TypeError: can't serialize tensor(62, device='cuda:0')

[ltc576935585]

I do not know why it happend.

/home/lts/.conda/envs/PCL/bin/python /home/lts/PycharmProject/mean-teacher/pytorch/main.py INFO:main:=> creating model 'cifar_shakeshake26' INFO:main:=> creating EMA model 'cifar_shakeshake26' INFO:main: List of model parameters:

module.conv1.weight 16 3 3 3 = 432 module.layer1.0.conv_a1.weight 96 16 3 3 = 13,824 module.layer1.0.bn_a1.weight 96 = 96 module.layer1.0.bn_a1.bias 96 = 96 module.layer1.0.conv_a2.weight 96 96 3 3 = 82,944 module.layer1.0.bn_a2.weight 96 = 96 module.layer1.0.bn_a2.bias 96 = 96 module.layer1.0.conv_b1.weight 96 16 3 3 = 13,824 module.layer1.0.bn_b1.weight 96 = 96 module.layer1.0.bn_b1.bias 96 = 96 module.layer1.0.conv_b2.weight 96 96 3 3 = 82,944 module.layer1.0.bn_b2.weight 96 = 96 module.layer1.0.bn_b2.bias 96 = 96 module.layer1.0.downsample.0.weight 96 16 1 1 = 1,536 module.layer1.0.downsample.1.weight 96 = 96 module.layer1.0.downsample.1.bias 96 = 96 module.layer1.1.conv_a1.weight 96 96 3 3 = 82,944 module.layer1.1.bn_a1.weight 96 = 96 module.layer1.1.bn_a1.bias 96 = 96 module.layer1.1.conv_a2.weight 96 96 3 3 = 82,944 module.layer1.1.bn_a2.weight 96 = 96 module.layer1.1.bn_a2.bias 96 = 96 module.layer1.1.conv_b1.weight 96 96 3 3 = 82,944 module.layer1.1.bn_b1.weight 96 = 96 module.layer1.1.bn_b1.bias 96 = 96 module.layer1.1.conv_b2.weight 96 96 3 3 = 82,944 module.layer1.1.bn_b2.weight 96 = 96 module.layer1.1.bn_b2.bias 96 = 96 module.layer1.2.conv_a1.weight 96 96 3 3 = 82,944 module.layer1.2.bn_a1.weight 96 = 96 module.layer1.2.bn_a1.bias 96 = 96 module.layer1.2.conv_a2.weight 96 96 3 3 = 82,944 module.layer1.2.bn_a2.weight 96 = 96 module.layer1.2.bn_a2.bias 96 = 96 module.layer1.2.conv_b1.weight 96 96 3 3 = 82,944 module.layer1.2.bn_b1.weight 96 = 96 module.layer1.2.bn_b1.bias 96 = 96 module.layer1.2.conv_b2.weight 96 96 3 3 = 82,944 module.layer1.2.bn_b2.weight 96 = 96 module.layer1.2.bn_b2.bias 96 = 96 module.layer1.3.conv_a1.weight 96 96 3 3 = 82,944 module.layer1.3.bn_a1.weight 96 = 96 module.layer1.3.bn_a1.bias 96 = 96 module.layer1.3.conv_a2.weight 96 96 3 3 = 82,944 module.layer1.3.bn_a2.weight 96 = 96 module.layer1.3.bn_a2.bias 96 = 96 module.layer1.3.conv_b1.weight 96 96 3 3 = 82,944 module.layer1.3.bn_b1.weight 96 = 96 module.layer1.3.bn_b1.bias 96 = 96 module.layer1.3.conv_b2.weight 96 96 3 3 = 82,944 module.layer1.3.bn_b2.weight 96 = 96 module.layer1.3.bn_b2.bias 96 = 96 module.layer2.0.conv_a1.weight 192 96 3 3 = 165,888 module.layer2.0.bn_a1.weight 192 = 192 module.layer2.0.bn_a1.bias 192 = 192 module.layer2.0.conv_a2.weight 192 192 3 3 = 331,776 module.layer2.0.bn_a2.weight 192 = 192 module.layer2.0.bn_a2.bias 192 = 192 module.layer2.0.conv_b1.weight 192 96 3 3 = 165,888 module.layer2.0.bn_b1.weight 192 = 192 module.layer2.0.bn_b1.bias 192 = 192 module.layer2.0.conv_b2.weight 192 192 3 3 = 331,776 module.layer2.0.bn_b2.weight 192 = 192 module.layer2.0.bn_b2.bias 192 = 192 module.layer2.0.downsample.conv.weight 192 96 1 1 = 18,432 module.layer2.0.downsample.conv.bias 192 = 192 module.layer2.0.downsample.bn.weight 192 = 192 module.layer2.0.downsample.bn.bias 192 = 192 module.layer2.1.conv_a1.weight 192 192 3 3 = 331,776 module.layer2.1.bn_a1.weight 192 = 192 module.layer2.1.bn_a1.bias 192 = 192 module.layer2.1.conv_a2.weight 192 192 3 3 = 331,776 module.layer2.1.bn_a2.weight 192 = 192 module.layer2.1.bn_a2.bias 192 = 192 module.layer2.1.conv_b1.weight 192 192 3 3 = 331,776 module.layer2.1.bn_b1.weight 192 = 192 module.layer2.1.bn_b1.bias 192 = 192 