# Dynamic number of Branches

How to generate dynamic number of branches?
My code use a list to store the branches, but got

RuntimeError: Caught RuntimeError in replica 1 on device 1.

RuntimeError: Expected tensor for argument #1 ‘input’ to have the same device as tensor for argument #2 ‘weight’; but device 1 does not equal 0 (while checking arguments for cudnn_convolution)

I think it is because the weights for branches are not copied to other GPUs.

``````class ResNetONE(nn.Module):
def __init__(self, depth, num_classes=1000, num_branches=3, block_name='BasicBlock'):
super(ResNetONE, self).__init__()
# Model type specifies number of layers for CIFAR-10 model
if block_name.lower() == 'basicblock':
assert(depth - 2) % 6 == 0, 'When use basicblock, depth should be 6n+2, e.g. 20, 32, 44, 56, 110, 1202'
n = (depth - 2) // 6
block = BasicBlock
elif block_name.lower() == 'bottleneck':
assert (depth == 2) % 9 == 0, 'When use bottleneck, depth should be 9n + 2, e.g. 20, 29, 47, 56, 110, 1199'
n = (depth - 2) // 9
block = Bottleneck
else:
raise ValueError('block_name shinterval_sumould be Basicblock or Bottleneck')

self.inplanes = 16
self.num_branches = num_branches
self.num_classes = num_classes

self.conv1 = nn.Conv2d(3, 16, kernel_size=3, padding=1, bias=False)
self.bn1 = nn.BatchNorm2d(16)
self.relu = nn.ReLU(inplace=True)
self.layer1 = self._make_layer(block, 16, n)
self.layer2 = self._make_layer(block, 32, n, stride=2)
self.layer3 = self._make_layer(block, 64, n, stride=2)
self.avgpool = nn.AvgPool2d(8)
self.fc = nn.Linear(64 * block.expansion, self.num_classes)
self.branches = self._make_branches(self.layer3, self.avgpool)
self.gate = self._make_gate()

for m in self.modules():
if isinstance(m, nn.Conv2d):
n = m.kernel_size * m.kernel_size * m.out_channels
m.weight.data.normal_(0, math.sqrt(2./n))
# "normal_" is defined by torch._C._TensorBase
# https://pytorch.org/docs/stable/tensors.html#torch.Tensor.normal_
# https://zhuanlan.zhihu.com/p/100937718
elif isinstance(m, nn.BatchNorm2d):
m.weight.data.fill_(1)
m.bias.data.zero_()

def _make_layer(self, block, planes, blocks, stride=1):
downsample = None
if stride != 1 or self.inplanes != planes * block.expansion:
downsample = nn.Sequential(
nn.Conv2d(self.inplanes, planes * block.expansion,
kernel_size=1, stride=stride, bias=False),
nn.BatchNorm2d(planes * block.expansion)
)
layers = []
layers.append(block(self.inplanes, planes, stride, downsample))
self.inplanes = planes * block.expansion
for i in range(1, blocks):
layers.append(block(self.inplanes, planes))
return nn.Sequential(*layers)

def _make_branches(self, *layers):
branches = []
for i in range(self.num_branches):
branch = nn.Sequential(*layers)
branches.append(branch)
return branches
``````

Hey @Erica_Zheng, are you using `DataParallel`, `DistributedDataParallel` or `torch.distributed`? And can you include a min repro? Looks like the above example does not include the `forward()` function or how the model forward pass was launched?

@mrshenli Thank you, Shen!
No Runtime Error now by using
`branches = nn.ModuleList(branches)`