code cause an error...torch.manual_seed(8) torch.normal(0, -1, (10, )) # output # ...
torch.compilereturns output with WRONG SHAPE aftercat_slice_cat importtorchtorch.manual_seed(420)width=16height=16channels=3batch_size=2x=torch.randn(batch_size,channels,height,width)classModel(torch.nn.Module):def__init__(self):super().__init__()defforward(self,x):cat_output=torch.cat([...
kshitij12345deleted thefix/multinomial/fast-pathbranchSeptember 10, 2020 07:54 mruberryadded theMergedlabelOct 28, 2020 Sign up for freeto join this conversation on GitHub. Already have an account?Sign in to comment Labels Mergedopen sourcetriagedThis issue has been looked at a team member, and...