module.layer2.1.conv_b2.weight 192 192 3 3 = 331,776 module.layer2.1.bn_b2.weight 192 = 192 module.layer2.1.bn_b2.bias 192 = 192 module.layer2.2.conv_a1.weight 192 192 3 3 = 331,776 module.layer2.2.bn_a1.weight 192 = 192 module.layer2.2.bn_a1.bias 192 = 192 module.layer2.2.conv_a2.weight 192 192 3 3 = 331,776 module.layer2.2.bn_a2.weight 192 = 192 module.layer2.2.bn_a2.bias 192 = 192 module.layer2.2.conv_b1.weight 192 192 3 3 = 331,776 module.layer2.2.bn_b1.weight 192 = 192 module.layer2.2.bn_b1.bias 192 = 192 module.layer2.2.conv_b2.weight 192 192 3 3 = 331,776 module.layer2.2.bn_b2.weight 192 = 192 module.layer2.2.bn_b2.bias 192 = 192 module.layer2.3.conv_a1.weight 192 192 3 3 = 331,776 module.layer2.3.bn_a1.weight 192 = 192 module.layer2.3.bn_a1.bias 192 = 192 module.layer2.3.conv_a2.weight 192 192 3 3 = 331,776 module.layer2.3.bn_a2.weight 192 = 192 module.layer2.3.bn_a2.bias 192 = 192 module.layer2.3.conv_b1.weight 192 192 3 3 = 331,776 module.layer2.3.bn_b1.weight 192 = 192 module.layer2.3.bn_b1.bias 192 = 192 module.layer2.3.conv_b2.weight 192 192 3 3 = 331,776 module.layer2.3.bn_b2.weight 192 = 192 module.layer2.3.bn_b2.bias 192 = 192 module.layer3.0.conv_a1.weight 384 192 3 3 = 663,552 module.layer3.0.bn_a1.weight 384 = 384 module.layer3.0.bn_a1.bias 384 = 384 module.layer3.0.conv_a2.weight 384 384 3 3 = 1,327,104 module.layer3.0.bn_a2.weight 384 = 384 module.layer3.0.bn_a2.bias 384 = 384 module.layer3.0.conv_b1.weight 384 192 3 3 = 663,552 module.layer3.0.bn_b1.weight 384 = 384 module.layer3.0.bn_b1.bias 384 = 384 module.layer3.0.conv_b2.weight 384 384 3 3 = 1,327,104 module.layer3.0.bn_b2.weight 384 = 384 module.layer3.0.bn_b2.bias 384 = 384 module.layer3.0.downsample.conv.weight 384 192 1 1 = 73,728 module.layer3.0.downsample.conv.bias 384 = 384 module.layer3.0.downsample.bn.weight 384 = 384 module.layer3.0.downsample.bn.bias 384 = 384 module.layer3.1.conv_a1.weight 384 384 3 3 = 1,327,104 module.layer3.1.bn_a1.weight 384 = 384 module.layer3.1.bn_a1.bias 384 = 384 module.layer3.1.conv_a2.weight 384 384 3 3 = 1,327,104 module.layer3.1.bn_a2.weight 384 = 384 module.layer3.1.bn_a2.bias 384 = 384 module.layer3.1.conv_b1.weight 384 384 3 3 = 1,327,104 module.layer3.1.bn_b1.weight 384 = 384 module.layer3.1.bn_b1.bias 384 = 384 module.layer3.1.conv_b2.weight 384 384 3 3 = 1,327,104 module.layer3.1.bn_b2.weight 384 = 384 module.layer3.1.bn_b2.bias 384 = 384 module.layer3.2.conv_a1.weight 384 384 3 3 = 1,327,104 module.layer3.2.bn_a1.weight 384 = 384 module.layer3.2.bn_a1.bias 384 = 384 module.layer3.2.conv_a2.weight 384 384 3 3 = 1,327,104 module.layer3.2.bn_a2.weight 384 = 384 module.layer3.2.bn_a2.bias 384 = 384 module.layer3.2.conv_b1.weight 384 384 3 3 = 1,327,104 module.layer3.2.bn_b1.weight 384 = 384 module.layer3.2.bn_b1.bias 384 = 384 module.layer3.2.conv_b2.weight 384 384 3 3 = 1,327,104 module.layer3.2.bn_b2.weight 384 = 384 module.layer3.2.bn_b2.bias 384 = 384 module.layer3.3.conv_a1.weight 384 384 3 3 = 1,327,104 module.layer3.3.bn_a1.weight 384 = 384 module.layer3.3.bn_a1.bias 384 = 384 module.layer3.3.conv_a2.weight 384 384 3 3 = 1,327,104 module.layer3.3.bn_a2.weight 384 = 384 module.layer3.3.bn_a2.bias 384 = 384 module.layer3.3.conv_b1.weight 384 384 3 3 = 1,327,104 module.layer3.3.bn_b1.weight 384 = 384 module.layer3.3.bn_b1.bias 384 = 384 module.layer3.3.conv_b2.weight 384 384 3 3 = 1,327,104 module.layer3.3.bn_b2.weight 384 = 384 module.layer3.3.bn_b2.bias 384 = 384 module.fc1.weight 10 384 = 3,840 module.fc1.bias 10 = 10 module.fc2.weight 10 384 = 3,840 module.fc2.bias 10 = 10

