lvzhihan commented 6 years ago

Windows8.1, Python0.4.1, CUDA8.0 The same issue happened in the latest version Ubuntu

F:\appearance\DeblurGAN-master>python test.py --dataroot F:\appearance\DeblurGAN -master\myImages --model test --dataset_mode single --learn_residual ------------ Options ------------- aspect_ratio: 1.0 batchSize: 1 checkpoints_dir: ./checkpoints dataroot: F:\appearance\DeblurGAN-master\myImages dataset_mode: single display_id: 1 display_port: 8097 display_single_pane_ncols: 0 display_winsize: 256 fineSize: 256 gan_type: wgan-gp gpu_ids: [0] how_many: 5000 input_nc: 3 isTrain: False learn_residual: True loadSizeX: 640 loadSizeY: 360 max_dataset_size: inf model: test nThreads: 2 n_layers_D: 3 name: experiment_name ndf: 64 ngf: 64 no_dropout: False no_flip: False norm: instance ntest: inf output_nc: 3 phase: test resize_or_crop: resize_and_crop results_dir: ./results/ serial_batches: False which_direction: AtoB which_epoch: latest which_model_netD: basic which_model_netG: resnet_9blocks -------------- End ---------------- CustomDatasetDataLoader dataset [SingleImageDataset] was created E:\Anaconda3\lib\site-packages\torchvision\transforms\transforms.py:188: UserWar ning: The use of the transforms.Scale transform is deprecated, please use transf orms.Resize instead. "please use transforms.Resize instead.") Traceback (most recent call last): File "test.py", line 21, in model = create_model(opt) File "F:\appearance\DeblurGAN-master\models\models.py", line 11, in create_mod el model.initialize(opt) File "F:\appearance\DeblurGAN-master\models\test_model.py", line 21, in initia lize self.load_network(self.netG, 'G', which_epoch) File "F:\appearance\DeblurGAN-master\models\base_model.py", line 53, in load_n etwork network.load_state_dict(torch.load(save_path)) File "E:\Anaconda3\lib\site-packages\torch\nn\modules\module.py", line 721, in load_state_dict self.class.name, "\n\t".join(error_msgs))) RuntimeError: Error(s) in loading state_dict for ResnetGenerator: Unexpected running stats buffer(s) "model.2.running_mean" and "model.2.r unning_var" for InstanceNorm2d with track_running_stats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expected because InstanceNorm2d doe s not track running stats by default since 0.4.0. Please remove these keys from state_dict. If the running stats are actually needed, instead set trackrunning stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNo rm2d for details. Unexpected running stats buffer(s) "model.5.running_mean" and "model.5.r unning_var" for InstanceNorm2d with track_running_stats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expected because InstanceNorm2d doe s not track running stats by default since 0.4.0. Please remove these keys from state_dict. If the running stats are actually needed, instead set trackrunning stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNo rm2d for details. Unexpected running stats buffer(s) "model.8.running_mean" and "model.8.r unning_var" for InstanceNorm2d with track_running_stats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expected because InstanceNorm2d doe s not track running stats by default since 0.4.0. Please remove these keys from state_dict. If the running stats are actually needed, instead set trackrunning stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNo rm2d for details. Unexpected running stats buffer(s) "model.10.conv_block.2.running_mean" and "model.10.conv_block.2.running_var" for InstanceNorm2d with track_running_st ats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expecte d because InstanceNorm2d does not track running stats by default since 0.4.0. Pl ease remove these keys from state_dict. If the running stats are actually needed , instead set track_running_stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNorm2d for details. Unexpected running stats buffer(s) "model.10.conv_block.7.running_mean" and "model.10.conv_block.7.running_var" for InstanceNorm2d with track_running_st ats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expecte d because InstanceNorm2d does not track running stats by default since 0.4.0. Pl ease remove these keys from state_dict. If the running stats are actually needed , instead set track_running_stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNorm2d for details. Unexpected running stats buffer(s) "model.11.conv_block.2.running_mean" and "model.11.conv_block.2.running_var" for InstanceNorm2d with track_running_st ats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expecte d because InstanceNorm2d does not track running stats by default since 0.4.0. Pl ease remove these keys from state_dict. If the running stats are actually needed , instead set track_running_stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNorm2d for details. Unexpected running stats buffer(s) "model.11.conv_block.7.running_mean" and "model.11.conv_block.7.running_var" for InstanceNorm2d with track_running_st ats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expecte d because InstanceNorm2d does not track running stats by default since 0.4.0. Pl ease remove these keys from state_dict. If the running stats are actually needed , instead set track_running_stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNorm2d for details. Unexpected running stats buffer(s) "model.12.conv_block.2.running_mean" and "model.12.conv_block.2.running_var" for InstanceNorm2d with track_running_st ats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expecte d because InstanceNorm2d does not track running stats by default since 0.4.0. Pl ease remove these keys from state_dict. If the running stats are actually needed , instead set track_running_stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNorm2d for details. Unexpected running stats buffer(s) "model.12.conv_block.7.running_mean" and "model.12.conv_block.7.running_var" for InstanceNorm2d with track_running_st ats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expecte d because InstanceNorm2d does not track running stats by default since 0.4.0. Pl ease remove these keys from state_dict. If the running stats are actually needed , instead set track_running_stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNorm2d for details. Unexpected running stats buffer(s) "model.13.conv_block.2.running_mean" and "model.13.conv_block.2.running_var" for InstanceNorm2d with track_running_st ats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expecte d because InstanceNorm2d does not track running stats by default