RuntimeError: N == input.dim(0) assert faild. error BUT can‘t solve by existing solution
Christic9 opened this issue · comments
Hi, Thank you for the awesome work!
I have encoutered the same question as > #460 i don't use decoder blocks.
Here is my spconv class code:
class SparseConvNet(nn.Module):
def __init__(self,num_channels):
super(SparseConvNet, self).__init__()
self.conv0 = double_conv(num_channels, 16, 'subm0')
self.down0 = stride_conv(16, 32, 'down0')
self.conv1 = double_conv(32, 32, 'subm1')
self.down1 = stride_conv(32, 64, 'down1')
self.conv2 = triple_conv(64, 64, 'subm2')
self.down2 = stride_conv(64, 128, 'down2')
self.conv3 = triple_conv(128, 128, 'subm3')
self.down3 = stride_conv(128, 128, 'down3')
self.conv4 = triple_conv(128, 128, 'subm4')
def forward(self, x):
net = self.conv0(x)
net = self.down0(net)
net = self.conv1(net)
net1 = net.dense()
net = self.down1(net)
net = self.conv2(net)
net2 = net.dense()
net = self.down2(net)
net = self.conv3(net)
net3 = net.dense()
net = self.down3(net)
net = self.conv4(net)
net4 = net.dense()
volumes = [net1, net2, net3, net4]
return volumes
def single_conv(in_channels, out_channels, indice_key=None):
return spconv.SparseSequential(
spconv.SubMConv3d(in_channels,
out_channels,
1,
bias=False,
indice_key=indice_key),
nn.BatchNorm1d(out_channels, eps=1e-3, momentum=0.01),
nn.ReLU(),
)
def double_conv(in_channels, out_channels, indice_key=None):
return spconv.SparseSequential(
spconv.SubMConv3d(in_channels,
out_channels,
3,
bias=False,
indice_key=indice_key),
nn.BatchNorm1d(out_channels, eps=1e-3, momentum=0.01),
nn.ReLU(),
spconv.SubMConv3d(out_channels,
out_channels,
3,
bias=False,
indice_key=indice_key),
nn.BatchNorm1d(out_channels, eps=1e-3, momentum=0.01),
nn.ReLU(),
)
def triple_conv(in_channels, out_channels, indice_key=None):
return spconv.SparseSequential(
spconv.SubMConv3d(in_channels,
out_channels,
3,
bias=False,
indice_key=indice_key),
nn.BatchNorm1d(out_channels, eps=1e-3, momentum=0.01),
nn.ReLU(),
spconv.SubMConv3d(out_channels,
out_channels,
3,
bias=False,
indice_key=indice_key),
nn.BatchNorm1d(out_channels, eps=1e-3, momentum=0.01),
nn.ReLU(),
spconv.SubMConv3d(out_channels,
out_channels,
3,
bias=False,
indice_key=indice_key),
nn.BatchNorm1d(out_channels, eps=1e-3, momentum=0.01),
nn.ReLU(),
)
def stride_conv(in_channels, out_channels, indice_key=None):
return spconv.SparseSequential(
spconv.SparseConv3d(in_channels,
out_channels,
3,
2,
padding=1,
bias=False,
indice_key=indice_key),
nn.BatchNorm1d(out_channels, eps=1e-3, momentum=0.01), nn.ReLU())
Now I'm basically confirm the issue is caused by stride_conv block in loss backwards phase.
So how to solve this error?Appreciate your time!