all parameters sum of above = 26,197,316

/home/lts/.conda/envs/PCL/lib/python3.6/site-packages/torch/nn/_reduction.py:49: UserWarning: size_average and reduce args will be deprecated, please use reduction='sum' instead. warnings.warn(warning.format(ret)) /home/lts/PycharmProject/mean-teacher/pytorch/main.py:224: UserWarning: volatile was removed and now has no effect. Use with torch.no_grad(): instead. ema_input_var = torch.autograd.Variable(ema_input, volatile=True) INFO:main:Epoch: [0][0/22000] Time 10.247 (10.247) Data 3.054 (3.054) Class 2.2228 (2.2228) Cons 0.0015 (0.0015) Prec@1 3.000 (3.000) Prec@5 51.000 (51.000) Traceback (most recent call last): File "/home/lts/PycharmProject/mean-teacher/pytorch/main.py", line 426, in main(RunContext(file, 0)) File "/home/lts/PycharmProject/mean-teacher/pytorch/main.py", line 105, in main train(train_loader, model, ema_model, optimizer, epoch, training_log) File "/home/lts/PycharmProject/mean-teacher/pytorch/main.py", line 311, in train meters.sums() File "/home/lts/PycharmProject/mean-teacher/pytorch/mean_teacher/run_context.py", line 34, in record self._record(step, col_val_dict) File "/home/lts/PycharmProject/mean-teacher/pytorch/mean_teacher/run_context.py", line 45, in _record self.save() File "/home/lts/PycharmProject/mean-teacher/pytorch/mean_teacher/run_context.py", line 38, in save df.to_msgpack(self.log_file_path, compress='zlib') File "/home/lts/.conda/envs/PCL/lib/python3.6/site-packages/pandas/core/generic.py", line 1320, in to_msgpack kwargs) File "/home/lts/.conda/envs/PCL/lib/python3.6/site-packages/pandas/io/packers.py", line 154, in to_msgpack writer(fh) File "/home/lts/.conda/envs/PCL/lib/python3.6/site-packages/pandas/io/packers.py", line 150, in writer fh.write(pack(a, **kwargs)) File "/home/lts/.conda/envs/PCL/lib/python3.6/site-packages/pandas/io/packers.py", line 691, in pack use_bin_type=use_bin_type).pack(o) File "pandas/io/msgpack/_packer.pyx", line 230, in pandas.io.msgpack._packer.Packer.pack (pandas/io/msgpack/_packer.cpp:3642) File "pandas/io/msgpack/_packer.pyx", line 232, in pandas.io.msgpack._packer.Packer.pack (pandas/io/msgpack/_packer.cpp:3484) File "pandas/io/msgpack/_packer.pyx", line 191, in pandas.io.msgpack._packer.Packer._pack (pandas/io/msgpack/_packer.cpp:2605) File "pandas/io/msgpack/_packer.pyx", line 220, in pandas.io.msgpack._packer.Packer._pack (pandas/io/msgpack/_packer.cpp:3178) File "pandas/io/msgpack/_packer.pyx", line 191, in pandas.io.msgpack._packer.Packer._pack (pandas/io/msgpack/_packer.cpp:2605) File "pandas/io/msgpack/_packer.pyx", line 220, in pandas.io.msgpack._packer.Packer._pack (pandas/io/msgpack/_packer.cpp:3178) File "pandas/io/msgpack/_packer.pyx", line 227, in pandas.io.msgpack._packer.Packer._pack (pandas/io/msgpack/_packer.cpp:3348) TypeError: can't serialize tensor(62, device='cuda:0')

Process finished with exit code 1

[ghost]

That's because the labeled_minibatch_size is a Tensor object. Using the codes in https://github.com/bl0/mean-teacher/blob/master/pytorch/main.py should solve this problem.

[panzhengo1]

I got the same problem, do you know how to solve it yet?

[panzhengo1]

That's because the labeled_minibatch_size is a Tensor object. Using the codes in https://github.com/bl0/mean-teacher/blob/master/pytorch/main.py should solve this problem. The link is 404, do you have another one?

[NamlessM]

That's because the labeled_minibatch_size is a Tensor object. Using the codes in https://github.com/bl0/mean-teacher/blob/master/pytorch/main.py should solve this problem. The link is 404, do you have another one?

You could just change the code in main function. Use .item() method to change the tensor into int

[yingqing0317]

change “labeled_minibatch_size = target_var.data.ne(NO_LABEL).sum()” to “labeled_minibatch_size = target_var.data.ne(NO_LABEL).sum().item()”