since 0.4.0. Pl ease remove these keys from state_dict. If the running stats are actually needed , instead set track_running_stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNorm2d for details. Unexpected running stats buffer(s) "model.13.conv_block.7.running_mean" and "model.13.conv_block.7.running_var" for InstanceNorm2d with track_running_st ats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expecte d because InstanceNorm2d does not track running stats by default since 0.4.0. Pl ease remove these keys from state_dict. If the running stats are actually needed , instead set track_running_stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNorm2d for details. Unexpected running stats buffer(s) "model.14.conv_block.2.running_mean" and "model.14.conv_block.2.running_var" for InstanceNorm2d with track_running_st ats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expecte d because InstanceNorm2d does not track running stats by default since 0.4.0. Pl ease remove these keys from state_dict. If the running stats are actually needed , instead set track_running_stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNorm2d for details. Unexpected running stats buffer(s) "model.14.conv_block.7.running_mean" and "model.14.conv_block.7.running_var" for InstanceNorm2d with track_running_st ats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expecte d because InstanceNorm2d does not track running stats by default since 0.4.0. Pl ease remove these keys from state_dict. If the running stats are actually needed , instead set track_running_stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNorm2d for details. Unexpected running stats buffer(s) "model.15.conv_block.2.running_mean" and "model.15.conv_block.2.running_var" for InstanceNorm2d with track_running_st ats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expecte d because InstanceNorm2d does not track running stats by default since 0.4.0. Pl ease remove these keys from state_dict. If the running stats are actually needed , instead set track_running_stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNorm2d for details. Unexpected running stats buffer(s) "model.15.conv_block.7.running_mean" and "model.15.conv_block.7.running_var" for InstanceNorm2d with track_running_st ats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expecte d because InstanceNorm2d does not track running stats by default since 0.4.0. Pl ease remove these keys from state_dict. If the running stats are actually needed , instead set track_running_stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNorm2d for details. Unexpected running stats buffer(s) "model.16.conv_block.2.running_mean" and "model.16.conv_block.2.running_var" for InstanceNorm2d with track_running_st ats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expecte d because InstanceNorm2d does not track running stats by default since 0.4.0. Pl ease remove these keys from state_dict. If the running stats are actually needed , instead set track_running_stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNorm2d for details. Unexpected running stats buffer(s) "model.16.conv_block.7.running_mean" and "model.16.conv_block.7.running_var" for InstanceNorm2d with track_running_st ats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expecte d because InstanceNorm2d does not track running stats by default since 0.4.0. Pl ease remove these keys from state_dict. If the running stats are actually needed , instead set track_running_stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNorm2d for details. Unexpected running stats buffer(s) "model.17.conv_block.2.running_mean" and "model.17.conv_block.2.running_var" for InstanceNorm2d with track_running_st ats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expecte d because InstanceNorm2d does not track running stats by default since 0.4.0. Pl ease remove these keys from state_dict. If the running stats are actually needed , instead set track_running_stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNorm2d for details. Unexpected running stats buffer(s) "model.17.conv_block.7.running_mean" and "model.17.conv_block.7.running_var" for InstanceNorm2d with track_running_st ats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expecte d because InstanceNorm2d does not track running stats by default since 0.4.0. Pl ease remove these keys from state_dict. If the running stats are actually needed , instead set track_running_stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNorm2d for details. Unexpected running stats buffer(s) "model.18.conv_block.2.running_mean" and "model.18.conv_block.2.running_var" for InstanceNorm2d with track_running_st ats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expecte d because InstanceNorm2d does not track running stats by default since 0.4.0. Pl ease remove these keys from state_dict. If the running stats are actually needed , instead set track_running_stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNorm2d for details. Unexpected running stats buffer(s) "model.18.conv_block.7.running_mean" and "model.18.conv_block.7.running_var" for InstanceNorm2d with track_running_st ats=False. If state_dict is a checkpoint saved before 0.4.0, this may be expecte d because InstanceNorm2d does not track running stats by default since 0.4.0. Pl ease remove these keys from state_dict. If the running stats are actually needed , instead set track_running_stats=True in InstanceNorm2d to enable them. See the documentation of InstanceNorm2d for details. Unexpected running stats buffer(s) "model.20.running_mean" and "model.20 .running_var" for InstanceNorm2d with track_running_stats=False. If state_dict i s a checkpoint saved before 0.4.0, this may be expected because InstanceNorm2d d oes not track running stats by default since 0.4.0. Please remove these keys fro m state_dict. If the running stats are actually needed, instead set track_runnin g_stats=True in InstanceNorm2d to enable them. See the documentation of Instance Norm2d for details. Unexpected running stats buffer(s) "model.23.running_mean" and "model.23 .running_var" for InstanceNorm2d with track_running_stats=False. If state_dict i s a checkpoint saved before 0.4.0, this may be expected because InstanceNorm2d d oes not track running stats by default since 0.4.0. Please remove these keys fro m state_dict. If the running stats are actually needed, instead set track_runnin g_stats=True in InstanceNorm2d to enable them. See the documentation of Instance Norm2d for details.