Here is my error infor:
Traceback (most recent call last):
File "E:\anaconda\envs\neuralbody\lib\site-packages\torch\autograd\function.py", line 399, in wrapper
outputs = fn(ctx, *args)
File "E:\anaconda\envs\neuralbody\lib\site-packages\torch\cuda\amp\autocast_mode.py", line 135, in decorate_bwd
return bwd(*args, **kwargs)
File "E:\anaconda\envs\neuralbody\lib\site-packages\spconv\pytorch\functional.py", line 283, in backward
raise e
File "E:\anaconda\envs\neuralbody\lib\site-packages\spconv\pytorch\functional.py", line 273, in backward
fp32_accum=fp32_accum)
File "E:\anaconda\envs\neuralbody\lib\site-packages\spconv\pytorch\ops.py", line 1739, in implicit_gemm_backward
use_tf32=constants.SPCONV_ALLOW_TF32)
RuntimeError: d:\a\spconv\spconv\build\temp.win-amd64-cpython-37\release\spconv\build\core_cc\src\cumm\conv\main\convmainunittest\convmainunittest_matmul_split_simt_f32f32f32_0.cu(227)
N == input.dim(0) assert faild. error
Process finished with exit code 1
Hi, Thank you for the awesome work! I have encoutered the same question as > #460 i don't use decoder blocks. Here is my spconv class code:
class SparseConvNet(nn.Module): def __init__(self,num_channels): super(SparseConvNet, self).__init__() self.conv0 = double_conv(num_channels, 16, 'subm0') self.down0 = stride_conv(16, 32, 'down0') self.conv1 = double_conv(32, 32, 'subm1') self.down1 = stride_conv(32, 64, 'down1') self.conv2 = triple_conv(64, 64, 'subm2') self.down2 = stride_conv(64, 128, 'down2') self.conv3 = triple_conv(128, 128, 'subm3') self.down3 = stride_conv(128, 128, 'down3') self.conv4 = triple_conv(128, 128, 'subm4') def forward(self, x): net = self.conv0(x) net = self.down0(net) net = self.conv1(net) net1 = net.dense() net = self.down1(net) net = self.conv2(net) net2 = net.dense() net = self.down2(net) net = self.conv3(net) net3 = net.dense() net = self.down3(net) net = self.conv4(net) net4 = net.dense() volumes = [net1, net2, net3, net4] return volumes def single_conv(in_channels, out_channels, indice_key=None): return spconv.SparseSequential( spconv.SubMConv3d(in_channels, out_channels, 1, bias=False, indice_key=indice_key), nn.BatchNorm1d(out_channels, eps=1e-3, momentum=0.01), nn.ReLU(), ) def double_conv(in_channels, out_channels, indice_key=None): return spconv.SparseSequential( spconv.SubMConv3d(in_channels, out_channels, 3, bias=False, indice_key=indice_key), nn.BatchNorm1d(out_channels, eps=1e-3, momentum=0.01), nn.ReLU(), spconv.SubMConv3d(out_channels, out_channels, 3, bias=False, indice_key=indice_key), nn.BatchNorm1d(out_channels, eps=1e-3, momentum=0.01), nn.ReLU(), ) def triple_conv(in_channels, out_channels, indice_key=None): return spconv.SparseSequential( spconv.SubMConv3d(in_channels, out_channels, 3, bias=False, indice_key=indice_key), nn.BatchNorm1d(out_channels, eps=1e-3, momentum=0.01), nn.ReLU(), spconv.SubMConv3d(out_channels, out_channels, 3, bias=False, indice_key=indice_key), nn.BatchNorm1d(out_channels, eps=1e-3, momentum=0.01), nn.ReLU(), spconv.SubMConv3d(out_channels, out_channels, 3, bias=False, indice_key=indice_key), nn.BatchNorm1d(out_channels, eps=1e-3, momentum=0.01), nn.ReLU(), ) def stride_conv(in_channels, out_channels, indice_key=None): return spconv.SparseSequential( spconv.SparseConv3d(in_channels, out_channels, 3, 2, padding=1, bias=False, indice_key=indice_key), nn.BatchNorm1d(out_channels, eps=1e-3, momentum=0.01), nn.ReLU())Now I'm basically confirm the issue is caused by stride_conv block in loss backwards phase. So how to solve this error?Appreciate your time!
Here is my error infor:
Traceback (most recent call last): File "E:\anaconda\envs\neuralbody\lib\site-packages\torch\autograd\function.py", line 399, in wrapper outputs = fn(ctx, *args) File "E:\anaconda\envs\neuralbody\lib\site-packages\torch\cuda\amp\autocast_mode.py", line 135, in decorate_bwd return bwd(*args, **kwargs) File "E:\anaconda\envs\neuralbody\lib\site-packages\spconv\pytorch\functional.py", line 283, in backward raise e File "E:\anaconda\envs\neuralbody\lib\site-packages\spconv\pytorch\functional.py", line 273, in backward fp32_accum=fp32_accum) File "E:\anaconda\envs\neuralbody\lib\site-packages\spconv\pytorch\ops.py", line 1739, in implicit_gemm_backward use_tf32=constants.SPCONV_ALLOW_TF32) RuntimeError: d:\a\spconv\spconv\build\temp.win-amd64-cpython-37\release\spconv\build\core_cc\src\cumm\conv\main\convmainunittest\convmainunittest_matmul_split_simt_f32f32f32_0.cu(227) N == input.dim(0) assert faild. error Process finished with exit code 1
Hi，did you have solved this question? I have met the same error as yours!
Hi，did you have solved this question? I have met the same error as yours!