lvzhihan commented 6 years ago

Work on Python0.3.0 and latest Ubuntu. The problem is the version problem of pytorch.

paulpoloniewicz commented 6 years ago

I have the same problem, but I am also trying to run in non-cuda mode, and if I install python < 3.6, it has a problem with pytorch-cpu. Ivzhihan, can you tell me the packages and versions I need to have installed to run non-cuda?

lvzhihan commented 6 years ago

paulpoloniewicz, glad to know you are using nonCUDA system, the same as me. You need to remove or comment all of 'cuda' in all of the scripts. Since you have the same problem as me, you should already detour the GPU problem.

lvzhihan commented 6 years ago

paulpoloniewicz, you should uninstall pytorch 0.4 and then install pytorch 0.3.0 as the command below. And then you can solve this problem. conda install pytorch=0.3.0 -c soumith

paulpoloniewicz commented 6 years ago

Ivzhihan, thank you for the quick response. I currently have pytorch 0.3.1 installed: pytorch: 0.3.1-py36_cuda8.0.61_cudnn7.0.5_2 soumith Definitely needs 0.3.0?

paulpoloniewicz commented 6 years ago

Not sure I understand commenting out all mention of cuda in the scripts. Shouldn't setting --gpu_ids -1 take care of that?

paulpoloniewicz commented 6 years ago

Switched to pytorch 0.3.0 and now get:

ubi@ubi-VirtualBox:~/Dev/DeblurGAN-master$ python test.py --dataroot ./images --model test --dataset_mode single --learn_residual --gpu_ids -1 Traceback (most recent call last): File "test.py", line 3, in from options.test_options import TestOptions File "/home/ubi/Dev/DeblurGAN-master/options/test_options.py", line 1, in from .base_options import BaseOptions File "/home/ubi/Dev/DeblurGAN-master/options/base_options.py", line 3, in from util import util File "/home/ubi/Dev/DeblurGAN-master/util/util.py", line 2, in import torch File "/home/ubi/anaconda3/lib/python3.6/site-packages/torch/init.py", line 315, in import torch.distributions File "/home/ubi/anaconda3/lib/python3.6/site-packages/torch/distributions/init.py", line 77, in from .categorical import Categorical File "/home/ubi/anaconda3/lib/python3.6/site-packages/torch/distributions/categorical.py", line 2, in from torch._six import nan ImportError: cannot import name 'nan'

lvzhihan commented 6 years ago

paulpoloniewicz, I didn't meet this problem. I guess it should install 'nan' by pip install nan

paulpoloniewicz commented 6 years ago

Ivzhihan, can you please send me the output of 'conda list' from your system?

lvzhihan commented 6 years ago

lvzhihan@ubuntu:~/Documents/DeblurGAN-master$ conda list

packages in environment at /home/lvzhihan/anaconda3:

#

asanakoy commented 5 years ago

Try conda uninstall pytorch and then conda install pytorch torchvision -c pytorch.

opentld commented 5 years ago

Work on Python0.3.0 and latest Ubuntu. The problem is the version problem of pytorch.

does pytorch1.0 work?

zzh928253398 commented 5 years ago

my torch version is 0.4.0, python3.6 cuda 9.0 in the file models/test_model.py line 54
def load_network(): added strict=False attribute to load_state_dict(). change network.load_state_dict(torch.load(save_path)) to network.load_state_dict(torch.load(save_path), strict=False) it works for me.

KupynOrest / DeblurGAN

load_network issue #94

packages in environment at /home/lvzhihan/anaconda3:

Name Version Build Channel